)]}'
{
  "log": [
    {
      "commit": "c58e8996185ab5b1499d34889c476d548899ef5a",
      "tree": "c97c6b8576d7d293ca877d998665cbc8b3696515",
      "parents": [
        "09895ff783f4b7a52121e575150fe1b97ea42079"
      ],
      "author": {
        "name": "Edward Lee",
        "email": "edwarddl@google.com",
        "time": "Fri Nov 20 18:22:32 2026 +0000"
      },
      "committer": {
        "name": "Edward Lee",
        "email": "edwarddl@google.com",
        "time": "Fri Nov 20 17:09:07 2026 -0800"
      },
      "message": "entity-manager: Backport EEPROM address size detection modes\n\nfru-device decides whether an EEPROM needs 1 or 2 address bytes in\nisDevice16Bit(): it writes a single 0x00 byte and then reads 8 bytes,\nexpecting a 2-byte EEPROM to return different data. What the write sets the address pointer to is undefined if we only send 1 bit. This issue has been solved upstream in openBMC via a new mechanism to check if a EEPROM is 8 bit/16 bit. (https://github.com/openbmc/entity-manager/commit/dac2dfc3a64063b062fabd9c4d6e2857c1185eaf)\n\nBackport the upstream entity-manager change that adds the\nfru-device-16bitdetectmode meson option:\n- MODE_1 (default): current behavior.\n- MODE_2: for offsets 0..7, write a 2-byte offset and read 1 byte after\n  a repeated START. A 2-byte EEPROM always decodes this as a new\n  address, independent of the previous pointer.\n\nisDevice16Bit() now takes the target address, so downstream patches that\ntouch its call site are rebased under the same topic.\n\nTested:\n- bitbake entity-manager passes with no patch fuzz on 16\n  machines covering every layer that patches fru_device.cpp.\n- With MODE_2 enabled, the previously missing 2-byte FRU is decoded on\n  every rescan on a real machine, and 8-bit FRU EEPROM contents are\n  unchanged after repeated scans.\n- Tested on other configs to ensure that we have no fru device population issues (platform18)\n\nfusion2: N/A\n\nGoogle-Bug-Id: 587128916\nGoogle-Bug-Id: 571164059\nChange-Id: If4c4386c4787d0b9105b419ff4c683a8d628417c\nSigned-off-by: Edward Lee \u003cedwarddl@google.com\u003e\n"
    },
    {
      "commit": "09895ff783f4b7a52121e575150fe1b97ea42079",
      "tree": "d8e4f8b3a1cc97f6e7ac7b0cadf846c5a1accffd",
      "parents": [
        "00efdce80c3d7cdbc5d547e52f3e68a99d3a09f1"
      ],
      "author": {
        "name": "Hamza Sheikh",
        "email": "hamzasheikh@google.com",
        "time": "Fri Nov 20 14:27:13 2026 +0000"
      },
      "committer": {
        "name": "Hamza Sheikh",
        "email": "hamzasheikh@google.com",
        "time": "Fri Nov 20 13:13:06 2026 -0800"
      },
      "message": "gbmcweb master: SRCREV bump 72a06d099e..18e87222a7\n\nNan Zhou (1):\n      Validate syslog target IP in CredentialManager\n\ngBMC Team (2):\n      Decouple tlBMC EntityConfig headers from collector and HTTP libraries.\n      tlbmc: Support per-channel PGA configuration for ADC sensors\n\nFusion-Link: fusion2 N/A\nTested: N/A\nGoogle-Bug-Id: 570170500\nSigned-off-by: Hamza Sheikh \u003chamzasheikh@google.com\u003e\nChange-Id: I9cd119b0dc6343b75bfb44b41fb2736c93da3d24\n"
    },
    {
      "commit": "00efdce80c3d7cdbc5d547e52f3e68a99d3a09f1",
      "tree": "f8804feab11587e10a76289b24369e151e22ec32",
      "parents": [
        "3bc1297a13be56bb3f6d3294f2566d247b409433"
      ],
      "author": {
        "name": "Ryan Wang",
        "email": "ryantwang@google.com",
        "time": "Thu Nov 19 23:19:36 2026 +0000"
      },
      "committer": {
        "name": "Yuxiao Zhang",
        "email": "yuxiaozhang@google.com",
        "time": "Fri Nov 20 10:52:04 2026 -0800"
      },
      "message": "emmc-image: create mutable upper dirs before merge\n\nsystemd-sysext/confext --mutable\u003dauto only make /usr and /etc writable\nif /run/initramfs/rw/cow/usr and /run/initramfs/rw/cow/etc exist. After\na fresh install nothing has written under /usr yet, so the merged /usr\ncomes up read-only. Create both directories before the merge.\n\nTested: On a freshly installed platform23 26.46.6.0 BMC, cow/usr was\nmissing and /usr was ro. Creating it and running\n`systemd-sysext refresh --force --mutable\u003dauto` made /usr writable.\nFusion-Link: http://fusion2/d1dc1be8-a3bb-3ccc-9325-45747cafca5a\nGoogle-Bug-Id: 581023604\nChange-Id: Ic7bdedeeed44ec776bcc98b68f3cf7ef48a7f3ae\nSigned-off-by: Ryan Wang \u003cryantwang@google.com\u003e\n"
    },
    {
      "commit": "3bc1297a13be56bb3f6d3294f2566d247b409433",
      "tree": "5fd251ca55d97f9c72048fb36c86d9e60edd1bc1",
      "parents": [
        "796c32c78e0c1487163b255c219602201902bb69"
      ],
      "author": {
        "name": "Yuli Fiterman",
        "email": "fiterman@google.com",
        "time": "Tue Nov 17 03:22:37 2026 +0000"
      },
      "committer": {
        "name": "Yuli Fiterman",
        "email": "fiterman@google.com",
        "time": "Thu Nov 19 17:52:22 2026 -0800"
      },
      "message": "entity-manager: Fix OOB reads in FRU area parsing\n\nfru-device still parses FRU EEPROMs that fail validateFRUAreas(), so\nthat FRUs with a bad area checksum keep being published. Those parsers\ntrusted the area offsets and length bytes in the data:\n\n- validateFRUAreas() checked the area length against the buffer but\n  formatIPMIFRU() did not, and in both a zero length made the checksum\n  walk run past the end.\n- parseMultirecordOem() and parseMultirecordUUID() dereferenced the\n  MultiRecord offset before checking it.\n- findFruAreaLocationAndField() and copyRestFRUArea() indexed past the\n  buffer and followed field lengths out of the area. updateFRUProperty()\n  sizes its writes from the locations they return, so it then wrote\n  past the area or past the end of the vector.\n\nAdd a downstream patch that moves the per-area header checks into one\nhelper shared by validateFRUAreas() and formatIPMIFRU(), does the same\nfor the MultiRecord offset, and keeps the property-update field walks\ninside the area.\n\nFRUs with a bad area checksum are still parsed and published. A\nzero-length area, or one that extends past the bytes read from the\nEEPROM, is now rejected. A property update is refused unless the field\nand the end-of-fields marker are both found before the area checksum.\n\nTested:\n- Built src/fru_utils.cpp before and after the patch with ASan and\n  UBSan on x86_64 and ran both over 4,567 truncated and byte-mutated\n  FRU images through validate, format and the update helpers (13,701\n  runs): 1,591 sanitizer errors before, 0 after.\n- On every input the old code handled without a memory error,\n  validateFRUAreas() and formatIPMIFRU() return the same results.\n- Property writes: took four FRU images, set each byte to each of the\n  256 values, and ran readFRUContents(), formatIPMIFRU() and the body\n  of updateFRUProperty() (also with the platform patch that makes the\n  Rework ID writable) for writes of 2, 8 and 63 characters, 1,036,800\n  runs per build: 11,886 sanitizer errors before, 0 after. Results on\n  the unmodified images are the same. The only writes that used to\n  complete and are now refused (611) are on areas whose fields do not\n  end with the end-of-fields marker.\n- Ran the bad-checksum EEPROM dump from b/492137013 through both\n  builds: still parsed with a warning and all fields returned.\n- bitbake entity-manager for one armv7 and one aarch64 machine with\n  this patch on meta-gbmc-staging 5e6707d5.\n- Applied the patch series from this bbappend to 16c0ba4e with\n  patch -F0; src/fru_utils.cpp matches the tested file.\n\nFusion-Link: fusion2 N/A\nGoogle-Bug-Id: 552515924\nGoogle-Bug-Id: 552515652\nChange-Id: I0171ee65a29c22c8a853812545c3dc218645ecd3\nSigned-off-by: Yuli Fiterman \u003cfiterman@google.com\u003e\n"
    },
    {
      "commit": "796c32c78e0c1487163b255c219602201902bb69",
      "tree": "3414c333b9db07f630707224b98a080c69dc46da",
      "parents": [
        "22131d6082f54c80df9622aef088006d2b5ef331"
      ],
      "author": {
        "name": "Bart Van Assche",
        "email": "bvanassche@google.com",
        "time": "Wed Nov 18 14:24:00 2026 -0800"
      },
      "committer": {
        "name": "Bart Van Assche",
        "email": "bvanassche@google.com",
        "time": "Thu Nov 19 17:40:44 2026 -0800"
      },
      "message": "linux-gbmc: Remove C0 network card PACKAGECONFIG\n\nRemove the C0 network card workaround PACKAGECONFIG flag, the platform\nsupport variables and check, and the conditional npcm7xx SRC_URI entry\nfrom linux-gbmc.inc because no machine enables this option anymore and\nthe C0 network card workaround has been reverted on the gbmc-5.15\nkernel branch. Also remove the C0 network card workaround patch file\nsince this change removes the only reference to that patch file.\n\nTested: Mimik + compared source trees with diff -ru\nGoogle-Bug-Id: 584826527\nChange-Id: I071ea2159b4bad67d3c9fbcb6fd731e1b89bda26\nSigned-off-by: Bart Van Assche \u003cbvanassche@google.com\u003e\n"
    },
    {
      "commit": "22131d6082f54c80df9622aef088006d2b5ef331",
      "tree": "860d833b285eb34523f17a756cdbecf41f0ab173",
      "parents": [
        "0d2bb2e16d34bf3f98407a27805ba0e699327597"
      ],
      "author": {
        "name": "George Kola",
        "email": "georgekola@google.com",
        "time": "Thu Nov 19 22:10:48 2026 +0000"
      },
      "committer": {
        "name": "George Kola",
        "email": "georgekola@google.com",
        "time": "Thu Nov 19 15:26:48 2026 -0800"
      },
      "message": "faultlog: Fix CPER UAF, host filtering, and ASan tests\n\nHarden the existing agentless FaultLog patches (0001, 0003, 0004, 0006)\nin place so downstream layers that strip 0001-0004 and 0006-0008 keep\nclean do_patch compatibility:\n\n- Replace unchecked dynamic_cast and dangling std::string_view bindings\n  (CWE-416) across createDump, saveEarliestEntry,\n  deleteEntryByTypeAndPrimaryLogId, deleteEntryByTypePrimaryId, and\n  registerCrashdumpDeleteAllMatch, and delegate\n  deleteEntryByTypePrimaryId across saved CPER and Crashdump maps.\n- Check msg.get_path() for null in createDumpFromCrashdumpSignal, add\n  missing return on is_method_error(), and wrap noexcept sd-bus match\n  callbacks in try/catch across registerCperLogMatch, crashdump\n  listeners, and secondary host CPER listeners.\n- Filter decoded secondary CPER object paths by hostId in multi-host\n  mode while preserving root ObjectManager path_namespace compatibility.\n- Use dumpDir in Manager::createDump, pass FAULTLOG_DUMP_PATH instead\n  of FAULTLOG_DUMP_OBJ_ENTRY in single-host dump_manager_main.cpp, wire\n  dump_types and per-PID FAULTLOG_DUMP_PATH / CRASHDUMP_FORMAT\n  overrides into test/meson.build, and add hermetic 1,600-CPER\n  ObjectMapper + ASan/LSan Intel and AMD crashdump unit tests.\n\nBuilt and ran faultlog_dump_test (7/7) and faultlog_amd_crashdump_test\n(2/2) under ASan+LSan over dbus-run-session with 0 ObjectMapper leaks\nand 0 ASan/LSan errors, and verified patch --fuzz\u003d0 across 0001-0008.\n\nTested: Ran faultlog_dump_test and faultlog_amd_crashdump_test (ASan)\nFusion-Link: fusion2 N/A\nGoogle-Bug-Id: 587980888\nGoogle-Bug-Id: 578037346\nGoogle-Bug-Id: 570097398\nGoogle-Bug-Id: 583984589\nSigned-off-by: George Kola \u003cgeorgekola@google.com\u003e\nChange-Id: I2b9983efb00ece35b605b0deaa48538325a7406e\n"
    },
    {
      "commit": "0d2bb2e16d34bf3f98407a27805ba0e699327597",
      "tree": "ab8d952bd079cf45321a46a3b1d2601437acca0f",
      "parents": [
        "5469ed9135f2d7e5f371f95c525abdbcad4eee2a"
      ],
      "author": {
        "name": "Luka Strizic",
        "email": "lstrz@google.com",
        "time": "Tue Nov 03 21:01:08 2026 +0000"
      },
      "committer": {
        "name": "Luka Strižić",
        "email": "lstrz@google.com",
        "time": "Thu Nov 19 10:29:09 2026 -0800"
      },
      "message": "mctp: Retry EID pool allocation on timeout\n\nendpoint_allocate_eids() sends Allocate Endpoint IDs (0x08) only once\nwith a 250 ms timeout, whereas query_peer_properties() retries up to 4\nattempts on -ETIMEDOUT. During BMC boot or OOB firmware update\ncontention on USB MCTP bridges, a single transient delay exceeding\n250 ms aborts downstream EID pool allocation.\n\nRetry endpoint_send_allocate_endpoint_ids() up to 4 times on -ETIMEDOUT\nwhile reusing the same request IID across retransmissions per DSP0237 so\ndelayed replies from an earlier attempt are still accepted and repeat\nallocation requests are idempotent.\n\nTested: Deployed full 5-patch mctpd + mctpreactor stack on hardware.\nVerified clean peer cleanup and auto-recovery on transient bridge\ntimeout, atomic Bridge1 pool publication before InterfacesAdded, live\nEndpoint1.Remove recovery, 5/5 cold-start stress cycles, and\nPLDM/Redfish discovery across both USB bridges and all 4 downstream\nbridged NICs: https://paste.googleplex.com/6755399436596849\n\nFusion-Link: fusion2 N/A\nGoogle-Bug-Id: 508310010\nChange-Id: I5297323f3e50e8e840ba2c28f571f28407f2ae88\nSigned-off-by: Luka Strizic \u003clstrz@google.com\u003e\n"
    },
    {
      "commit": "5469ed9135f2d7e5f371f95c525abdbcad4eee2a",
      "tree": "0fd322810e0c8be198c2387575f5732cd4516a67",
      "parents": [
        "ee309dc76953bf5acb8107a0eea97d29fd424a55"
      ],
      "author": {
        "name": "Luka Strizic",
        "email": "lstrz@google.com",
        "time": "Mon Nov 09 23:20:14 2026 +0000"
      },
      "committer": {
        "name": "Luka Strižić",
        "email": "lstrz@google.com",
        "time": "Thu Nov 19 10:28:56 2026 -0800"
      },
      "message": "mctp: Discard stale out-of-order responses\n\nWhen a control request times out after 250 ms and closes its socket, the\nnext request can reuse the kernel MCTP tag and receive the delayed reply\nfrom the timed-out request, aborting the active request instead of\nwaiting for its real response.\n\nDrain stale control and mismatched-type packets within a single\nmctp_timeout window on the receive socket so late responses from earlier\ntimed-out requests are ignored rather than failing the active request.\n\nTested: Deployed full mctpd + mctpreactor stack on hardware. Verified\nclean peer cleanup and auto-recovery on transient bridge timeout, atomic\nBridge1 pool publication before InterfacesAdded, live Endpoint1.Remove\nrecovery, 5/5 cold-start stress cycles, and PLDM/Redfish discovery\nacross both USB bridges and all 4 downstream bridged NICs:\nhttps://paste.googleplex.com/6755399436596849\n\nFusion-Link: fusion2 N/A\nGoogle-Bug-Id: 508310010\nChange-Id: I88acf842ce1c7d217371aa55814132052f697d3e\nSigned-off-by: Luka Strizic \u003clstrz@google.com\u003e\n"
    },
    {
      "commit": "ee309dc76953bf5acb8107a0eea97d29fd424a55",
      "tree": "8c5f4873595e8ae4015a196f0711f1f862663707",
      "parents": [
        "1704903bdbba11ea08b44cc1bfcc26adecbc2775"
      ],
      "author": {
        "name": "Rudransh Agrawal",
        "email": "rudyagrawal@google.com",
        "time": "Wed Nov 18 23:27:47 2026 +0000"
      },
      "committer": {
        "name": "Rudransh Agrawal",
        "email": "rudyagrawal@google.com",
        "time": "Thu Nov 19 00:54:20 2026 +0000"
      },
      "message": "emmc-image: stage verity image in /run tmpfs\n\nOpening the dm-verity image from /mnt/luks-mmcblk0_fs keeps the eMMC\nbusy while any daemon holds the /usr or /etc overlay. Stopping with\nclear_mount_ref and sleep 120 also restarts active daemons (including\nsshd) in ExecStopPost, misses pre-merge openers, and can stall until\nTimeoutStopSec.\n\nCopy the verity image to /run/emmc-image tmpfs before veritysetup open\n(skipping the copy if veritysetup close --cancel-deferred succeeds on\nan existing mapping, and failing without eMMC fallback if cp or\nveritysetup fails) and unlink it right after. Unlike reclaimable eMMC\npage cache, this copy uses unreclaimable tmpfs memory for the full\nuptime, and after stop it stays pinned until the last process holding\nthe old /usr or /etc exits (potentially until reboot). Delete\nclear_mount_ref and sleep 120 so stop never restarts daemons, and exit\n0 at the end of stop. Only two high-DRAM configs enable\ngbmc-image-on-emmc (1.55 to 3.12 MiB, under 1% of /run tmpfs), so zero\nlow-DRAM BMCs run this script.\n\nTested: Live platform17 jtmy2-nfd01 (dev 26.47.201.0) with /usr holder:\n```\nstop_s: 134s -\u003e 3.56s (3.40-3.64s across 5 cycles)\nrestarts: msvfud+6+sshd killed -\u003e 0 (all PIDs unchanged)\numount /mnt/luks-mmcblk0_fs: rc\u003d32 busy -\u003e rc\u003d0 (dm-1 Open\u003d1)\nautoclear close: /run 9128 -\u003e 7544 KiB (-1584 KiB exact)\nGuitar E2E (4/4 PASSED): emmc_erase, coredumps, psdmd 3/3, diskerase 9/9\n```\n\nFusion-Link: http://fusion2/eb09fd58-42d9-3356-b665-ac053bc140c9\nGoogle-Bug-Id: 548837432\nGoogle-Bug-Id: 582658313\nChange-Id: Ia27c81ddca3072e08080ed4c0d367bee22f87bc4\nSigned-off-by: Rudransh Agrawal \u003crudyagrawal@google.com\u003e\n"
    },
    {
      "commit": "1704903bdbba11ea08b44cc1bfcc26adecbc2775",
      "tree": "68c26f9547cf947ceadf22e7c73d3233a718380d",
      "parents": [
        "3f799f41f1f2b99725146c4e8ff289e09012875b"
      ],
      "author": {
        "name": "Ryan Wang",
        "email": "ryantwang@google.com",
        "time": "Wed Nov 18 19:19:16 2026 +0000"
      },
      "committer": {
        "name": "Ryan Wang",
        "email": "ryantwang@google.com",
        "time": "Wed Nov 18 19:21:40 2026 +0000"
      },
      "message": "g3-shared-libs master: SRCREV bump b055c05286..f037f9a399\n\ngBMC Team (7):\n      BEGIN_PUBLIC Expose kRebootStart checkpoint constant in boottime_api.\n      Make gBMC settings writes atomic to survive power loss and write errors.\n      Remove printproto from g3_shared_libs\n      Add a reported core count to gBMC BIOS settings.\n      Prevent SIGSEGV on null RedfishIterable and substitute predicate variables\n      Generalize some details in the codebase to avoid external confusion due to internal terms and links. These details do not change functionality.\n      Add insecure_skip_verify option to HttpCredential proto\n\nFusion-Link: fusion2 N/A\nTested: N/A\nGoogle-Bug-Id: 556304358\nChange-Id: I4c0bd36cd79409524e8f399a04eb97a8e76ce835\nSigned-off-by: Ryan Wang \u003cryantwang@google.com\u003e\n"
    },
    {
      "commit": "3f799f41f1f2b99725146c4e8ff289e09012875b",
      "tree": "e86750a99e3e38c861cb1005986122aa4a454f82",
      "parents": [
        "082dae248a00240dfb39b52bd8338393660b8b50"
      ],
      "author": {
        "name": "Munawar Hussain",
        "email": "munawarhussain@google.com",
        "time": "Tue Nov 17 23:37:59 2026 +0000"
      },
      "committer": {
        "name": "Munawar Hussain",
        "email": "munawarhussain@google.com",
        "time": "Tue Nov 17 20:36:21 2026 -0800"
      },
      "message": "nvmed: SRCREV bump f57557ee45..9af36a5698\n\nMunawar Hussain (1):\n      nvmesensor: fix fw commit needs-reset check\n\nTested: In each commit\n\nFusion-Link: fusion2 N/A\nGoogle-Bug-Id: 578681755\nChange-Id: I2f8d2b3040a017674beab05dbfd41cececacbd01\nSigned-off-by: Munawar Hussain \u003cmunawarhussain@google.com\u003e\n"
    },
    {
      "commit": "082dae248a00240dfb39b52bd8338393660b8b50",
      "tree": "e82af3318f5200c78e24a971bb73c373cf3e1c1a",
      "parents": [
        "dd001e835c196b390396d825d2c5e1cecf889c2d"
      ],
      "author": {
        "name": "Yuxiao Zhang",
        "email": "yuxiaozhang@google.com",
        "time": "Mon Nov 16 15:53:53 2026 -0800"
      },
      "committer": {
        "name": "Yuxiao Zhang",
        "email": "yuxiaozhang@google.com",
        "time": "Tue Nov 17 19:28:37 2026 -0800"
      },
      "message": "u-boot: Quiesce AST2700 ftgmac100 DMA on boot\n\nAcross a WDT0 SOC warm reset on AST2700 (even with MAC0-2 set in\nWDT_RST_MASK_3), IO-die FTGMAC100 controllers that were active in Linux\nremain out of reset (SCU1 modrst1_ctrl bit 0) with their prior\nMAC_MADR/MAC_LADR, MACCR, and RX DMA ring pointers intact. In the kdump\nreboot flow, the crash kernel disables unused MAC clocks (SCU1\nclkgate_ctrl1) without resetting the MAC, and U-Boot\u0027s ftgmac100_probe()\nlater ungates the clocks via ast2700_soc1_clk_enable() without pulsing\nSCU1 reset, resuming stale RX DMA into DRAM across kernel load (~1%\nsecondary kernel panic rate). AST2600 is unaffected because\nast2600_configure_mac() hard-resets the MAC whenever ungating its clock.\n\nAdd an AST2700 U-Boot board_init() patch to quiesce any out-of-reset\nFTGMAC100 instance right after ahbc_init(): temporarily ungate the MAC\nclock if gated, log pre-clear MACCR and SCU reset/clock state when\nactive, clear IER and MACCR, wait 2 ms to drain any in-flight frame,\nhold SCU1 MAC reset for 10 ms with the clock running, and restore\nMAC_MADR/MAC_LADR and any prior clock gate.\n\nTested:\n- Verified checkpatch.pl passes with 0 errors and 0 warnings, and the\n  aspeed-g7 SRC_URI patch stack applies with 0 fuzz.\n- Built, signed, and flashed 26.50.99.17 onto AST2700 BMC hardware;\n  verified cold boot (gbmc-psu-hardreset.target), warm reboot, and 5/5\n  consecutive sysrq -\u003e kdump -\u003e reboot cycles with kcore saved, \"Read\n  MAC address ... from chip\" preserved on both 14050000.ethernet and\n  14060000.ethernet, ncsi/front/gbmcbr UP, and zero kernel panics.\n\nFusion-Link: fusion2/N/A\nGoogle-Bug-Id: 544584165\nSigned-off-by: Yuxiao Zhang \u003cyuxiaozhang@google.com\u003e\nTAG\u003dagy\nCONV\u003de6cb2203-cbc6-4656-8044-8157fd27597b\nChange-Id: Ib2aa8ba03ca9adfb7a7b3e435160e1554d3ae9f6\n"
    },
    {
      "commit": "dd001e835c196b390396d825d2c5e1cecf889c2d",
      "tree": "4e8c676de9db9cc10f74805b9c240941f6a978e7",
      "parents": [
        "b0c04393b85fbb845f03a2be494e6dadaa62fdca"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Wed Nov 11 08:44:43 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Tue Nov 17 17:45:06 2026 -0800"
      },
      "message": "gbmc-usb-reset: Add USB hub reset shell library\n\nProvide /usr/share/gbmc-usb-reset/lib.sh so platforms stop open-coding\nUSB hub resets. It covers per-controller flock serialization, automatic\nehci-platform and npcm7xx-ehci driver resolution, signal-safe GPIO and\nEHCI restore, active-low and active-high libgpiod or sysfs GPIOs, and\nroot-port enumeration retries.\n\nHub reset GPIOs are passed as an ordered --gpio list: all are asserted\ntogether and then released one at a time in the order given, so any\ndepth of cascaded hubs is described by listing them from the root down.\nA \":optional\" suffix skips lines that don\u0027t exist on unpopulated hubs.\nAll timeouts, poll counts, and retry counts are GBMC_USB_* variables at\nthe top of the library.\n\nTested:\n- bash -n and shellcheck -x passed; bash test.sh passed all 8 unit tests\n  covering sysfs resets, ready-GPIO waits, pre-unbound recovery,\n  active-high and multi-controller resets, 3-level release order,\n  optional GPIOs, invalid modifiers, enumeration timeout retries, and\n  SIGTERM cleanup.\n- BitBake do_compile passed across multiple AST2600 and NPCM machines.\n- Verified on live NPCM and AST2600 hardware across sequential resets,\n  concurrent flock, mid-reset SIGTERM, pre-unbound EHCI recovery, tray\n  AC powercycles, and warm BMC reboots with 0 failures (with the earlier\n  upstream/downstream GPIO options, which released hubs in the same\n  order).\n\nFusion-Link: fusion2 N/A\nGoogle-Bug-Id: 574912243\nChange-Id: I592e9ccbb084485d05f70776827474e65d4610a4\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "b0c04393b85fbb845f03a2be494e6dadaa62fdca",
      "tree": "b1ff9a477740ae14ab4bd54e77756f43950ab055",
      "parents": [
        "4e60eb9f2e4cf0088738474237649ffa731ec7cb"
      ],
      "author": {
        "name": "Jorge Cardona",
        "email": "jorgecardona@google.com",
        "time": "Fri Nov 13 09:05:29 2026 -0800"
      },
      "committer": {
        "name": "Jorge Cardona Livas",
        "email": "jorgecardona@google.com",
        "time": "Tue Nov 17 17:01:11 2026 -0800"
      },
      "message": "strowger master: SRCREV bump 3b5878dfd8..333a43423d\n\nJorge Cardona (1):\n      Fail server initialization if secure credentials cannot be\n      obtained.\n\ngBMC Team (1):\n      Fix 9 ClangInliner findings\n\nFusion-Link:\nplatform5: https://fusion2.corp.google.com/f49ff52a-eb9f-3bfa-9df5-588a946694e8\nplatform15: https://fusion2.corp.google.com/d36b3cff-cb91-32b5-8e4a-64f2ae2901bb\nplatform17: https://fusion2.corp.google.com/2620d806-86c0-3c0b-b3a7-5cb0e6f861b9\nplatform11+ext: https://fusion2.corp.google.com/cd425c14-50de-30ba-8de3-f02a8b080f53\nplatform11: https://fusion2.corp.google.com/26e99e07-86a4-371a-bcd3-3c8bf7e1f1a7\nGoogle-Bug-Id: 505525030\nTested: Ran presubmit tests\nChange-Id: Ibd4878fb7761ea5ec8d87df96169d10014a28a6f\nSigned-off-by: Jorge Cardona \u003cjorgecardona@google.com\u003e\n"
    },
    {
      "commit": "4e60eb9f2e4cf0088738474237649ffa731ec7cb",
      "tree": "15a69ea2454c083e43058dedb59d36246059e2e9",
      "parents": [
        "459dfa44d648c2404a440335977a8272339ed7f9"
      ],
      "author": {
        "name": "Mo Elbadry",
        "email": "elbadrym@google.com",
        "time": "Fri Nov 13 17:43:34 2026 +0000"
      },
      "committer": {
        "name": "Mo Elbadry",
        "email": "elbadrym@google.com",
        "time": "Tue Nov 17 08:50:48 2026 -0800"
      },
      "message": "gbmc-bsp-common: Check enabled DTB USB 1.1 HCDs\n\nReject unallowlisted enabled USB 1.1 companion controllers (`uhci` /\n`ohci`, `COMPANION_USB11_HCD_ENABLED`) in base `.dtb` trees, wire\n`GBMC_ALLOWED_USB11_NODES` (`--allow-usb11-node`) in\n`linux-common.inc`, and warn on unused allowlist entries\n(`UNUSED_USB11_ALLOWLIST`).\n\nFusion-Link: fusion2 N/A\nTested: GbmcDtbValidatorTester.py (9/9 unit tests pass).\nGoogle-Bug-Id: 579089739\nChange-Id: I5364758617039485762314059687231405968703\nSigned-off-by: Mo Elbadry \u003celbadrym@google.com\u003e\n"
    },
    {
      "commit": "459dfa44d648c2404a440335977a8272339ed7f9",
      "tree": "d1a9c86d94c64f17188f7cbf62cec5f7f80e1d3a",
      "parents": [
        "d82c8b91e70bdec6d4ba33d5c2b884b24c17eb5e"
      ],
      "author": {
        "name": "Guangzong Chen",
        "email": "guangzong@google.com",
        "time": "Tue Nov 10 06:35:19 2026 +0000"
      },
      "committer": {
        "name": "Guangzong Chen",
        "email": "guangzong@google.com",
        "time": "Mon Nov 16 11:32:09 2026 -0800"
      },
      "message": "nvmed: SRCREV bump 82471714da..f57557ee45\n\nAgrim Bharat (1):\n      nvmed: Set MCDA when creating host telemetry log\n\nFusion-Link: fusion2/N/A\nTested: Verified recipe SRCREV bump for target hardware.\nGoogle-Bug-Id: 586518423\nChange-Id: I7c8d9e0123456789abcdef0123456789abcdef01\nSigned-off-by: Guangzong Chen \u003cguangzong@google.com\u003e\nTAG\u003dagy\nCONV\u003d514f125f-7ea2-4789-96e7-0c0d23e673e4\n"
    },
    {
      "commit": "d82c8b91e70bdec6d4ba33d5c2b884b24c17eb5e",
      "tree": "d5de5457be58abd81e8f7eadcfb8a686a55bd21e",
      "parents": [
        "0dd9c80992ca93e466027cf9cd0b5980e0243b29"
      ],
      "author": {
        "name": "Alex Lai",
        "email": "alexlai@google.com",
        "time": "Mon Nov 16 03:30:34 2026 +0000"
      },
      "committer": {
        "name": "Alex Lai",
        "email": "alexlai@google.com",
        "time": "Sun Nov 15 21:52:57 2026 -0800"
      },
      "message": "gbmcweb master: SRCREV bump bd8c636fc1..72a06d099e\n\nAnshul Nigham (1):\n      Automated Code Change\n\ngBMC Team (8):\n      Use mkstemp instead of std::tmpnam for temporary file creation in unit tests.\n      #tlbmc: Implement HalSensor for polling HAL SensorReaders\n      Preserve thermal zone manual mode across failsafe and expose Redfish FanMode.\n      Resolve sensor RelatedItem system_id from parent board topology when unset\n      Stop and join PowerControlThreadManager before destroying PowerControlGpio.\n      Use heterogeneous I2cGpioKey lookup for GpioCollector::i2c_input_gpios_.\n      Harden UpdateService URI/JSON validation and TaskService task parsing.\n      Make AmdCpuSensor depends on OS state instead of power state\n\nFusion-Link:\nplatform5: https://fusion2.corp.google.com/dc74a82a-8f82-3957-9154-cda64a62c521\nplatform15: https://fusion2.corp.google.com/28097951-96fc-39f7-b812-75fd0290c3c2\nplatform18: https://fusion2.corp.google.com/39f748fc-b471-3e5a-9cbb-4a5eb51ba269\nplatform17: https://fusion2.corp.google.com/8feeedfe-80c2-3344-b8fd-196e53d829ea\nplatform11+ext: https://fusion2.corp.google.com/f8260adc-03f4-30b7-a242-8e751f9f0dd3\nplatform11: https://fusion2.corp.google.com/5bc9b60b-178d-344d-aac2-f1d9f1ec00bd\nplatform25: https://fusion2.corp.google.com/fe7b2535-3122-3222-b4ee-2ceb75fb8926\nplatform26: https://fusion2.corp.google.com/e4b1d9ca-521d-3b27-a64b-9cf288a0c4ea\nTested: N/A\nGoogle-Bug-Id: 543446945\nGoogle-Bug-Id: 559017050\nGoogle-Bug-Id: 576982516,576982769\nGoogle-Bug-Id: 578510037\nGoogle-Bug-Id: 581513419\nGoogle-Bug-Id: 585336659\nChange-Id: Ibe8b0ca58c3fe6e0e33cd56287a78b31d9670e5c\nSigned-off-by: Alex Lai \u003calexlai@google.com\u003e\n"
    },
    {
      "commit": "0dd9c80992ca93e466027cf9cd0b5980e0243b29",
      "tree": "d2ab05dd390e6eb46be693dafbf0aff4b871da4a",
      "parents": [
        "1a910c62e8e8271eae6a780763997fac7e635510"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Sun Nov 15 14:24:12 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Sun Nov 15 17:59:28 2026 -0800"
      },
      "message": "hothd master: SRCREV bump 2307ab799f..ebad3a1128\n\nWilliam A. Kennington III (1):\n      hothd-usb: Fix template expansion and start rate limits\n\nPlatforms-Affected: All-TTF\nFusion-Link: fusion2 N/A\nTested: N/A\nGoogle-Bug-Id: 585587833\nChange-Id: I94756a4875e9c3d7d3d4c2fa244aab5f0bcd14a3\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "1a910c62e8e8271eae6a780763997fac7e635510",
      "tree": "13968e352e699ca91a19cd9ea959d285ced8dcce",
      "parents": [
        "6ed54f60983af040809eea489a7a12713671d3b5"
      ],
      "author": {
        "name": "Sam Agazaryan",
        "email": "samagazaryan@google.com",
        "time": "Sun Nov 15 00:10:39 2026 +0000"
      },
      "committer": {
        "name": "Sam Agazaryan",
        "email": "samagazaryan@google.com",
        "time": "Sun Nov 15 00:10:39 2026 +0000"
      },
      "message": "linux gbmc-6.18-ast27xx: SRCREV bump 63deff6a9a..c74565c05a\n\nSam Agazaryan (1):\n      i3c: mipi-i3c-hci: Don\u0027t rely only on IBI_READY\n\nTested: Validated on AST2750 A1 (65-min stress and 239 boots).\nFusion-Link: fusion2 N/A\nGoogle-Bug-Id: 581910830\nChange-Id: I02f6065077f837bac3d83400795926911005a441\nSigned-off-by: Sam Agazaryan \u003csamagazaryan@google.com\u003e\n"
    },
    {
      "commit": "6ed54f60983af040809eea489a7a12713671d3b5",
      "tree": "14e0aaeaf46630eebaffc9570cf92325960e2f19",
      "parents": [
        "0e864548a3a58046e482d9a4641e66ba137f4684"
      ],
      "author": {
        "name": "Joseph CT Chan",
        "email": "josephctchan@google.com",
        "time": "Mon Nov 09 01:31:28 2026 +0000"
      },
      "committer": {
        "name": "Joseph Chan",
        "email": "josephctchan@google.com",
        "time": "Fri Nov 13 16:32:46 2026 -0800"
      },
      "message": "linux-gbmc: npcm8xx: fix debounce register offset\n\nBackport upstream commit 602ee6c9447e (\"pinctrl: npcm8xx: fix debounce\nregister selection\") to linux-gbmc 6.12, 6.14, and 6.15-rc1\n(linux-gbmc_lts.bb, linux-gbmc_stable.bb, and linux-gbmc_mainline.bb).\n\nIn the Nuvoton NPCM845 pinctrl driver, debounce_timing_setting()\npreviously computed the GPnDBNCS register address using (gpio / 4)\ninstead of (gpio / 16) * 4. For GPIO offsets 4..15 (GPnDBNCS0) and\n20..31 (GPnDBNCS1), dividing by 4 produces unaligned register offsets\n(+0x81..+0x83 and +0x85..+0x87), causing 32-bit MMIO reads and writes\nin npcm_gpio_set() and npcm_gpio_clr() to fault on ARM64 Device memory.\n\nApply the upstream fix so that the GPnDBNCS0/1 register address is\n4-byte aligned.\n\nTested: Verified patch applies with 0 fuzz across linux-gbmc_lts.bb,\n  linux-gbmc_stable.bb, and linux-gbmc_mainline.bb.\nFusion-Link: fusion2 N/A\nGoogle-Bug-Id: 578568109\nChange-Id: I3e9f7b9054cb676cc48b919dbf311eae14a59bfb\nSigned-off-by: Joseph CT Chan \u003cjosephctchan@google.com\u003e\n"
    },
    {
      "commit": "0e864548a3a58046e482d9a4641e66ba137f4684",
      "tree": "b2011498b9f3bc839dafa074af052b00bd6017e0",
      "parents": [
        "66b9f6dd35bc8e0cf0def45a5ed4629afc7a4401"
      ],
      "author": {
        "name": "Jane Kramer",
        "email": "jkrm@google.com",
        "time": "Thu Nov 12 21:41:08 2026 +0000"
      },
      "committer": {
        "name": "Jane Kramer",
        "email": "jkrm@google.com",
        "time": "Fri Nov 13 12:16:00 2026 -0800"
      },
      "message": "phosphor-objmgr: Point SRC_URI at gbmc fork\n\nSwitch phosphor-objmgr to fetch from gbmc/phosphor-objmgr on the\ngoogle-dev branch and drop the two local .patch files, which are now\ncommitted directly on google-dev (80b58463d7).\n\nDrop FILESEXTRAPATHS since no local files remain for this recipe, and\nuse a plain SRCREV assignment so machine-level overrides continue to\ntake precedence.\n\nTested:\n- Verified bitbake -e phosphor-objmgr resolves gbmc/phosphor-objmgr\n  (google-dev @ 80b58463d7) with 0 .patch files, and diff -ruN against\n  google-dev shows 0 byte differences.\n- Built obmc-phosphor-image (verified phosphor-objmgr and libmapper1\n  1.0+git0+80b58463d7 in image manifest).\n- Ran gbmc-bumptool --external --recipes\u003dphosphor-objmgr --dryrun.\n\nFusion-Link: fusion2 N/A\nGoogle-Bug-Id: 572860608\nChange-Id: I033eb74091e37620cae6ddf0a7337db293c81cb3\nSigned-off-by: Jane Kramer \u003cjkrm@google.com\u003e\n"
    },
    {
      "commit": "66b9f6dd35bc8e0cf0def45a5ed4629afc7a4401",
      "tree": "68e74d95f9d7d71bf512a578700a8681b2469620",
      "parents": [
        "0649c730086bc593ca8c90105fd9a9bc85390869"
      ],
      "author": {
        "name": "Yuxiao Zhang",
        "email": "yuxiaozhang@google.com",
        "time": "Tue Nov 10 20:51:16 2026 -0800"
      },
      "committer": {
        "name": "Yuxiao Zhang",
        "email": "yuxiaozhang@google.com",
        "time": "Fri Nov 13 10:16:42 2026 -0800"
      },
      "message": "gbmc-bridge: Persist trip ID on empty RWFS purge\n\nWhen an external workflow (such as MAAF QuickErase or MachineDecom) or\nan older gBMC image without trip-ID logic places a 0-byte\n/var/google/do-rwfs-purge flag, init wipes RWFS on boot and leaves an\nempty /run/initramfs/rwfs-purged file without a trip ID.\n\nOnly skip purging via /run/initramfs/rwfs-purged when neither\ngbmc_psu_hardreset_needed nor gbmc_reboot_needed is set, and only if\nthe flag is empty (populating it with trip_id so 53-clear-purge-rwfs.sh\npersists it to /var/google/rwfs-purged), matches trip_id, or no trip\nID is available (e.g. the metadata fetch fell back to the v1 flow).\n\nTested:\n- Ran 49-purge-rwfs.sh and 53-clear-purge-rwfs.sh through 21 simulated\n  install flows (empty, matching, mismatched and missing trip IDs).\n  The MAAF 0-byte flag case needs one BMC powercycle instead of two;\n  only a flag left by a different trip purges where master did not.\n- Two full-machine RRIs on qtbid11 with test images built from this\n  change, 26.36.18.0 -\u003e 80.11.10.1 -\u003e 26.99.99.1, passed with one\n  netboot powercycle each, the same as with 26.36.18.0. They cover\n  the normal netboot path, not the empty-flag branch:\n  http://go/install-log/14249339843\n  http://go/install-log/14273464057\n- shellcheck reports no new warnings.\n\nFusion-Link: fusion2 N/A\nGoogle-Bug-Id: 584470391\nChange-Id: I87bfd070b0962f91b731f0e38dedacd144aac7c3\nSigned-off-by: Yuxiao Zhang \u003cyuxiaozhang@google.com\u003e\n"
    },
    {
      "commit": "0649c730086bc593ca8c90105fd9a9bc85390869",
      "tree": "5feaa5a54c34cdb5c75b605a9213bb8dd182c15d",
      "parents": [
        "14de6ac7c7ae6ba4fbe37d8bef492bf6e54c2715"
      ],
      "author": {
        "name": "Mo Elbadry",
        "email": "elbadrym@google.com",
        "time": "Fri Nov 13 03:33:33 2026 +0000"
      },
      "committer": {
        "name": "Mo Elbadry",
        "email": "elbadrym@google.com",
        "time": "Fri Nov 13 08:51:44 2026 -0800"
      },
      "message": "gbmc-bsp-common: Check DTB status properties\n\nExtend `GbmcDtbValidator.py` with `Severity`, `RuleId`, `Finding`, and\n`status` property validation (`INVALID_STATUS_PROPERTY`) across enabled\nDevicetree subtrees, catching non-standard `status` values (`\"disable\"`,\n`\"enabled\"`) while relying on `dtc` `-Estatus_is_string` and\n`-Ecompatible_is_string_list` for wire-format string checks.\n\nFusion-Link: fusion2 N/A\nTested: GbmcDtbValidatorTester.py (5/5 unit tests pass).\nGoogle-Bug-Id: 579089739\nChange-Id: I6273849506928374651203948576120394857602\nSigned-off-by: Mo Elbadry \u003celbadrym@google.com\u003e\n"
    },
    {
      "commit": "14de6ac7c7ae6ba4fbe37d8bef492bf6e54c2715",
      "tree": "21a0649db3f192daa018e83a87ea15bbb0678929",
      "parents": [
        "18ba69dadd5e3565ba9ec72d442a25b9ae2dcaaa"
      ],
      "author": {
        "name": "Mo Elbadry",
        "email": "elbadrym@google.com",
        "time": "Fri Nov 13 03:19:43 2026 +0000"
      },
      "committer": {
        "name": "Mo Elbadry",
        "email": "elbadrym@google.com",
        "time": "Fri Nov 13 08:51:44 2026 -0800"
      },
      "message": "gbmc-bsp-common: Add FDT parser unit tests\n\nAdd `GbmcDtbValidatorTester.py` using `libfdt.FdtSw` (`serialize_fdt`)\nto construct test FDT binaries and test `parse_fdt_blob` property\ndecoding and malformed FDT header/structure rejection via\n`libfdt.fdt_check_full`.\n\nFusion-Link: fusion2 N/A\nTested: GbmcDtbValidatorTester.py (2/2 unit tests pass).\nGoogle-Bug-Id: 579089739\nChange-Id: I7182930495817263540192837465019283746501\nSigned-off-by: Mo Elbadry \u003celbadrym@google.com\u003e\n"
    },
    {
      "commit": "18ba69dadd5e3565ba9ec72d442a25b9ae2dcaaa",
      "tree": "491a6bb05c898321413cfdd621c6baa78417c406",
      "parents": [
        "1d48311ada786553afc068aaad807df7006154ab"
      ],
      "author": {
        "name": "Mo Elbadry",
        "email": "elbadrym@google.com",
        "time": "Fri Nov 13 03:19:43 2026 +0000"
      },
      "committer": {
        "name": "Mo Elbadry",
        "email": "elbadrym@google.com",
        "time": "Fri Nov 13 08:51:44 2026 -0800"
      },
      "message": "linux-common: Add presubmit DTB validator\n\nDevice Tree Source (.dts/.dtsi) regressions in gBMC kernel trees can\ncompile without syntax errors under `dtc` yet break hardware probing\nat runtime. For example, leaving a companion USB 1.1 controller\n(`uhci`/`ohci`) enabled alongside `ehci` causes synchronous UHCI/OHCI\nprobing to claim `usb1` before asynchronous EHCI registration\n(b/579089739, b/574912243), and `status` typos (`\"disable\"`,\n`\"enabled\"`), `/aliases/usb*`, missing USB `#address-cells`, or\nduplicate unit addresses silently misconfigure devices.\n\nAdd `GbmcDtbValidator.py` (`pylibfdt`-backed Flattened Device Tree\nvalidator CLI) and wire `do_gbmc_dtbvalidate` into `linux-common.inc`:\n- Validate compiled `.dtb`/`.dtbo` binaries via `libfdt.fdt_check_full`\n  and `libfdt.FdtRo` (`python3-dtc-native` via\n  `gbmc-bsp-common-native`).\n- Append `-Estatus_is_string`, `-Ecompatible_is_string_list`, and\n  `-Ealias_paths` to `KERNEL_DTC_FLAGS` so `dtc` enforces property\n  types at compile time.\n- Enable `GBMC_DTBCHECK:dev ??\u003d \"1\"` in warnings-only mode\n  (`GBMC_DTBCHECK_STRICT ??\u003d \"0\"`), routing non-zero exit output to\n  `bbfatal` and warning output to `bbwarn`.\n\nFusion-Link: fusion2 N/A\nTested: GbmcDtbValidatorTester.py in the next CL in this stack.\nGoogle-Bug-Id: 579089739\nChange-Id: Ifc4c85bc2adb03592aa23f0a2c1b3d3ffbabe992\nSigned-off-by: Mo Elbadry \u003celbadrym@google.com\u003e\n"
    },
    {
      "commit": "1d48311ada786553afc068aaad807df7006154ab",
      "tree": "7d0d36034761087465b6c8db8df6ccbcf25f56b8",
      "parents": [
        "b466679c15d04d6519f7b288d04f618761c5b823"
      ],
      "author": {
        "name": "Vivekanand Veeracholan",
        "email": "vveerach@google.com",
        "time": "Thu Nov 12 17:51:10 2026 -0800"
      },
      "committer": {
        "name": "Vivekanand Veeracholan",
        "email": "vveerach@google.com",
        "time": "Thu Nov 12 21:52:24 2026 -0800"
      },
      "message": "google/libhoth main: SRCREV bump 9d5f1a561d..3e8b236b1a\n\nKor Nielsen (2):\n      opentitan_version: Parse bundle manifests with a struct\n      opentitan dfu: Extract version from 88k-offset bundles\n\nMichael Field (7):\n      [transports] Fix ioctl pointer argument in spi_nor_busy_wait\n      rust: Add protobuf and crates needed by tlvtool\n      [protocol] Widen host command buffers to support 6 KiB mailbox\n      [transports] Support 6 KiB requests across USB FIFO and spidev\n      [protocol] Add OpenTitan key rotation record write-commit dispatch\n      Zero Initialize variables causing msan failures\n      Fix for downstream build error\n\nRoyce (1):\n      payload_update: Base the progress ETA on sent bytes (#326)\n\nSteven Portley (1):\n      rust: Add initial rust setup to libhoth\n\nesnguyen (2):\n      [examples] Convert remaining htool.c commands to use the new compact error format (#318)\n      [protocol] Remove libhoth_hostcmd_exec_v2 strangler fig (#321)\n\nFusion-Link: fusion2 N/A\nGoogle-Bug-Id: 501181776\nTested: N/A\nChange-Id: Iddddcaf0bac3000ea6a288d6d08448d6f677112c\nSigned-off-by: Vivekanand Veeracholan \u003cvveerach@google.com\u003e\n"
    },
    {
      "commit": "b466679c15d04d6519f7b288d04f618761c5b823",
      "tree": "7198d2cb9efd6311cee76967eab177b795cd9fbb",
      "parents": [
        "1e29323400b5fe4874988af7317b00f7495b34fb"
      ],
      "author": {
        "name": "Hamza Sheikh",
        "email": "hamzasheikh@google.com",
        "time": "Thu Nov 12 19:43:58 2026 +0000"
      },
      "committer": {
        "name": "Hamza Sheikh",
        "email": "hamzasheikh@google.com",
        "time": "Thu Nov 12 19:13:54 2026 -0800"
      },
      "message": "gbmcweb master: SRCREV bump 2002f9c446..bd8c636fc1\n\nBump gbmcweb SRCREV to bd8c636fc1 (including the narrowed insecure\ntransport blocklist in c717ae31b5) and enable\n-Ddisable-mutable-redfish-services-from-insecure-transport\u003denabled for\nprod builds only (dev and MFG builds are unaffected).\n\nNan Zhou (1):\n  c9df2c983c Add build flag to disable mutable Redfish services over\n             insecure transport\n\ngBMC Team (15):\n  008b07e110 Refactor installation of generated .pb.h\n  486935e6a2 Omit Cable Assembly link when assembly ownership is\n             disabled\n  d99e72d058 Fix TrayPowerCycle PSU checkpoint name during root chassis\n             reset\n  d660d16c7e Disable unused mutable Redfish services over insecure\n             transports\n  f5e3e35089 Create SolveNonNegativeLeastSquares for FRE\n  f9971616d9 gbmcweb: Harden dump attachment download in\n             downloadDumpEntry\n  44d0970ef9 Make GeneratePointsWithLogSpacing a template\n  4d4826120b gbmcweb: Expose FormatActionInfo for NVMe Format (Storage\n             scope)\n  cef776e9c5 gbmcweb: Expose FormatActionInfo for NVMe Format\n             (Controller scope)\n  c717ae31b5 Narrow mutable services disabled over insecure transports\n  2c64991c29 tlbmc: Add an SSE parser and streaming client\n  c40641d9f5 tlbmc: Extract LogService helpers from the SEL route\n  dbe380325c Create GenerateDerivativeFilterTimeConstants for FRE\n  90f37d3ca5 Create GenerateBlockDiagonalMatrix for FRE\n  bd8c636fc1 tlbmc: Add an fsync option to FileManager::WriteToFile\n\nTested: Verified on hardware across Port 80, Port 3995, and Port 443.\nFusion-Link: fusion2 N/A\nPlatforms-Affected: All\nGoogle-Bug-Id: 395219125\nGoogle-Bug-Id: 502998640\nGoogle-Bug-Id: 545329676\nGoogle-Bug-Id: 549163378\nGoogle-Bug-Id: 559754428\nGoogle-Bug-Id: 564573479\nGoogle-Bug-Id: 580979817\nGoogle-Bug-Id: 580980114\nGoogle-Bug-Id: 581902351\nGoogle-Bug-Id: 582151177\nGoogle-Bug-Id: 583625916\nGoogle-Bug-Id: 584397825\nSigned-off-by: Hamza Sheikh \u003chamzasheikh@google.com\u003e\nChange-Id: I9c4d7a18e2b5046f83910c2d7e5a8b6f412039e5\n"
    },
    {
      "commit": "1e29323400b5fe4874988af7317b00f7495b34fb",
      "tree": "b54e2e4590af7070e478e21cb69e751088b3c555",
      "parents": [
        "aa0f1db16ae96a4c8a5e7739bc99ff989e577d34"
      ],
      "author": {
        "name": "Matthew Radosavljevic",
        "email": "mradosav@google.com",
        "time": "Wed Nov 11 23:29:40 2026 +0000"
      },
      "committer": {
        "name": "Matthew Radosavljevic",
        "email": "mradosav@google.com",
        "time": "Thu Nov 12 14:07:59 2026 -0800"
      },
      "message": "Remove printproto recipe\n\nThis recipe is not use by anything so it should be removed\n\nChange-Id: I96d1d517dd572dcfeab9ccce9dc6a3c0016fd718\nFusion-Link: fusion2 n/a\nGoogle-Bug-Id: 581927589\nSigned-off-by: Matthew Radosavljevic \u003cmradosav@google.com\u003e\nTested: n/a\n"
    },
    {
      "commit": "aa0f1db16ae96a4c8a5e7739bc99ff989e577d34",
      "tree": "f3301477b5d929f6053be7792a8bc42a41b19d7a",
      "parents": [
        "c7ab867f0131ea99c69cd9120daf15d30c48f095"
      ],
      "author": {
        "name": "cmchang",
        "email": "cmchang@google.com",
        "time": "Thu Nov 12 10:09:45 2026 +0000"
      },
      "committer": {
        "name": "cmchang",
        "email": "cmchang@google.com",
        "time": "Thu Nov 12 10:09:46 2026 +0000"
      },
      "message": "linux gbmc-6.18-ast27xx: SRCREV bump 9962c14a6f..63deff6a9a\n\nAdrian Hunter (1):\n      i3c: mipi-i3c-hci: Factor out DMA mapping from queuing path\n\nFusion-Link: fusion2 N/A\nTested: N/A\nGoogle-Bug-Id: 549948410\nChange-Id: I4f5e835f1c14af4572f4e28ffdf00d980458b2b1\nSigned-off-by: cmchang \u003ccmchang@google.com\u003e\n"
    },
    {
      "commit": "c7ab867f0131ea99c69cd9120daf15d30c48f095",
      "tree": "3d6df92301e5244750c71bc27d3737c074fee458",
      "parents": [
        "ca5103dbc39edf320241d43ea005ee627dee2934"
      ],
      "author": {
        "name": "Yolkee Huang",
        "email": "cweihuang@google.com",
        "time": "Tue Nov 03 19:54:40 2026 +0800"
      },
      "committer": {
        "name": "Yolkee Huang",
        "email": "cweihuang@google.com",
        "time": "Tue Nov 10 18:30:22 2026 -0800"
      },
      "message": "mctp-reactor: Unmanage bridged devices on removal\n\nWhen entity-manager removes and re-adds a bridged MCTP device,\nmctpreactor fails to re-register its endpoint and logs an\nunsynchronised endpoint reinitialisation warning. Today, the inventory\nremoval handler unmanages direct bus devices but omits bridged\ndownstream devices, leaving stale endpoint state in the reactor.\nUpdate the removal handler to match bridged devices so endpoints are\ncleanly unmanaged before re-enumeration.\n\nBackground \u0026 Story:\nOn multi-NIC gBMC systems, ConnectX-8 NICs sit behind a USB MCTP\nbridge and are published by entity-manager as\nMCTPBridgeDownstreamDevice objects (CX8_0 through CX8_3). During a FRU\nrescan or hardware reset, MCTPReactorMain.cpp processes D-Bus\ninventory signals in three steps:\n\n  +------------------------------------------------------------------+\n  | Step 1: Inventory Added (addInventory -\u003e deviceFromConfig)       |\n  |   Matches: I2C | I3C | USB | BridgeDownsteamMCTPDDevice          |\n  |   -\u003e reactor-\u003emanageMCTPDevice() registers endpoint \u0026 EID        |\n  +------------------------------------------------------------------+\n    |\n    v\n  +------------------------------------------------------------------+\n  | Step 2: Inventory Removed (removeInventory on InterfacesRemoved) |\n  |   - Before: Checks only I2C | I3C | USB (omits Bridged!)         |\n  |             -\u003e CX8_0..CX8_3 remain stale in reactor-\u003edevices     |\n  |   - After : Also checks BridgeDownsteamMCTPDDevice::match()      |\n  |             -\u003e reactor-\u003eunmanageMCTPDevice() clears state        |\n  +------------------------------------------------------------------+\n    |\n    v\n  +------------------------------------------------------------------+\n  | Step 3: Inventory Re-added (addInventory after rescan/reset)     |\n  |   - Before: Hits EBUSY (unsynchronised endpoint reinitialisation)|\n  |   - After : Cleanly re-registers endpoint and assigns EID        |\n  +------------------------------------------------------------------+\n\nRoot Cause:\n- While Step 1 (deviceFromConfig()) matches all four device types,\n  Step 2 (removeInventory()) omits\n  BridgeDownsteamMCTPDDevice::match(removed) and leaves removed\n  bridged devices in reactor-\u003edevices and reactor-\u003edeferred.\n\nFix:\n- Add BridgeDownsteamMCTPDDevice::match(removed) to removeInventory()\n  in MCTPReactorMain.cpp so bridged MCTP devices call\n  reactor-\u003eunmanageMCTPDevice(path.str) on removal.\n- Add unit tests in test_MCTPEndpoint.cpp covering\n  BridgeDownsteamMCTPDDevice::match() against empty, irrelevant, and\n  matching SensorData and removed-interface sets.\n\nTested:\n- Built mctp-reactor cleanly with BitBake.\n- Passed test_mctp_suite (17/17) under qemu-arm.\n- Verified on yucbfsr-bct9-bmc-01 and yucbfsr-eav13-bmc-01 that\n  removing and re-adding CX8_2 and CX8_3 cleanly unmanages and\n  restores their bridged MCTP endpoints with zero unsynchronised\n  reinitialisation errors:\n  https://paste.googleplex.com/6755399436353937\n\nFusion-Link: fusion2 n/a\nGoogle-Bug-Id: 571350975\nChange-Id: I3b6b652f8c32743583f61dbf66ab0ea537ee6d5a\nSigned-off-by: Yolkee Huang \u003ccweihuang@google.com\u003e\n"
    },
    {
      "commit": "ca5103dbc39edf320241d43ea005ee627dee2934",
      "tree": "bd1438babc4670179692e2d8a31104f13b803718",
      "parents": [
        "2ffdbb8d9bd15b427aac2006e62afc1565d84b8b"
      ],
      "author": {
        "name": "Jane Kramer",
        "email": "jkrm@google.com",
        "time": "Sat Nov 07 00:45:32 2026 +0000"
      },
      "committer": {
        "name": "Jane Kramer",
        "email": "jkrm@google.com",
        "time": "Tue Nov 10 13:37:47 2026 -0800"
      },
      "message": "entity-manager: Fix use-after-free in i2cFruScan\n\ni2cFruScan keeps references to failedAddresses[bus] and\nfruAddresses[bus] during a bus scan. If rootBus is not in the flat_map,\noperator[] allocates a new buffer and frees the old buffer. The next\nwrite through the old reference corrupts freed memory and crashes\nfru-device.\n\nLook up rootBus in failedAddresses and fruAddresses with find() instead\nof operator[] so failedItems and foundItems stay valid for the whole\nscan.\n\nTested:\n- Ran fru-device on HumuBeach (fden88) without the bus 132 blocklist:\n  https://paste.googleplex.com/6755399436784302\nhttps://paste.googleplex.com/6755399436855557\n- Verified bitbake -c patch entity-manager with downstream patches.\n- Ran the unit test with ASan and UBSan on x86_64 and qemu-aarch64.\n\nFusion-Link: fusion2 N/A\nGoogle-Bug-Id: 581508240\nChange-Id: I3f83a82229ac3e75e64013cf6d60a45ef59ad1cb\nSigned-off-by: Jane Kramer \u003cjkrm@google.com\u003e\n"
    },
    {
      "commit": "2ffdbb8d9bd15b427aac2006e62afc1565d84b8b",
      "tree": "c6c3b76a0b083edd756ba57c2ffd356bb3d077fa",
      "parents": [
        "976a7bf3cb0041856f6bcef24b6f487dd527ca37"
      ],
      "author": {
        "name": "Luka Strizic",
        "email": "lstrz@google.com",
        "time": "Mon Nov 09 23:20:14 2026 +0000"
      },
      "committer": {
        "name": "Luka Strižić",
        "email": "lstrz@google.com",
        "time": "Tue Nov 10 10:55:46 2026 -0800"
      },
      "message": "mctp: Restore control IID check with legacy IID-0 quirk\n\n0001-mctpd-ignore-IIDs-validation-for-control-responses.patch disabled\nIID rejection on all MCTP control responses to support legacy SSD mux\nfirmware returning IID 0 on both success and error responses\n(b/437126395). Accepting arbitrary IID mismatches lets delayed replies\nfrom earlier timed-out requests satisfy newer requests.\n\nReplace 0001-mctpd-ignore-IIDs-validation-for-control-responses.patch\nwith a narrow exception allowing IID 0 only when command_code matches,\nand skip IID 0 on outgoing requests so delayed replies from compliant\nendpoints never match the legacy exception.\n\nTested: Deployed full mctpd + mctpreactor stack on hardware. Verified\nclean peer cleanup and auto-recovery on transient bridge timeout, atomic\nBridge1 pool publication before InterfacesAdded, live Endpoint1.Remove\nrecovery, 5/5 cold-start stress cycles, and PLDM/Redfish discovery\nacross both USB bridges and all 4 downstream bridged NICs:\nhttps://paste.googleplex.com/6755399436596849\n\nFusion-Link: fusion2 N/A\nGoogle-Bug-Id: 437126395\nGoogle-Bug-Id: 508310010\nChange-Id: I89e64dd2918da63f189c1af93659700136ccfa2b\nSigned-off-by: Luka Strizic \u003clstrz@google.com\u003e\n"
    },
    {
      "commit": "976a7bf3cb0041856f6bcef24b6f487dd527ca37",
      "tree": "41cf7d47c139d24279285769a397f9251b6a9ea6",
      "parents": [
        "8c92deef2125d49f0e6c22858a7cdc472dbd6852"
      ],
      "author": {
        "name": "Jane Kramer",
        "email": "jkrm@google.com",
        "time": "Sat Nov 07 00:04:57 2026 +0000"
      },
      "committer": {
        "name": "Jane Kramer",
        "email": "jkrm@google.com",
        "time": "Mon Nov 09 19:01:24 2026 -0800"
      },
      "message": "phosphor-ipmi-net: Point SRC_URI at gbmc fork\n\nSwitch phosphor-ipmi-net to fetch from gbmc/phosphor-net-ipmid on the\ngoogle-dev branch and drop the local HMAC-SHA1 revert .patch file, which\nis now committed directly on google-dev (09025a9007).\n\nKeep 99-gbmcbr-interface.rules and FILESEXTRAPATHS in place, and use a\nplain SRCREV assignment so machine-level overrides continue to take\nprecedence.\n\nTested:\n- Verified bitbake -e phosphor-ipmi-net resolves gbmc/phosphor-net-ipmid\n  (google-dev @ 09025a9007) + 99-gbmcbr-interface.rules with 0 .patch\n  files, and diff -ruN against google-dev shows 0 byte differences.\n- Built obmc-phosphor-image on targets with and without obmc-net-ipmi\n  enabled (verified phosphor-ipmi-net 1.0+git0+09025a9007, netipmid, and\n  99-gbmcbr-interface.rules in rootfs).\n- Verified bumptool dry-run resolves gbmc/phosphor-net-ipmid google-dev\n  and generates SRCREV bump commits cleanly.\n\nFusion-Link: fusion2 N/A\nGoogle-Bug-Id: 572860567\nChange-Id: I78530db65d543a8eae628e30dc48ebec010bf31c\nSigned-off-by: Jane Kramer \u003cjkrm@google.com\u003e\n"
    },
    {
      "commit": "8c92deef2125d49f0e6c22858a7cdc472dbd6852",
      "tree": "563872aace488c73d67dbbf1b954ef3cbff1b1e2",
      "parents": [
        "001b9d2592c1fe096276cf6f908958b9b977ed61"
      ],
      "author": {
        "name": "Ramon Rakow",
        "email": "rakow@google.com",
        "time": "Thu Nov 05 21:42:53 2026 +0000"
      },
      "committer": {
        "name": "Ramon Rakow",
        "email": "rakow@google.com",
        "time": "Mon Nov 09 15:23:26 2026 -0800"
      },
      "message": "psdmd master: SRCREV bump 64f6a53747..b2f72827e7\n\nGeorge Kola (1):\n      psdmd: Fix timeout overflow, status file, history\n\nRamon Rakow (2):\n      psdmd: Fix clang-tidy config and GCC 15 build\n      psdmd: Do not clobber NSS when SKM EEPROM is unresponsive\n\ngBMC Team (9):\n      Remove deprecated TPC ProtoDB annotations from .proto files\n      Update doc comment for `preserve_user_data`\n      Remove go_legacy_channel_api from remaining protos\n      [DPCY WS1] Add thinmint field to DeviceActionRequest in storage_device_action.proto\n      psdmd: Fix Copybara migration and OpenBMC proto export.\n      Add firmware_version field to DeviceStatus for erasure audit logging.\n      Add STRICT_SANITIZE enum value to Erase.Type for NVMe Quick Erase.\n      Merge Astra tsp proto_compare annotation overrides into main annotations.\n      Delete szl_api_option from all proto files in google3 (no-op).\n\nTested: Built psdmd (b2f72827e7) via `bitbake psdmd` (2842/2842 tasks\nsucceeded) and verified SKM EEPROM blocking/recovery on target BMC\nddcbt8-nfd01.prod.google.com (gbmc-release-26.36.18.0):\nhttps://paste.googleplex.com/6755399436723237\n\nAlso verified via Guitar Mimik workflows (10/10 passing),\ngbmc_oob_psdmd_test on real hardware (passing), hardware presubmit\nworkflows (2/2 passing), and MSV real-hardware workflows (5 passing;\nremaining failures checked against clean HEAD).\n\nFusion-Link: http://fusion2/ci/kokoro/prod%3Agbmc%2Ffirmware-build%2Ffirmware%2Fgerrit_upstreamed_platforms/activity/6c431a84-703d-4598-809f-1e0fe5c7c5f1\nFusion-Link: https://fusion2.corp.google.com/929edc51-bbac-450a-bc88-2cdc45cac221\nFusion-Link: https://fusion2.corp.google.com/9e0fb8a1-059b-3fee-93eb-1004330f8e55\nFusion-Link: https://fusion2.corp.google.com/b0c1a37f-6363-300e-aa08-d1693dc7163b\nFusion-Link: https://fusion2.corp.google.com/8839c99d-a8b5-3c8b-915f-c072f7410850\nFusion-Link: https://fusion2.corp.google.com/19a6533e-18e7-316b-81be-754015e5ad37\nFusion-Link: https://fusion2.corp.google.com/c641dabf-534a-3e21-b795-93790ee39b6b\nFusion-Link: https://fusion2.corp.google.com/826f29e9-5b94-382a-a1d6-3e3231ae8ac6\nFusion-Link: https://fusion2.corp.google.com/40095d18-006e-3f1a-8f28-8e35ed9335aa\nFusion-Link: https://fusion2.corp.google.com/d842eb12-fbb4-3056-a94a-f5e411ae8420\nFusion-Link: https://fusion2.corp.google.com/03e8b359-6f55-35e1-b9a2-b6a21502b335\nFusion-Link: https://fusion2.corp.google.com/0f60da2d-96d5-3a43-8ace-c43d9ab6720d\nFusion-Link: https://fusion2.corp.google.com/a6b5b535-8471-36e3-a77b-a72621698f27\nFusion-Link: https://fusion2.corp.google.com/ed1eacea-d4b4-313b-915e-672fcd926c54\nFusion-Link: https://fusion2.corp.google.com/bfac183b-75de-3468-9e18-f5df242417b5\nFusion-Link: https://fusion2.corp.google.com/d54aa698-7a2c-388a-b7f8-3d599bceacb6\nFusion-Link: https://fusion2.corp.google.com/cfc4e8a4-f0d9-3066-9d96-04a1e8532565\nFusion-Link: https://fusion2.corp.google.com/58f11afb-c923-3a66-bea0-a4bbf7eb4ea1\nFusion-Link: https://fusion2.corp.google.com/29c47e61-0890-3df2-87a2-4a09a54041ee\nGoogle-Bug-Id: 468433665\nChange-Id: I576649bfeb918d684ee0cfd3c56413b40dba5cfa\nSigned-off-by: Ramon Rakow \u003crakow@google.com\u003e\nTAG\u003dagy\nCONV\u003d42437f37-d165-4a88-8c7f-e05947fb1ef6\n"
    },
    {
      "commit": "001b9d2592c1fe096276cf6f908958b9b977ed61",
      "tree": "85aaa91e312e1824bf03ddeb15c33c930c9e77af",
      "parents": [
        "89647cea448014b29a42efa3e4a91d14566dbf6a"
      ],
      "author": {
        "name": "Luka Strizic",
        "email": "lstrz@google.com",
        "time": "Tue Nov 03 21:01:08 2026 +0000"
      },
      "committer": {
        "name": "Luka Strižić",
        "email": "lstrz@google.com",
        "time": "Mon Nov 09 12:50:57 2026 -0800"
      },
      "message": "mctp: Preserve pool allocation error code\n\nWhen endpoint_send_allocate_endpoint_ids() fails,\nendpoint_allocate_eids() cleans up the pre-installed downstream gateway\nroute via mctp_nl_route_del() and overwrites rc with its return value.\nWhen route deletion succeeds, endpoint_allocate_eids() returns 0 and\nmasks the pool allocation failure, preventing callers from handling\nfailed bridge setup.\n\nPreserve the allocation error code across route cleanup and return\nnegative errno values instead of -1 in preparation for propagating pool\nallocation failures to callers.\n\nTested: Deployed full 5-patch mctpd + mctpreactor stack on hardware.\nVerified clean peer cleanup and auto-recovery on transient bridge\ntimeout, atomic Bridge1 pool publication before InterfacesAdded, live\nEndpoint1.Remove recovery, 5/5 cold-start stress cycles, and\nPLDM/Redfish discovery across both USB bridges and all 4 downstream\nbridged NICs: https://paste.googleplex.com/6755399436596849\n\nFusion-Link: fusion2 N/A\nGoogle-Bug-Id: 508310010\nChange-Id: I38638ed14326d054668015bad39abd406f59718c\nSigned-off-by: Luka Strizic \u003clstrz@google.com\u003e\n"
    },
    {
      "commit": "89647cea448014b29a42efa3e4a91d14566dbf6a",
      "tree": "c72ab4089d231ee0492aee31a30a74eff6783c0f",
      "parents": [
        "f36e49fa8c22de7dffb7b076f8c48203a0d7c4b8"
      ],
      "author": {
        "name": "Matthew Radosavljevic",
        "email": "mradosav@google.com",
        "time": "Wed Nov 04 18:22:19 2026 +0000"
      },
      "committer": {
        "name": "Matthew Radosavljevic",
        "email": "mradosav@google.com",
        "time": "Mon Nov 09 09:29:47 2026 -0800"
      },
      "message": "gbmcweb-google-plugins master: SRCREV bump 14c9d7a16f..899ba9c1ea\n\nMatthew Radosavljevic (1):\n      Remove crashdump support\n\nFusion-Link: fusion2 N/A\nGoogle-Bug-Id: 557231843\nTested: N/A\nChange-Id: I8cd5e771e9582cc0df771d6726b23824f110df24\nSigned-off-by: Matthew Radosavljevic \u003cmradosav@google.com\u003e\n"
    },
    {
      "commit": "f36e49fa8c22de7dffb7b076f8c48203a0d7c4b8",
      "tree": "27c6aa5f8b19643776287c06850a9824d104ee66",
      "parents": [
        "93e01e16e4b97d41fe1e57b15c7221bb5ee0885e"
      ],
      "author": {
        "name": "Munawar Hussain",
        "email": "munawarhussain@google.com",
        "time": "Sat Nov 07 14:56:31 2026 +0000"
      },
      "committer": {
        "name": "Munawar Hussain",
        "email": "munawarhussain@google.com",
        "time": "Sat Nov 07 07:03:50 2026 -0800"
      },
      "message": "dbus-sensors google-dev: SRCREV bump c0165b7fdb..82471714da\n\nGuangzong Chen (1):\n      nvmed: expose NVMe.Format capability properties\n\nHao Jiang (7):\n      nvmesensor: bounce lockdown tasks via io thread\n      nvmesensor: gate runtime MI tasks via tryPost()\n      nvmesensor: guard epOptimize with optimizeTimer\n      nvmesensor: run adminFwCommit on worker thread\n      nvmesensor: avoid endpoint capture in worker\n      nvmesensor: make MI callbacks cancellable\n      nvmesensor: address controllers by id\n\nFusion-Link: fusion2 N/A\nTested: on each commit.\nGoogle-Bug-Id: 560407116\nGoogle-Bug-Id: 564573285\nGoogle-Bug-Id: 573764995\nGoogle-Bug-Id: 573831528\nGoogle-Bug-Id: 578681755\nChange-Id: I6e47e7fedf9620ff67214e816af5262009315ddf\nSigned-off-by: Munawar Hussain \u003cmunawarhussain@google.com\u003e\n"
    },
    {
      "commit": "93e01e16e4b97d41fe1e57b15c7221bb5ee0885e",
      "tree": "771f310c7c44653f7ce06c3f7bfa39ac66fe02db",
      "parents": [
        "e76cf4aedd35c455c2ae19cb24549d8439d2fec5"
      ],
      "author": {
        "name": "Munawar Hussain",
        "email": "munawarhussain@google.com",
        "time": "Sat Nov 07 14:56:30 2026 +0000"
      },
      "committer": {
        "name": "Munawar Hussain",
        "email": "munawarhussain@google.com",
        "time": "Sat Nov 07 07:03:40 2026 -0800"
      },
      "message": "libnvme: Backport nvme_mi_ctrl_id() accessor\n\ndbus-sensors 8247171 (\"nvmesensor: address controllers by id\") calls\nnvme_mi_ctrl_id() to address NVMe-MI controllers by ID instead of\ncasting into the private struct nvme_mi_ctrl. The upstream subproject\nwrap provides it, but Yocto builds nvmed with wrap_mode\u003dnodownload\nagainst the recipe-sysroot libnvme, so nvmed fails in do_compile\n(\u0027nvme_mi_ctrl_id\u0027 was not declared in this scope) without this\nbackport.\n\nBackport upstream eaa7ad4eb224 (\"mi: Add nvme_mi_ctrl_id to retrieve\ncontroller ID\") as 0017.\n\nTested: N/A.\n\nFusion-Link: fusion2 N/A\nGoogle-Bug-Id: 560407116\nChange-Id: Ie6df408a55fa5d7d68f1e4d3cd1f2e53792a8f77\nSigned-off-by: Munawar Hussain \u003cmunawarhussain@google.com\u003e\n"
    },
    {
      "commit": "e76cf4aedd35c455c2ae19cb24549d8439d2fec5",
      "tree": "39a760562c32b0e56513836463372b9233354771",
      "parents": [
        "83ee34059bd5c0b4502f9b04faf6ba3916dd4b6d"
      ],
      "author": {
        "name": "Edward Lee",
        "email": "edwarddl@google.com",
        "time": "Fri Nov 06 21:45:14 2026 +0000"
      },
      "committer": {
        "name": "Edward Lee",
        "email": "edwarddl@google.com",
        "time": "Fri Nov 06 21:45:21 2026 +0000"
      },
      "message": "hothd master: SRCREV bump 1618dbca16..2307ab799f\n\nEdward Lee (1):\n      hothd-usb: Bind service lifetime to its USB device\n\nEllis Sarza-Nguyen (1):\n      Add firmware mauv to the dbus state\n\nLovepreet Singh (1):\n      Update HothState to show chip information for OpenTitan chips as well\n\nShou-En Hsiao (1):\n      payload_update: Raise SendPayload timeout to 20m\n\nSteven Portley (1):\n      hoth_updater: Add retries\n\nPlatforms-Affected: All-TTF\nFusion-Link: fusion2 N/A\nTested: N/A\nGoogle-Bug-Id: 484142258\nGoogle-Bug-Id: 575099179\nGoogle-Bug-Id: 581687485\nChange-Id: Ifcf01786e3e987a3e6f0c33e4fb66b693f237305\nSigned-off-by: Edward Lee \u003cedwarddl@google.com\u003e\n"
    },
    {
      "commit": "83ee34059bd5c0b4502f9b04faf6ba3916dd4b6d",
      "tree": "c10032d333351739ca9bc09f4c37445d946f60e3",
      "parents": [
        "1e4a58f2e23dddbea5212ac725aa008b723fb086"
      ],
      "author": {
        "name": "Sandesh",
        "email": "sandeshharidas@google.com",
        "time": "Thu Nov 05 21:19:39 2026 +0000"
      },
      "committer": {
        "name": "Sandesh",
        "email": "sandeshharidas@google.com",
        "time": "Thu Nov 05 21:19:55 2026 +0000"
      },
      "message": "gbmcweb master: SRCREV bump f0506b3981..2002f9c446\n\ngBMC Team (23):\n      Harden GbmcSel filter parsing and JSON Id validation.\n      Fix PowerControlInotify stop deadlock and state/inotify parsing bounds.\n      gbmcweb: Ignore unknown fields in devpath_plugin mapping config\n      Allow GET ManagerDiagnosticData in Bloom recovery authz\n      Skip missing or failed GPIOs during GpioCollector::Create\n      Support setting folder path for crashdumps\n      Cap HFT sampling at the collector\u0027s ambient polling interval\n      Create non-negative least squares solver for FRE\n      Keep BootOptions across tray power cycles A tray power cycle cleared all persisted BIOS settings, including BootOptions, so clients could not set BootNext until the host re-registered them on the next boot. Switch the tray power cycle from SettingsManager::ClearAllBiosSettings() to ClearBiosSettings(settings) with an explicit list: core count, BootNext, memory capacity, and boot source override target/enabled. BootOptions are not listed, so they are kept.\n      tlbmc: Treat NaN shm sensor values as unavailable\n      Replace spaces with underscores in shared memory sensor names.\n      storage: Drive Status.State follows OperationalStatus\n      Release gpiod_line handle on error paths in Gpio::Create.\n      Store tlBMC sensor history as inline SensorSample structs.\n      Fix sensor collector, sensor polling, and thermal zone manager safety bugs.\n      #tlbmc: Add HalSensorConfig proto and EntityConfigJsonImpl parser\n      Harden IPMI FRU bounds checks and config parser safety.\n      Refine FRE utilities\n      tlbmc: Support HFT telemetry collection for overridden sensors.\n      Revert \"bmcweb: start at a very high priority level\"\n      gbmcweb: Handle persistent data file write errors gracefully\n      tlbmc: Remove unused #include directives\n      Add support for MP29436 PSU sensor in tlbmc.\n\nFusion-Link: fusion2 N/A\nTested: N/A\nGoogle-Bug-Id: 515437010\nGoogle-Bug-Id: 521979698\nGoogle-Bug-Id: 557231843\nGoogle-Bug-Id: 559754428\nGoogle-Bug-Id: 566915949\nGoogle-Bug-Id: 572540383\nGoogle-Bug-Id: 574233913\nGoogle-Bug-Id: 574561109\nGoogle-Bug-Id: 576323466\nGoogle-Bug-Id: 576678366\nGoogle-Bug-Id: 576805857\nGoogle-Bug-Id: 576982516,576982769\nGoogle-Bug-Id: 578163043\nGoogle-Bug-Id: 579053156\nGoogle-Bug-Id: 579509670\nGoogle-Bug-Id: 579513401\nGoogle-Bug-Id: 581052097\nChange-Id: Ie2d5209eb59bc3ba2636840b7194a4a96bd35086\nSigned-off-by: Sandesh \u003csandeshharidas@google.com\u003e\n"
    },
    {
      "commit": "1e4a58f2e23dddbea5212ac725aa008b723fb086",
      "tree": "b1013050ce78be9289183da65fb476cfb7cebddd",
      "parents": [
        "7d849eb526e3ef8301ee2c804713923df3550232"
      ],
      "author": {
        "name": "Sandesh",
        "email": "sandeshharidas@google.com",
        "time": "Thu Nov 05 21:19:32 2026 +0000"
      },
      "committer": {
        "name": "Sandesh",
        "email": "sandeshharidas@google.com",
        "time": "Thu Nov 05 21:19:33 2026 +0000"
      },
      "message": "g3-shared-libs master: SRCREV bump cd5f6a508f..b055c05286\n\nAnshul Nigham (1):\n      Fix unused-variable warnings in //third_party/ecclesia.\n\nChris Kennelly CA (1):\n      Stop assuming Protobuf descriptor `string_view`s are null-terminated.\n\ngBMC Team (14):\n      Fix 1 IncludeCleaner finding: * Used header \"platforms/gbmc/g3_shared_libs/subprojects/gbmc_settings/gbmc_settings.proto.h\" is not included directly\n      Automatic clean-up of unused code by go/sensenmann. Cleanup reasons: go/code-deletion-why.\n      Add GPUViewCPULimitWatts as a new ecclesia Redfish property definition\n      Enable RedPath deep cache on [REDACTED] Borglet query router with selective leaf power sensor freshness.\n      BEGIN_PUBLIC Add gbmc_structured_log header-only library to g3_shared_libs END_PUBLIC\n      Wire gbmc_structured_log into the g3_shared_libs meson build\n      Add a payload schema and a metric/event registry to gbmc_structured_log\n      Delete szl_api_option from all proto files in google3 (no-op).\n      Add SettingsManager::ClearBiosSettings() to clear selected BIOS settings.\n      gbmc_structured_log: GBMC_LOG_* macros, per-metric timestamps, batching\n      Keep BootOptions across tray power cycles.\n      Remove SettingsManager::ClearAllBiosSettings().\n      tlbmc: Treat NaN shm sensor values as unavailable\n      gbmc_structured_log: Add GBMC_STRUCTURED_LOG_ENABLED switch, off by default\n\nFusion-Link: fusion2 N/A\nTested: N/A\nGoogle-Bug-Id: 566915949\nChange-Id: I0a7e3e356583be38a14637769d042e67280120c5\nSigned-off-by: Sandesh \u003csandeshharidas@google.com\u003e\n"
    },
    {
      "commit": "7d849eb526e3ef8301ee2c804713923df3550232",
      "tree": "f6a513bd412e5f1cb3e5f4f3713a2c99e2521917",
      "parents": [
        "876c1ffb707b4ccc56551ac8f2e513591a728212"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Thu Nov 05 06:45:42 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Thu Nov 05 10:51:53 2026 -0800"
      },
      "message": "systemd: Skip spawning TTY password agent if missing\n\nWhen systemd-tty-ask-password-agent is omitted from the target image,\ninteractive systemctl and machinectl invocations on a TTY still attempt\nto fork and exec /usr/bin/systemd-tty-ask-password-agent, printing an\nENOENT error to /dev/tty. Check whether the binary is executable before\nforking the agent.\n\nFusion-Link: fusion2 n/a\nTested: Verified patch applies cleanly to systemd 256.7.\nGoogle-Bug-Id: 580947649\nChange-Id: If7e4401a07a5ee40db75a36cd5988f1612e2e046\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "876c1ffb707b4ccc56551ac8f2e513591a728212",
      "tree": "d53f45deee4a6cdf3a4c50193339a05c04d224af",
      "parents": [
        "dec33d47b1195dcc77d6103628fedd2bd48f0180"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Mon Nov 02 21:36:25 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Thu Nov 05 10:51:35 2026 -0800"
      },
      "message": "libgpiod: Backport disappearing chip iteration fix\n\nWhen hot-pluggable USB-backed GPIO chips (such as Silicon Labs CP2105\nor FTDI FT232R USB-to-GPIO bridges) are removed during concurrent USB\nhost controller unbind/reset operations, libgpiod v1.6.4\ngpiod_chip_iter_new() and gpiod_line_find() abort with ENOENT or\nENODEV, causing parallel GPIO callers to fail.\n\nBackport upstream libgpiod v1.6.x commit 40f18f16f573 (\"core:\ngracefully handle disappearing chips during iteration\") so chip\niteration skips disappearing /dev/gpiochip* nodes and gpiod_line_find()\nignores ENODEV.\n\nTested: Applied with --fuzz\u003d0 against libgpiod-1.6.4.tar.xz, built and\ndeployed gBMC images across AST2600 and AST2750 multi-host platforms,\nand verified 0 gpiofind failures across 196/196 tight-loop iterations\nduring live FTDI FT232R USB gpiochip hot-unplug/re-enumeration and\n770/770 iterations across concurrent AST2750 EHCI unbind/re-bind\ncycles.\n\nFusion-Link: fusion2 n/a\nGoogle-Bug-Id: 580895287\nGoogle-Bug-Id: 574912243\nChange-Id: Ic041d75ab2ca9d04718d7696a8e6f1008a7f3fcb\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "dec33d47b1195dcc77d6103628fedd2bd48f0180",
      "tree": "92aa2fa9a6d1ccd1f6461f49516df83c3c06d272",
      "parents": [
        "861535a8c76bcb22b78b0da956cb1f7287c065a3"
      ],
      "author": {
        "name": "Kenneth Lamar",
        "email": "kennethlamar@google.com",
        "time": "Thu Nov 05 00:18:43 2026 +0000"
      },
      "committer": {
        "name": "Luka Strižić",
        "email": "lstrz@google.com",
        "time": "Thu Nov 05 10:16:19 2026 -0800"
      },
      "message": "pldm: pldm-scripts: Add --force to pldm_update_fw.sh\n\nA PLDM firmware device compares the comparison stamp of the component it\nis offered with the one it runs and declines the UpdateComponent unless\nthe stamp is higher, so the same bundle cannot be re-flashed and a\ndevice cannot be downgraded. DSP0267 lets the update agent ask the\ndevice to bypass that policy (\"Request Force Update\", UpdateOptionFlags\nbit 0), and pldmd now exposes that request as the ForceUpdate property\nof xyz.openbmc_project.Software.UpdatePolicy on\n/xyz/openbmc_project/software (gbmc/pldm change\nIb535803f707ad0992475735055fdd6078f80b22d). Until now the only way to\nget the bit on the wire was to edit the released .fwpkg.\n\nAdd a --force option that sets the property before the bundle is staged\nand resets it as soon as pldmd has created the software object for the\nbundle (pldmd samples the policy when it processes the package), on\nevery failure path in between, and when the run is interrupted. The\nproperty is sticky and shared, so it is set explicitly - true or false -\nbefore every staging, which also clears a value left behind by an\ninterrupted run. Without --force nothing changes for pldmd builds that\ndo not implement the interface; with --force such a build is refused up\nfront instead of running an update the devices would decline.\n\nTested: Evaluated on a pldmd that carries\nIb535803f707ad0992475735055fdd6078f80b22d, switch MCU running\n0033.01.0355.0000, bundle 0033.01.0344.0000:\n- yudlsqd-oy5 (provisioned MCU): pldm_update_fw.sh --force\nexits 0; ForceUpdate reads true while the bundle is staged\nand false again 8 s later once the software object exists;\npldmd logs FORCE_UPDATE\u003dTrue for the package and the\nUpdateComponent, the MCU accepts the downgrade\n(UpdateOptionFlagsEnabled\u003d0x1), verify and apply complete,\nthe Activation reaches Active after 10 s and the MCU reports\nthe pending version 0033.01.0344.0000. The object is removed\nand the staging directory is empty afterwards.\n- mves9 (unprovisioned DVT MCU): same up to the transfer, then\nthe MCU fails VerifyComplete with 0x93 (its own anti-rollback\ncheck); the existing failure handling removes the object,\ncancels the device, retries 3x and exits 1 with\nForceUpdate\u003dfalse, no object, an empty staging directory and\nthe MCU IDLE.\n- Without --force the MCU declines the same bundle (comparison\nstamp lower) and pldmd fails the attempt at once. An unknown\noption exits 1 with the usage line.\n- Mock unit tests verify single-dash flags (-f, -h) exit 1 and\nD-Bus timeouts on ForceUpdate fail rather than being ignored.\n\nFusion-Link: fusion2,n/a\nGoogle-Bug-Id: 479201442, 581968602\nChange-Id: Ibf5f1ef3e450e0fc9ffafaaab4de7eecbb7a2761\nSigned-off-by: Kenneth Lamar \u003ckennethlamar@google.com\u003e\n"
    },
    {
      "commit": "861535a8c76bcb22b78b0da956cb1f7287c065a3",
      "tree": "f75ace43af21e6a42383eff856d4606f0a6722fb",
      "parents": [
        "c5998806607ff00d4aebd65e652d035636c8ad8e"
      ],
      "author": {
        "name": "Kenneth Lamar",
        "email": "kennethlamar@google.com",
        "time": "Thu Nov 05 00:23:38 2026 +0000"
      },
      "committer": {
        "name": "Luka Strižić",
        "email": "lstrz@google.com",
        "time": "Thu Nov 05 10:16:19 2026 -0800"
      },
      "message": "pldm: pldm-scripts: Clean up after failed or interrupted updates\n\nAn update through pldm_update_fw.sh involves state in three places:\nthe software object pldmd creates for the staged package, the PLDM\nfirmware devices pldmd puts into update mode, and the bundle in the\nRAM-backed staging directory /tmp/pldm-images. The script only ever\ncleaned up the first, and only when things went well, so a failed,\nstuck or interrupted update left state behind that blocked every later\nupdate of any device until pldmd was restarted or the BMC rebooted.\n\nSoftware objects stuck in Activating\n\n  pldmd removes the Object.Delete interface from the package object\n  when its activation starts and only restores it on Active or Failed.\n  An activation that never completes - the device stopped requesting\n  firmware data, or reported a failed TransferComplete, VerifyComplete\n  or ApplyComplete, which pldmd logs without failing the activation -\n  therefore leaves an undeletable object behind, and pldmd rejects\n  every later bundle (\"already in progress\") while it exists. The\n  script\u0027s pre-update cleanup looked for objects implementing\n  Object.Delete and so skipped it (\"Clean slate\"); its post-attempt\n  cleanup only deleted when the interface was present.\n\n  Look for leftovers implementing Software.Activation without\n  Software.Version instead (persistent inventory objects implement\n  Software.Version - downstream device objects such as optical\n  transceivers implement both - and stay untouched), and add\n  delete_software_object(), which forces an object without Delete to\n  Failed - pldmd re-exposes Delete on that transition - and then\n  deletes it.\n\nActivations that can no longer complete\n\n  Give an activation up at the next poll once pldmd logs that a device\n  reported a failed update step (\"Transfer of the component failed\",\n  \"Component verification failed\", \"Component apply failed\"): pldmd\n  neither fails the activation nor retries after these.\n\n  pldmd exposes no transfer progress on D-Bus and never times a device\n  out, so a device that silently stops requesting data (e.g. it\n  declined the component) kept the attempt waiting for the whole\n  budget. Once an object has been Activating for a minute, each poll\n  looks for evidence that the update is still progressing: pldmd\u0027s\n  transfer trace since the previous poll, MCTP transmit traffic well\n  above the idle baseline measured before the activation, or a matched\n  device reporting VERIFY, APPLY or ACTIVATE (the phases it works\n  through on its own; a device sitting in DOWNLOAD without requesting\n  data is stuck, not busy). Both the packet and the byte transmit\n  counters are watched because the mctp-usb driver under-reports\n  tx_bytes (~12 bytes per packet, so an 8-NIC transfer looks like\n  sensor polling) while its tx_packets count is exact. The evidence is\n  sparse for a slow device (pldmd logs its RequestFirmwareData trace\n  once per MiB per device, every ~50 s for a NIC behind an MCTP\n  bridge), and a false \"stuck\" verdict aborts a healthy transfer\n  whereas a late one only delays the retry, so five minutes without\n  any evidence are required before the activation is declared stuck,\n  forced to Failed and deleted. An activation that is still\n  progressing when the attempt runs out of time is rejoined by the\n  next attempt instead of being re-staged; one that outlives the last\n  attempt is left to pldmd and removed by the next invocation.\n\nDevices left in update mode\n\n  Deleting the object only drops pldmd\u0027s side of the update. A device\n  left in update mode answers the RequestUpdate of the next attempt\n  with ALREADY_IN_UPDATE_MODE (CC 0x81) until its DSP0267 timeout\n  expires, which fails that attempt too. Send CancelUpdate to the\n  devices pldmd matched to the bundle (taken from its journal) whenever\n  an attempt is given up, and repeat it until they report IDLE: a\n  device acknowledges CancelUpdate before it has finished aborting.\n  Devices are also left behind without any object - by an earlier\n  invocation, or by pldmd being restarted (killed, crashed)\n  mid-transfer - so before staging a bundle, ask every PLDM endpoint\n  for its GetStatus state and cancel those in update mode. Endpoints of\n  other MCTP message types (the NVMe-MI drives on the same network) and\n  devices that do not answer are left alone: neither can be in a PLDM\n  update, and waiting for them to report IDLE would never end.\n\nBundles left in the staging directory\n\n  pldmd only unlinks the package file of an object it created and\n  later deleted; bundles it rejected (invalid header or size, no\n  matching device) or ignored (no device discovered yet) stay in\n  /tmp/pldm-images, and the script copied a fresh one there on every\n  attempt, leaking BMC memory (~10 MiB per NIC bundle) until reboot.\n  Clear the directory before staging and release what a failed attempt\n  staged on every exit path, deleting the object first (pldmd keeps a\n  rejected package open until then). Since the directory is now\n  cleared, refuse a bundle that is given from inside it: it would be\n  deleted before it could be copied.\n\nInterrupted runs\n\n  The script installed no signal handler, so Ctrl-C, \"systemctl stop\"\n  of a unit running it or a lost session left all of the above behind.\n  Trap SIGINT, SIGTERM and SIGHUP: delete the object of the activation\n  in progress (so that pldmd stops streaming), cancel the matched\n  devices (or, if none are known yet, whatever device is in update\n  mode), remove the staged bundle and re-raise the signal so the caller\n  sees the usual 128 + N status. bash runs the handler once the running\n  foreground command returns, i.e. within one 10 s poll. Ignore SIGPIPE\n  so that losing the reader of the script\u0027s output does not kill it in\n  the middle of a transfer or of this cleanup. (A run started as a\n  background job of a non-interactive shell has SIGINT ignored by POSIX\n  rules, which a trap cannot undo; SIGTERM works in every case.)\n\nConcurrent runs\n\n  Nothing stopped two runs from overlapping (an operator and a\n  provisioning job, say), and with the cleanup above each run would\n  delete the other\u0027s object, cancel its devices and wipe its bundle.\n  Hold an exclusive flock on /run/lock/pldm_update_fw.lock for the\n  lifetime of the run and refuse to start, with a clear message and\n  exit status 1, while another run holds it; the kernel releases the\n  lock however the holder exits.\n\nTested: on mves9-bmc-01 (DVT tray; its switch MCU, EID 14, has a blank\nCFPA and deterministically fails verification of the 0386 bundle with\n0x93) and yudlsqd-oy5-bmc-01 (production tray) with this script\ninstalled over gbmc-release-10.14, and on yudlsqd-oy4-bmc-01 over\ngbmc-release-11.05 with downstream optical transceiver inventory\nobjects present:\n- test_cl118804.sh (https://paste.googleplex.com/6755399435979600), a\n  self-checking script that runs the scenarios of this change on the\n  BMC - orphaned file and directory in the staging directory, a bundle\n  given from inside it, a concurrent run, SIGTERM before staging and\n  30 s into an 8-NIC transfer, the switch MCU refusing its update, a\n  run SIGKILLed 30 s into an 8-NIC transfer followed by a recovering\n  run, pldmd SIGKILLed 30 s into an 8-NIC transfer - and prints every\n  command, the script\u0027s output and a PASS/FAIL per check: 68 passed,\n  0 failed. https://paste.googleplex.com/6755399435979598\n- Switch MCU downgrade declined by the device (lower comparison stamp,\n  no error reported) on yudlsqd-oy5: the stall watchdog (\"No sign of\n  transfer progress\") caught it, the object was cleaned up and\n  CancelUpdate sent. https://paste.googleplex.com/6755399435925430\n- 8 NICs, upgrade 82.48.1688 -\u003e 82.48.1908 and forced downgrade,\n  ~500-550 s each: the packet rate (~670 packets/s against a ~340-380\n  packets/s threshold) counts as progress on every poll; no idle poll\n  was recorded. An earlier revision without the pre-flight CancelUpdate\n  saw the RequestUpdate of the run after a pldmd crash answered with CC\n  0x81 (ALREADY_IN_UPDATE_MODE).\n  https://paste.googleplex.com/6755399435931528,\n  https://paste.googleplex.com/6755399435931530\n- SIGHUP during a switch MCU activation: same cleanup as SIGTERM, exit\n  status 129. https://paste.googleplex.com/6755399435931535\n- Bundle that matches no device (other PSID), three attempts: object\n  deleted and /tmp/pldm-images empty on exit, where previously a\n  9.6 MiB file and an Invalid object remained.\n  https://paste.googleplex.com/6755399435882044\n- The stale-device scan queries only the 15 PLDM endpoints (~4 s per\n  attempt); the 4 NVMe-MI endpoints on the same MCTP network are\n  skipped.\n- shellcheck reports nothing.\n\nFusion-Link: fusion2,n/a\nGoogle-Bug-Id: 479201442\nChange-Id: I14bee82ed4de4b59b8fb944d0f5192a32821393e\nSigned-off-by: Kenneth Lamar \u003ckennethlamar@google.com\u003e\n"
    },
    {
      "commit": "c5998806607ff00d4aebd65e652d035636c8ad8e",
      "tree": "ffb6d7adfaed618baac9ce006ac86b9a959e3616",
      "parents": [
        "ffc2aa2b6c63b41774bbec26286c41aee7c86f3f"
      ],
      "author": {
        "name": "Kenneth Lamar",
        "email": "kennethlamar@google.com",
        "time": "Thu Oct 15 19:35:28 2026 +0000"
      },
      "committer": {
        "name": "Luka Strižić",
        "email": "lstrz@google.com",
        "time": "Thu Nov 05 10:16:19 2026 -0800"
      },
      "message": "pldm: pldm-scripts: Harden the wait loops of pldm_update_fw.sh\n\npldm_update_fw.sh stages a PLDM firmware update package for pldmd,\nwaits for the software object pldmd creates for it, requests the\nactivation and waits for it to complete, retrying the whole sequence up\nto three times. Several assumptions in those two wait loops do not hold\non real hardware; each turned a recoverable situation into a lost\nattempt or an undiagnosable failure.\n\nHandle Activating/Active states before activation\n\n  The script assumed that a freshly discovered software object is\n  always Ready and failed the attempt when it found the object already\n  Activating or Active. Both states are legitimate: a previous attempt\n  may have requested the activation and lost track of it (e.g. its\n  status poll failed), and pldmd builds without inotify package\n  monitoring activate a package as soon as it is staged. Join an\n  in-progress activation by proceeding straight to the completion wait,\n  treat an Active object as a completed update, and only fail the\n  attempt for states that cannot complete (Invalid, Failed, ...),\n  logging the state. When looking for the staged package object, skip\n  persistent firmware inventory objects by excluding Software.Version:\n  downstream device inventory objects (such as optical transceivers)\n  implement both Software.Version and Software.Activation with\n  Activation\u003dActive, and would otherwise be mistaken for a completed\n  package update.\n\nTolerate transient status read failures while waiting\n\n  pldmd serves the firmware transfer (RequestFirmwareData) from the\n  same single-threaded event loop that answers D-Bus, so a status read\n  can time out while a transfer is running. The script treated the\n  empty result as an unexpected state and started the next attempt\n  while the previous transfer was still running in pldmd. Treat an\n  empty/null status as a read failure instead: keep polling as long as\n  the same pldmd instance is running, and give up only after three\n  consecutive unreadable polls, when pldmd is no longer active, or as\n  soon as its main PID has changed (sampled before triggering\n  activation, as systemd restarts a crashed pldmd within a second and\n  the new instance knows nothing of the activation).\n\nPoll at a fixed interval\n\n  Both wait loops slept for exponentially growing periods (up to 60 s\n  for the object, up to 512 s for the activation), so after a couple of\n  minutes the script was blind for 4 and then 8 minutes at a time: no\n  status, no journal, no reaction to a completed, failed or stuck\n  activation, to pldmd dying, or to a termination signal (bash defers\n  traps until the running sleep returns). Poll every 5 s for up to\n  120 s for the object and every 10 s for up to 1020 s (the previous\n  17 minute budget) for the activation, and anchor the incremental\n  pldmd journal printed after each poll to the time of the previous\n  poll so that consecutive windows stay contiguous however long a poll\n  takes.\n\nDo not consume a retry when starting pldmd\n\n  When pldmd.service was not running, the script started it, slept\n  60 s and unconditionally restarted the retry loop, losing one of the\n  three attempts although pldmd was up. A fixed delay is also the wrong\n  tool: pldmd silently ignores a bundle staged before it has discovered\n  any PLDM firmware device, fails association with \"No matching devices\n  found\" if a slower endpoint is still being discovered, and discovery\n  takes seconds on a warm tray but minutes after a BMC boot. Whenever\n  pldmd exposes no software object yet or was (re)started within the\n  5 minute discovery window, wait for its software object count to\n  stabilize across polls before staging; if an attempt still fails with\n  \"No matching devices found\" while pldmd is within that startup\n  window, wait for additional devices to appear and settle before\n  retrying; only give the attempt up up front if pldmd is not running.\n\nSurface pldmd\u0027s reason when a bundle is rejected\n\n  pldmd keeps the reason a bundle was rejected (unparsable header,\n  package size mismatch, no device matching the descriptors, no device\n  discovered yet) in its journal only; the script printed \"is not ready\n  to activate\" or \"not recognized by pldmd within the timeout\", so MSV\n  post-install logs could not tell these failure modes apart. Dump the\n  pldmd journal covering the discovery window whenever the bundle is\n  not discovered in time or the discovered object cannot be activated.\n\nStop waiting once pldmd has rejected the bundle\n\n  pldmd rejects some bundles without creating a software object (a\n  file it cannot open or that is too short for a package header, an\n  unknown header format UUID, any bundle staged while another package\n  is Activating), logging \"Error (-1) processing image \u003cpath\u003e\" and\n  removing the file. The script nevertheless waited the full discovery\n  timeout on each attempt, so an empty or corrupt bundle took over five\n  minutes to fail. Record when the bundle was staged and give the\n  attempt up as soon as that line shows up in the pldmd journal.\n\nThe earlier patchsets of this change also trimmed vendor signature\ntrailers from DSP0267 packages on the BMC. That is now done at Blaze\nbuild time when the MSV package is assembled (google3\n//platforms/gpus/tools:pldm_fwpkg_trim, cl/998974461) and has been\ndropped from the script.\n\nTested: on mves9-bmc-01 (DVT tray) and yudlsqd-oy5-bmc-01\n(pre-provisioned production tray) with this script installed over\ngbmc-release-10.14, and on yudlsqd-oy4-bmc-01 over gbmc-release-11.05\nwith downstream optical transceiver inventory objects present. The\ndevices are the PCIe switch MCU (MCXN556), four NIC SMA MCUs and eight\nNICs behind four MCTP bridges.\n- test_cl108585.sh (https://paste.googleplex.com/6755399435978667), a\n  self-checking script that runs the scenarios of this change on the\n  BMC - typical SMA MCU update, zero-byte, random and truncated\n  bundles, pldmd stopped beforehand, a failed status read injected\n  during an 8-NIC transfer, pldmd killed during an 8-NIC transfer - and\n  prints every command, the script\u0027s output and a PASS/FAIL per check:\n  57 passed, 0 failed. https://paste.googleplex.com/6755399435978664\n- Switch MCU upgrade 0033.01.0355.0000 -\u003e 0033.01.0386.0000 on\n  yudlsqd-oy5: object discovered after 5 s, transfer 3 s, verify and\n  apply passed, Active after 12 s, rc\u003d0 (baseline restored afterwards).\n  https://paste.googleplex.com/6755399435924338\n- 8 NICs on mves9 upgraded 82.48.1688 -\u003e 82.48.1908 and downgraded\n  again with a ForceUpdate bundle, ~500-550 s each, rc\u003d0, status and\n  pldmd journal reported every 10 s.\n  https://paste.googleplex.com/6755399435971664 (upgrade),\n  https://paste.googleplex.com/6755399435971822 (downgrade)\n- An attempt that ran out of time while the object was still\n  Activating was followed by one that found it Activating and joined\n  it (\"Activation already in progress\") instead of re-staging.\n  https://paste.googleplex.com/6755399435882047\n- pldmd frozen with SIGSTOP for ~40 s during a wait: \"Failed to read\n  the Activation status ... (1/3); pldmd is still running, retrying.\";\n  the wait resumed on the same activation after SIGCONT.\n  https://paste.googleplex.com/6755399435882154\n- Bundle for a different PSID: the journal dump shows \"No matching\n  devices found with the PLDM firmware update package\"; bundle with\n  1024 bytes appended: \"Package size does not match calculated package\n  size\". https://paste.googleplex.com/6755399435882044,\n  https://paste.googleplex.com/6755399435882049\n- shellcheck reports nothing.\n\nFusion-Link: fusion2,n/a\nGoogle-Bug-Id: 479201442\nChange-Id: I90236473924a44cfca71cacac86de6b8d8af2647\nSigned-off-by: Kenneth Lamar \u003ckennethlamar@google.com\u003e\n"
    },
    {
      "commit": "ffc2aa2b6c63b41774bbec26286c41aee7c86f3f",
      "tree": "7adedaa62f50d29185b1189d92b4651e535c757b",
      "parents": [
        "f6f9c557d97b052e1dfa58c6d9e191c18fb83c3b"
      ],
      "author": {
        "name": "Sam Agazaryan",
        "email": "samagazaryan@google.com",
        "time": "Thu Nov 05 06:42:28 2026 +0000"
      },
      "committer": {
        "name": "Sam Agazaryan",
        "email": "samagazaryan@google.com",
        "time": "Thu Nov 05 06:44:47 2026 +0000"
      },
      "message": "linux gbmc-6.18-ast27xx: SRCREV bump bdf539a9f7..9962c14a6f\n\nSam Agazaryan (2):\n      i3c: master: Drain IBI slot pool before freeing\n      i3c: master: Preserve IBI workqueue across DAA\n\nTested: Built an image with both patches applied on top of bdf539a9f7\nand booted it on AST27xx hardware. fusion2 n/a\nGoogle-Bug-Id: 572553250\nChange-Id: Iae8b388ca8693aa1da33383b5876781780a45c1e\nSigned-off-by: Sam Agazaryan \u003csamagazaryan@google.com\u003e\n"
    },
    {
      "commit": "f6f9c557d97b052e1dfa58c6d9e191c18fb83c3b",
      "tree": "df675bce25f8dc0806a3f22d54f62f8953045967",
      "parents": [
        "dbb5eebc1ef485e6448f28e7d540ed3309703efd"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Thu Nov 05 00:29:01 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Wed Nov 04 22:14:29 2026 -0800"
      },
      "message": "linux-gbmc: Enable CONFIG_KALLSYMS\n\nBecause linux-gbmc builds with --allnoconfig and CONFIG_EXPERT\u003dy,\nCONFIG_KALLSYMS defaults to disabled in non-dev builds, leaving kernel\ncrash backtraces without function names. Enable CONFIG_KALLSYMS\u003dy in\ngbmc_defconfig so all images print symbolic stack traces.\n\nFixes: 9f3286057dc9 (\"linux-gbmc: Use EXPERT config\")\nTested: Built kernel and verified CONFIG_KALLSYMS\u003dy in .config\nFusion-Link: fusion2 N/A\nGoogle-Bug-Id: n/a\nChange-Id: Ib5222c3a9b1ab3feb9adca1be81cf376509da077\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "dbb5eebc1ef485e6448f28e7d540ed3309703efd",
      "tree": "0f18198912dcdd99e47a9e2173ac9e2f3830ed38",
      "parents": [
        "d2c273c7a647b76096967bebb65bbfacbb47d471"
      ],
      "author": {
        "name": "Bart Van Assche",
        "email": "bvanassche@google.com",
        "time": "Wed Nov 04 14:00:49 2026 -0800"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Wed Nov 04 22:14:29 2026 -0800"
      },
      "message": "linux-gbmc: Enable CONFIG_DEBUG_LIST\n\nWithout CONFIG_DEBUG_LIST, linked list corruption, e.g. caused by a\nuse-after-free or by a stray memory write, is not detected when it\nhappens. Instead, such corruption causes a crash later on, typically\nin code that is not related to the code that caused the corruption.\nEnable CONFIG_DEBUG_LIST such that the list_add() and list_del()\nfunctions verify the list pointers and report corruption when list\nentries are added or removed. This makes memory corruption easier to\ndiagnose. The reports are printed via WARN() and hence require\nCONFIG_BUG\u003dy.\n\nTested: Mimik\nGoogle-Bug-Id: 581702595\nChange-Id: Icc6c20fe03fe92bcb57f43487b4317c9f1205c26\nSigned-off-by: Bart Van Assche \u003cbvanassche@google.com\u003e\n"
    },
    {
      "commit": "d2c273c7a647b76096967bebb65bbfacbb47d471",
      "tree": "9808929e3b6e559e05e48bd728e862a3177cb61c",
      "parents": [
        "6e78f114069dde9d8be91592b603b8ee60dccf91"
      ],
      "author": {
        "name": "Bart Van Assche",
        "email": "bvanassche@google.com",
        "time": "Wed Nov 04 14:00:49 2026 -0800"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Wed Nov 04 22:14:29 2026 -0800"
      },
      "message": "linux-gbmc: Enable CONFIG_DEBUG_BUGVERBOSE\n\nWithout CONFIG_DEBUG_BUGVERBOSE, BUG() and WARN() reports only include\nthe function name and offset of the BUG() or WARN() statement.\nTranslating that information into a source code location requires\nthe vmlinux file with debug information that matches the kernel that\nreported the problem. Enable CONFIG_DEBUG_BUGVERBOSE such that BUG()\nand WARN() reports include the file name and line number of the BUG()\nor WARN() statement. This makes it easier to analyze kernel problems\nfrom the kernel log alone.\n\nTested: Mimik\nGoogle-Bug-Id: 581702595\nChange-Id: Iae430bc7a1a903474ac02a6ca8b1d66e8ca49e52\nSigned-off-by: Bart Van Assche \u003cbvanassche@google.com\u003e\n"
    },
    {
      "commit": "6e78f114069dde9d8be91592b603b8ee60dccf91",
      "tree": "03d8d49e921535049c98c62056a77e58c61ea182",
      "parents": [
        "941701ec4ca73da2561c080513e3c12a530aba82"
      ],
      "author": {
        "name": "Bart Van Assche",
        "email": "bvanassche@google.com",
        "time": "Wed Nov 04 13:42:46 2026 -0800"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Wed Nov 04 22:14:29 2026 -0800"
      },
      "message": "linux-gbmc: Enable CONFIG_BUG\n\nThe gBMC kernels are configured with KCONFIG_MODE\u003d\"--allnoconfig\" and\nCONFIG_EXPERT\u003dy. With CONFIG_EXPERT\u003dy, CONFIG_BUG becomes a visible\noption and hence allnoconfig disables it. Without CONFIG_BUG, WARN(),\nWARN_ON() and WARN_ONCE() evaluate their condition but print nothing.\nThis silences, among others, the refcount_t saturation warnings and\nall other kernel consistency checks that report problems via WARN*().\nAs an example, on NPCM7xx BMCs that crashed in unix_stream_recvmsg()\nbecause of a corrupted sk_buff, skb_get() silently saturated a zero\nreference count instead of reporting \"refcount_t: addition on 0;\nuse-after-free\". That warning would have identified the corruption\nearlier.\n\nEnable CONFIG_BUG such that WARN*() reports are printed again.\n\nFixes: 9f3286057dc9 (\"linux-gbmc: Use EXPERT config\")\nTested: Mimik\nGoogle-Bug-Id: 581702595\nChange-Id: I6273e904a38887bb5ec59b9c0c7d684286fe6277\nSigned-off-by: Bart Van Assche \u003cbvanassche@google.com\u003e\n"
    },
    {
      "commit": "941701ec4ca73da2561c080513e3c12a530aba82",
      "tree": "34932bdd0b52a99ef6cd868ed13c47f66c0ec689",
      "parents": [
        "adb9a921981bbe66ca9cb1c67b2cfa0988f60958"
      ],
      "author": {
        "name": "Bart Van Assche",
        "email": "bvanassche@google.com",
        "time": "Wed Oct 28 07:59:22 2026 -0700"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Wed Nov 04 22:14:29 2026 -0800"
      },
      "message": "linux-gbmc: Enable the vDSO on 32-bit ARM\n\nCONFIG_VDSO defaults to \u0027y\u0027 only if CONFIG_ARM_ARCH_TIMER\u003dy. The\nNPCM7xx BMCs have a Cortex-A9 CPU without an ARM architected timer,\nso the vDSO is disabled on these BMCs and every clock_gettime() call\nis a system call. perf shows that clock_gettime() system calls\nconsume a significant fraction of all CPU cycles on these BMCs.\n\nEnable CONFIG_VDSO on all 32-bit ARM BMCs. Only the NPCM7xx BMCs are\naffected since the AST2600 BMCs have an ARM architected timer and\nhence already have CONFIG_VDSO\u003dy. Restrict this setting to armv7\nbecause CONFIG_VDSO does not exist on arm64 and because\nConfigValidator.py reports an error for options that are missing from\nthe generated .config.\n\nThis change alone does not eliminate the clock_gettime() system calls\non NPCM7xx BMCs: if the architected timer is not usable, the kernel\nremoves __vdso_clock_gettime(), __vdso_clock_gettime64() and\n__vdso_gettimeofday() from the vDSO at boot time. This change is a\nprerequisite for a kernel patch that keeps __vdso_clock_gettime() and\n__vdso_clock_gettime64() such that CLOCK_REALTIME_COARSE and\nCLOCK_MONOTONIC_COARSE can be read without a system call.\n\nTested: Verified that CONFIG_VDSO\u003dy in the kernel .config and ran a test\nin Mimik. Verified that ConfigValidator.py accepts the arm64 kernel\n.config.\n\nTested: Mimik\nGoogle-Bug-Id: 576323466\nChange-Id: I1cdba83ebb042da26f42e4acc648df5bf5a57796\nSigned-off-by: Bart Van Assche \u003cbvanassche@google.com\u003e\n"
    },
    {
      "commit": "adb9a921981bbe66ca9cb1c67b2cfa0988f60958",
      "tree": "e2a2df67f1ef61b9bd2621477276255b7e8c0742",
      "parents": [
        "dfe19515d64841d803ccbeaf88172770dfab1538"
      ],
      "author": {
        "name": "Guangzong Chen",
        "email": "guangzong@google.com",
        "time": "Tue Nov 03 19:06:15 2026 +0000"
      },
      "committer": {
        "name": "Guangzong Chen",
        "email": "guangzong@google.com",
        "time": "Wed Nov 04 19:44:06 2026 -0800"
      },
      "message": "libnvme: set req_hdr dlen and flags for set-features with payload\n\nPer NVMe-MI §6.1, when an NVMe Admin Command transfers data from the\nManagement Controller to the NVMe Storage Device (Request Data, such as\nSet Features with data payload), the NVMe-MI Admin Command Request\nHeader DLEN field must contain the transfer length, and Flags bit 0\n(DLEN valid) must be set to 0x1.\n\nPreviously, nvme_mi_admin_set_features() cleared req_hdr via\nnvme_mi_admin_init_req() and attached the payload buffer to\nreq.data / req.data_len, but never populated req_hdr.dlen or\nreq_hdr.flags. Consequently, on features with data payloads such as\nOcpLatencyMonitor (FID 0xC5, transferring a 4096-byte payload), the\nrequest header on the wire declared 0 payload bytes (72 bytes expected)\nwhile 4096 payload bytes were transmitted (4168 bytes wire size),\nviolating NVMe-MI §6.1 and causing NVMe-MI controllers to reject the\nrequest (e.g. MI status 0x4 Invalid Parameter) or time out over MCTP.\n\nAdd patch 0016-mi-set-req_hdr-dlen-and-flags-for-features-with-payload.patch\nto libnvme in meta-gbmc-staging to set req_hdr.dlen and req_hdr.flags\nin nvme_mi_admin_set_features() when payload data is present\n(args-\u003edata \u0026\u0026 args-\u003edata_len). Note that nvme_mi_admin_get_features()\ndoes not require this because Get Features is a read operation with no\nrequest data payload (req.data is NULL, req.data_len is 0).\n\nTested: Verified on target hardware: http://gpaste/6755399436670492\nGoogle-Bug-Id: 581863515\nFusion-Link: http://fusion2/ci/kokoro/prod%3Agbmc%2Ffirmware-build%2Ffirmware%2Fgerrit_upstreamed_platforms/activity/44e32f2c-e7dc-46ce-9be1-00228546be0c\nChange-Id: I057d292b39b1a685f5f88dd02adc2ef5ff92e0b1\nSigned-off-by: Guangzong Chen \u003cguangzong@google.com\u003e\nTAG\u003dagy\nCONV\u003dc1107d2e-0b48-46b8-80af-81fdf79d0aee\n"
    },
    {
      "commit": "dfe19515d64841d803ccbeaf88172770dfab1538",
      "tree": "84288c8090031250d8dc5af5ae4ed310c20ccc45",
      "parents": [
        "4fe1a2571ffe78345be10c9750fd9d02b543bcce"
      ],
      "author": {
        "name": "Yichao Wang",
        "email": "wangyic@google.com",
        "time": "Mon Nov 02 23:40:53 2026 +0000"
      },
      "committer": {
        "name": "Yichao Wang",
        "email": "wangyic@google.com",
        "time": "Wed Nov 04 13:32:25 2026 -0800"
      },
      "message": "pldm rde-support: SRCREV bump 48156129f7148..c3f0a87927699\n\nYichao Wang (1):\n      rded: set cache expiration interval to 120s\n\nFusion-Link: fusion2 N/A\nTested: https://paste.googleplex.com/6359170526347264\nGoogle-Bug-Id: 565437927\nGoogle-Bug-Id: 578129605\nChange-Id: I963c48d560376d20d46ec212952505696ce7ba94\nSigned-off-by: Yichao Wang \u003cwangyic@google.com\u003e\n"
    },
    {
      "commit": "4fe1a2571ffe78345be10c9750fd9d02b543bcce",
      "tree": "bd56d232cff64972a91810be9079a2835fe81fbc",
      "parents": [
        "0d4513f9fa059af9ab136a699e5155d74b06ce28"
      ],
      "author": {
        "name": "Hao Jiang",
        "email": "jianghao@google.com",
        "time": "Tue Nov 03 22:27:20 2026 +0000"
      },
      "committer": {
        "name": "Munawar Hussain",
        "email": "munawarhussain@google.com",
        "time": "Wed Nov 04 06:51:03 2026 -0800"
      },
      "message": "dbus-sensors google-dev: SRCREV bump b9e0c44f33..c0165b7fdb\n\nGuangzong Chen (1):\n      nvmed: reset featureStore in NVMeSubsystem::markFunctional(false)\n\nHao Jiang (1):\n      nvmesensor: bind drive Functional to Start status\n\nMunawar Hussain (1):\n      nvmed: Move NVMeMiWorker to AsioHelper\n\nFusion-Link: fusion2 N/A\nTested: N/A\nGoogle-Bug-Id: 512215974\nGoogle-Bug-Id: 565728782\nGoogle-Bug-Id: 574561109\nChange-Id: I12e844f68adb97755b556c93d150d932381574a9\nSigned-off-by: Hao Jiang \u003cjianghao@google.com\u003e\n"
    },
    {
      "commit": "0d4513f9fa059af9ab136a699e5155d74b06ce28",
      "tree": "aa18244a6cd7fa7efb894517edc453af31ae0a1b",
      "parents": [
        "5e6707d5cb91d5bd2f7303331e7da3ee822e1d9e"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Tue Nov 03 21:18:44 2026 +0000"
      },
      "committer": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Wed Nov 04 02:59:38 2026 +0000"
      },
      "message": "phosphor-software-manager: Disable mutable D-Bus\n\nAdd a patch to phosphor-software-manager that throws SdBusError(EPERM)\nfor mutable D-Bus methods and property setters (Activation,\nRequestedActivation, Priority, Delete, DeleteAll, Reset,\nFieldModeEnabled, and DownloadViaTFTP) and rejects image uploads in\nphosphor-version-software-manager and phosphor-image-updater while\npreserving read-only version inventory reporting.\n\nTested: Verified on hardware that read-only D-Bus/Redfish software\ninventory works and mutable D-Bus operations return EPERM.\nFusion-Link: fusion2 N/A\nGoogle-Bug-Id: 581089189\nChange-Id: Ia1d3f9659c5d33b7cda769646aeec7ac799bdbf5\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "5e6707d5cb91d5bd2f7303331e7da3ee822e1d9e",
      "tree": "cc5ebab2a3aaacc9e7d59723807357c262f20826",
      "parents": [
        "18c16b56e78fc1a93320222d144be51e1503be0b"
      ],
      "author": {
        "name": "Bart Van Assche",
        "email": "bvanassche@google.com",
        "time": "Fri Oct 30 10:58:53 2026 -0700"
      },
      "committer": {
        "name": "Bart Van Assche",
        "email": "bvanassche@google.com",
        "time": "Mon Nov 02 15:15:56 2026 -0800"
      },
      "message": "estoraged: Drop the shorter sdbusplus type alias reverts\n\nRemove the 0001- and 0002-Revert-sdbusplus-use-shorter-type-aliases\npatches and the SRC_URI and FILESEXTRAPATHS entries that add them from\nestoraged_%.bbappend. Keep the e2fsprogs runtime dependency changes.\n\nUpstream estoraged commit 3e0046420658 (\"sdbusplus: Use shorter type\naliases only if supported by sdbusplus\") makes eStoraged check at\nconfigure time whether sdbusplus defines sdbusplus::match and\nsdbusplus::object_path and fall back to sdbusplus::bus::match_t and\nsdbusplus::message::object_path otherwise. Hence the revert patches are\nno longer needed. That commit also rewrites the lines in src/main.cpp\nand include/getConfig.hpp that the revert patches modify, which makes\ndo_patch fail for every estoraged SRCREV that includes it:\n\n  Hunk #1 FAILED at 190.\n  1 out of 1 hunk FAILED -- rejects in file src/main.cpp\n\nDo this to unblock the estoraged SRCREV bumps starting from\n2ddabfb2ad..47bffe27e6.\n\nFixes: debca065049d (\"estoraged: Revert shorter sdbusplus type aliases\")\nTested: Mimik\nGoogle-Bug-Id: 564588832\nChange-Id: Ia3fc39f51e983067b4b45d83594f81bd2d4a1e28\nSigned-off-by: Bart Van Assche \u003cbvanassche@google.com\u003e\n"
    },
    {
      "commit": "18c16b56e78fc1a93320222d144be51e1503be0b",
      "tree": "c1e8fff898848976a24b4948876ae5d73abbdf3e",
      "parents": [
        "e0d90e02345150c020adad3cdbe06959c27c8720"
      ],
      "author": {
        "name": "Yolkee Huang",
        "email": "cweihuang@google.com",
        "time": "Thu Oct 15 19:19:03 2026 +0800"
      },
      "committer": {
        "name": "Luka Strižić",
        "email": "lstrz@google.com",
        "time": "Mon Nov 02 10:15:57 2026 -0800"
      },
      "message": "mctp-reactor: Defer USBMCTPDDevice sysfs lookup to setup()\n\nDuring BMC boot or USB controller rebind, cascading USB hub enumeration\nruns asynchronously in background kernel worker threads for 4 to 5\nseconds. When entity-manager publishes MCTPUSBDevice objects before the\nUSB network interface (mctpusb0/mctpusb1) finishes enumerating in sysfs,\nmctpreactor logs \"No net device associated with USB device\" and\npermanently drops the USB MCTP bridge endpoints and all downstream\nbridged NIC endpoints until mctpreactor restarts. This happens because\nUSBMCTPDDevice::from() resolves sysfs synchronously during object\nconstruction and returns nullptr on MCTPException, so\nMCTPReactor::manageMCTPDevice() ignores the device without adding it to\nthe 5-second deferred retry queue.\n\nStore the USB topology properties (rootHubPath, port, configuration, and\ninterfaceNum) in USBMCTPDDevice during from() and defer\nbusFromRootHubPath() and interfaceFromSysfs() resolution to\nUSBMCTPDDevice::setup(). When the USB sysfs node or network interface is\nnot yet populated, invoke the setup callback with\nstd::errc::resource_unavailable_try_again so\nMCTPReactor::setupEndpoint() queues the device in deferred and retries\nevery 5 seconds (matching BridgeDownsteamMCTPDDevice::setup()).\n\nTested:\n- Unit tests: Built and ran test_MCTPEndpoint (14/14 passed) and\n  test_MCTPReactor (6/6 passed).\n- Before: Triggering an EHCI controller unbind/bind before starting\n  mctpreactor permanently dropped both USB MCTP bridges and all 4\n  downstream bridged endpoints (0 endpoints on D-Bus).\n- After: Patched mctpreactor deferred setup while USB hubs enumerated\n  and automatically recovered both USB MCTP bridges and all 4 downstream\n  bridged endpoints on subsequent 5-second tick() cycles.\n- Raw log: http://b/571351241\n\nFusion-Link: fusion2, n/a\nGoogle-Bug-Id: 571351241\nChange-Id: I4f1d718157ff77c4802b71c67df917c28787a045\nSigned-off-by: Yolkee Huang \u003ccweihuang@google.com\u003e\n"
    },
    {
      "commit": "e0d90e02345150c020adad3cdbe06959c27c8720",
      "tree": "4f3092746828bac629a9407cc3bfd181ca2692c1",
      "parents": [
        "9d18410e35f3b3928f8219cb25b7d72850ae04d5"
      ],
      "author": {
        "name": "Ron Vered",
        "email": "ronvered@google.com",
        "time": "Mon Oct 26 19:12:45 2026 +0000"
      },
      "committer": {
        "name": "Luka Strižić",
        "email": "lstrz@google.com",
        "time": "Mon Nov 02 10:09:09 2026 -0800"
      },
      "message": "gbmcweb-headers: Depend on protobuf-native\n\nSince gbmcweb commit cf1cf6a7f7, the plugins-headers branch of its\nmeson.build installs the tlBMC headers listed in\ntlbmc/redfish/routes/tlbmc_plugin_headers.txt and\ntlbmc/collector/bulk_sensor_collector_headers.txt, and generates the\n.pb.h headers for the protos they depend on when protoc is available\n(it skips them with a warning otherwise). Add protobuf-native so every\nplatform gets the generated headers once GBMCWEB_SRCREV includes that\ncommit.\n\nLocally, bitbake gbmcweb-headers for humubeach against that gbmcweb\ntree installs the listed headers, the .proto sources and the generated\n.pb.h files, and libmetis builds against them.\n\nTested: N/A\nFusion-Link: fusion2 N/A\nGoogle-Bug-Id: 545329676\nChange-Id: Ie087875895b9bab009e71a2ce39540488f661cf9\nSigned-off-by: Ron Vered \u003cronvered@google.com\u003e\n"
    },
    {
      "commit": "9d18410e35f3b3928f8219cb25b7d72850ae04d5",
      "tree": "c51eee7dc34d572aae0cbe562f04eb72e3ac2759",
      "parents": [
        "1cf87b468697e1f87a5e08bf170961ac2123aed2"
      ],
      "author": {
        "name": "Ramon Rakow",
        "email": "rakow@google.com",
        "time": "Thu Oct 29 21:59:10 2026 +0000"
      },
      "committer": {
        "name": "Ramon Rakow",
        "email": "rakow@google.com",
        "time": "Mon Nov 02 10:00:55 2026 -0800"
      },
      "message": "grpc-blobs: SRCREV bump to f81c9aa501\n\ngBMC Team (1):\n      Map single-host port 10166 (instance 2) to host0 in\n      BootTimeServiceImpl.\n\nTAG\u003dagy\nCONV\u003d38317951-3299-4411-8479-871465bce27e\nTested: https://paste.googleplex.com/6755399436493322\nFusion-Link: fusion2 n/a\nSigned-Off-By: Ramon Rakow \u003crakow@google.com\u003e\nGoogle-Bug-Id: 476486003\nChange-Id: I2e3d34e5845751a2fd612b628d10af1f26aad857\n"
    },
    {
      "commit": "1cf87b468697e1f87a5e08bf170961ac2123aed2",
      "tree": "fe66ea329ae3dc6702922d6d8a4a47d1bd103bbb",
      "parents": [
        "389396f08b6e98e4418c14c0e5c1f7249d9ea1b7"
      ],
      "author": {
        "name": "Jeff Lien",
        "email": "jeff.cf.lien@fii-na.corp-partner.google.com",
        "time": "Tue Oct 27 17:19:11 2026 +0800"
      },
      "committer": {
        "name": "Alan Hung",
        "email": "yung-sheng.huang@fii-na.corp-partner.google.com",
        "time": "Sun Nov 01 16:58:53 2026 -0800"
      },
      "message": "meta-aspeed-sdk: bmc-pb: SRCREV bump 5531ad7..2d615ea\n\nNeal Liu (15)\n    ast2700a2: update subtitle\n    ast2755a0: update cptra fw with fixing warm reset problem\n    ast2700a2: update caliptra-fw\n    ast2700a2: add vendor signature\n    ast2700a1: update caliptra-fw\n    ast2700a2: update caliptra-fw to rt-1.2.5\n    ast2700a1: update caliptra-fw to rt-1.2.5\n    ast2700a2: add caliptra-fw-v2 with updated vendor LMS keys 16~31\n    ast1040a0: update cptra-fw\n    ast1040a0: update BootROM XIP prebuilt binary\n    rename(ast2705a0): rename prebuilts directory from ast2755a0 to ast2705a0\n    ast1040a0: update caliptra-fw based on 6980f7d\n    ast1040a0: update caliptra-fw\n    ast2705a0: update caliptra-fw\n    ast2700a2: update cptra-fw to rt-1.3.0\n\nNeal (1):\n    Revert “ast2700a1: update caliptra-fw to rt-1.2.5\n\nchengyu_lin (11):\n    ast2755a0: add ssmcu prebuilt binaries\n    ast1040a0: add ssmcu prebuilt binaries\n    ast2755a0: update SSMCU prebuilt binary\n    ast1040a0: update SSMCU prebuilt binary\n    ast2705a0: update SSMCU prebuilt binary\n    ast1040a0: update SSMCU prebuilt binary\n    ast2755a0: fix ssmcu-rom-xip.bin file permission\n    ast2705a0: fix stale AST2755 references and refresh SSMCU binaries\n    ast1040a0: update SSMCU prebuilt binary\n    ast1040a0: update SSMCU prebuilt binary\n    ast2705a0: update SSMCU prebuilt binary\n\nJheng Shuo Lin (7):\n    ast2755a0: Add ddr, dp_fw, ssp, tsp, uefi prebuilt\n    ast2700a1\u0026a2: Add ssp \u0026 tsp\n    rename(ast2705a0): rename boot MCU XIP ROM binary and fix stale AST2755 references\n    ast2705a0: update SSMCU prebuilt binary\n    ast1040a0,ast2705a0: update caliptra-fw with OCP-LOCK support\n    ast1040a0,ast2705a0: fix typo in bootmcu-rom-xip.bin filename\n    ast1040a0: update SSMCU prebuilt binary\n\nJammy Huang (6):\n    ast2700a1\u0026a2/2755a0: Update DP_FW\n    ast2755: dp_fw: Adjust TxClk from 1G to 500M\n    ast2755a0: update SSMCU prebuilt binary\n    Revert “ast2755: dp_fw: Adjust TxClk from 1G to 500M”\n    ast2755a0: update SSMCU prebuilt binary\n    ast2705a0: update SSMCU prebuilt binary\n\nYunHsuanChung(2):\n    ast2755a0: add BootROM XIP prebuilt image\n    ast1040a0: add BootROM XIP prebuilt binary\n\nTested: The boot log is available on the issue tracker page.\nGoogle-Bug-Id: 569799697\nSigned-off-by: Jeff Lien \u003cjeff.cf.lien@fii-na.corp-partner.google.com\u003e\nChange-Id: I060665e5fc569d0cc3f78232348aa0d2e9a191a8\n"
    },
    {
      "commit": "389396f08b6e98e4418c14c0e5c1f7249d9ea1b7",
      "tree": "66b7d61713b56bb87ba37f92642c15c6b7b60a93",
      "parents": [
        "f04bcea3dc92c8d00adfff502ad373681725506d"
      ],
      "author": {
        "name": "Edward Lee",
        "email": "edwarddl@google.com",
        "time": "Fri Oct 30 00:23:21 2026 +0000"
      },
      "committer": {
        "name": "Edward Lee",
        "email": "edwarddl@google.com",
        "time": "Fri Oct 30 14:09:19 2026 -0700"
      },
      "message": "gbmcweb master: SRCREV bump ce08346e42..f0506b3981\n\ngBMC Team (5):\n      Apply LinearSetpointTransformer in PID controllers\n      Configure LinearSetpointTransformer for PID controllers\n      gbmcweb: Allow overriding ServiceLabel and PLC in devpath_plugin\n      Clean up LinkType fallback comment in network_adapter_test.\n      Create orgqr for FRE\n\nFusion-Link: fusion2 N/A\nTested: N/A\nGoogle-Bug-Id: 496763407\nGoogle-Bug-Id: 521979698\nGoogle-Bug-Id: 531491777\nGoogle-Bug-Id: 559754428\nChange-Id: I87f48638685a03454efc62f52fa7272b86154e27\nSigned-off-by: Edward Lee \u003cedwarddl@google.com\u003e\n"
    },
    {
      "commit": "f04bcea3dc92c8d00adfff502ad373681725506d",
      "tree": "37909e945e17f1c405fe32b6f4b5b15bb3b1cd5e",
      "parents": [
        "02c7d3efb964a6c1ee7402a4ad1ba6cb8ad8d4b1"
      ],
      "author": {
        "name": "Munawar Hussain",
        "email": "munawarhussain@google.com",
        "time": "Thu Oct 29 05:47:55 2026 +0000"
      },
      "committer": {
        "name": "Munawar Hussain",
        "email": "munawarhussain@google.com",
        "time": "Fri Oct 30 10:02:21 2026 -0700"
      },
      "message": "libnvme: Backport MI CRC mismatch rc/errno fix\n\nWhen nvme_mi_submit() encounters an MCTP MIC/CRC mismatch,\nnvme_mi_verify_resp_mic() returns 1 with errno\u003d0, causing callers such\nas NVMeMi::adminXfer() (which only treats rc \u003c 0 as a transport error)\nto mistake corrupted responses for NVMe-MI status 0x1.\n\nBackport upstream libnvme commit 2f722b1b0f3d (\"mi: set correct rc and\nerrno when crc mismatch\") so CRC mismatches set errno\u003dEBADMSG and\nreturn rc\u003d-1.\n\nFusion-Link:\nplatform5: https://fusion2.corp.google.com/f1463efb-b8fb-3b0d-a46c-433d96955a74\nplatform15: https://fusion2.corp.google.com/6f70eb9e-6c9d-37d0-a9c5-99a2ebde9fc8\nplatform18: https://fusion2.corp.google.com/781def8d-149a-3126-8e20-e34e28a0d6b8\nplatform17: https://fusion2.corp.google.com/2b3d54e5-865f-3a6e-9f71-5aa47bb567cf\nplatform11+ext: https://fusion2.corp.google.com/a200ef10-a310-312a-a5ad-8d69295895ed\nplatform11: https://fusion2.corp.google.com/f856f37b-7da4-34f2-acbf-69dd3941333f\nplatform25: https://fusion2.corp.google.com/4bc2c6f5-aefa-397e-a882-dc179a123987\nplatform26: https://fusion2.corp.google.com/b486d9c0-ddb9-3ea9-8e43-44ebdd9e7fdc\n\nTested: Built libnvme for raya.\nGoogle-Bug-Id: 573831528\nChange-Id: I55e37afb4bee99ab2d202c4ab81981ae45246b8f\nSigned-off-by: Munawar Hussain \u003cmunawarhussain@google.com\u003e\n"
    },
    {
      "commit": "02c7d3efb964a6c1ee7402a4ad1ba6cb8ad8d4b1",
      "tree": "1f234885fa7d305bc41d057dc404db93a8649782",
      "parents": [
        "a050e97eb9a130db29549ec50e59e5a8524a9fce"
      ],
      "author": {
        "name": "Sandesh",
        "email": "sandeshharidas@google.com",
        "time": "Thu Oct 29 17:15:08 2026 +0000"
      },
      "committer": {
        "name": "Sandesh",
        "email": "sandeshharidas@google.com",
        "time": "Thu Oct 29 17:15:08 2026 +0000"
      },
      "message": "gbmcweb master: SRCREV bump b48fc0f439..ce08346e42\n\ngBMC Team (15):\n      gbmcweb: Fix multi-host ComputerSystem Status and Asset overwrite\n      Create geqp for FRE, and make `BlasNormCalculation` a template\n      Add LinearSetpointTransformer for thermal controller outputs\n      Configure the RWFS credential paths for ClearAllCredentials().\n      Add Store::ClearAllCredentials(). Exposes CredentialManager::ClearAllCredentials() through the Store.\n      gbmcweb: Return 204 for EthernetInterface PATCH\n      Harden expression parser and config expression substitution.\n      Harden Redfish routing, response error handling, and $expand JSON safety.\n      Capture owned std::string in GpioCollector::SetActiveForDuration callback.\n      Fixes two bugs in `buildNetworkPortMetricDBus` when handling `xyz.openbmc_project.Metrics.PortECC`:\n      Add WipeBmcCredentials parameter to Google.Wipe. Adds an optional WipeBmcCredentials boolean to the Google.Wipe action on the BMC Manager. When true, tlBMC calls Store::ClearAllCredentials() to remove all owner credentials (trust bundle, owner verification, OS verification, server and syslog certificates, and BMC and serial console SSH CA keys) without wiping entire RWFS. The parameter is listed in WipeActionInfo and accepted only on platforms that enable the trust bundle install module. On other platforms, a true value returns 400. A non-boolean value returns 400 before any wipe runs. If the parameter is omitted, the response is unchanged for existing callers.\n      CX9 ports publish link, error, and packet counters on `xyz.openbmc_project.Metrics.IBPort`, but `buildNetworkPortMetricDBus` only unpacked `BitErrorRate` and `LinkDownCount`, leaving `PortMetrics` missing telemetry required by the consumers.\n      Populates FECHistogramBins in PortMetrics TransceiverTelemetry from D-Bus.\n      Populate PortInfo and MAC address properties on D-Bus network ports.\n      Populates Asset, Location, and PCIe Controllers on D-Bus NetworkAdapters.\n\nFusion-Link: fusion2 N/A\nTested: N/A\nGoogle-Bug-Id: 496763407\nGoogle-Bug-Id: 496763407,574295165\nGoogle-Bug-Id: 531491777\nGoogle-Bug-Id: 543610220\nGoogle-Bug-Id: 545764272\nGoogle-Bug-Id: 559754428\nGoogle-Bug-Id: 573237408\nGoogle-Bug-Id: 578970832\nChange-Id: I8d3c28043d553c8836f4bb9a2c20bb7c823c6e52\nSigned-off-by: Sandesh \u003csandeshharidas@google.com\u003e\n"
    },
    {
      "commit": "a050e97eb9a130db29549ec50e59e5a8524a9fce",
      "tree": "77e364b539e18848cd50fb41c3f0c0686ad6328c",
      "parents": [
        "46c1ce548862ec40ba96ecaf577f4ed9c85f0d6e"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Wed Oct 28 22:06:49 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Thu Oct 29 00:06:12 2026 -0700"
      },
      "message": "linux-gbmc: npcm7xx-lpc-bpc: Don\u0027t lose host reset events in poll()\n\nThe IRQ handler sets the host reset flag while poll() tests and clears\nit with plain accesses. A reset flagged between the test and the clear\nis lost, and the unannotated accesses are a data race.\n\nTest and clear the flag atomically with xchg(). It is widened from bool\nas not every architecture supports xchg() on a single byte.\n\nTested: Compile tested at this point in the chain (COMPILE_TEST, W\u003d1).\nThe complete chain was built for NPCM7xx (6.14) and NPCM8xx (5.15)\nkernels.\nhttps://sashiko.corp.goog/#/patchset/7dbafbdf8d3b660d0967effe7d37d2ceac7c9e32.1793225364.git.wak%40google.com\n\nGoogle-Bug-Id: 362348309\nChange-Id: If0b474003b6b5b47aebf37286c7e56380cc51958\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "46c1ce548862ec40ba96ecaf577f4ed9c85f0d6e",
      "tree": "ad9ccca4d93fdb7d2030a7b6c9ec818a76a86553",
      "parents": [
        "eb7f84752288c862fb8c6c683365e79ceeec7ec4"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Wed Oct 28 22:06:48 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Thu Oct 29 00:06:12 2026 -0700"
      },
      "message": "linux-gbmc: npcm7xx-lpc-bpc: Grow the FIFO to 1024 records\n\nA full FIFO now drops the newest data, so it only holds the 256 oldest\ndouble-word records while a reader is stalled. Readers commonly rate\nlimit how many POST codes they process per second, and a host can emit\ncodes faster than that during boot.\n\nGrow the FIFO from 1KiB to 4KiB per channel.\n\nTested: Compile tested at this point in the chain (COMPILE_TEST, W\u003d1).\nThe complete chain was built for NPCM7xx (6.14) and NPCM8xx (5.15)\nkernels.\n\nGoogle-Bug-Id: 362348309\nChange-Id: Ie48f7a4d447fcfd6f7014fe65f61484be0a87e45\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "eb7f84752288c862fb8c6c683365e79ceeec7ec4",
      "tree": "13f263bebebe950b3f631eb3e05da91273af417d",
      "parents": [
        "a8af73decc55e2662925975ba8dd0e953460c550"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Wed Oct 28 22:06:48 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Thu Oct 29 00:06:12 2026 -0700"
      },
      "message": "linux-gbmc: npcm7xx-lpc-bpc: Defer committing short writes to a timer\n\nIn double-word capture mode the hardware marks the first byte of a host\nwrite but not the last, so a byte or word write is only distinguishable\nfrom a double word whose remaining bytes are still in flight by waiting.\nThe IRQ handler instead treats a partial record as a short write as soon\nas the hardware FIFO looks empty. An LPC double word arrives as four\nseparate byte cycles, so when the handler drains the first byte before\nthe rest have arrived, the write is split into a zero extended record\nand dropped continuation bytes.\n\nThe gap between the byte cycles of one write isn\u0027t bounded by LPC. On\nan NPCM750 the bytes of split writes were up to 9.6us apart during host\nboot and up to 19.4us apart while the IRQ handler was busy with a\nburst, both including interrupt latency.\n\nLeave a partial record pending and arm an hrtimer that commits it as a\nshort write dw_timeout_us (default 100us) after the last byte. The next\nbyte of the same write cancels the timer, and the first byte of the next\nwrite commits the pending record immediately, so only genuinely short\nwrites wait. Nothing busy-waits in the IRQ handler.\n\nA host reset discards a pending partial record. Handle it before\ndraining the FIFO, so it can\u0027t also discard the start of a write the\nhost made after the reset.\n\nThe IRQ handler and the timer callback are the two producers, and they\nhand ownership of the partial record and the producer side of the kfifo\nto each other without a lock: the IRQ handler cancels the timer on\nentry, which waits for a running callback to finish even on another\nCPU, and arms it as its last access. This relies on the timer\nexpiring in hard interrupt context, where the IRQ handler can\u0027t\ninterrupt the callback on the same CPU. The wait queue lock can\u0027t be\ntaken there on PREEMPT_RT, so the callback leaves waking readers to an\nirq_work. Interrupts that are only for the other users of the shared\nline return early without touching the timer.\n\nTested: Compile tested at this point in the chain (COMPILE_TEST, W\u003d1).\nThe complete chain was built for NPCM7xx (6.14) and NPCM8xx (5.15)\nkernels.\n\nGoogle-Bug-Id: 362348309\nChange-Id: I68ca3bc41f782c7bd4b8e406a9f11748322ec0f5\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "a8af73decc55e2662925975ba8dd0e953460c550",
      "tree": "bf4f639b35e8adb12293a6a2732fcd34f2539409",
      "parents": [
        "cb95e1cad031a5c8210148bdb842dff4bfb5917c"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Wed Oct 28 22:06:48 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Thu Oct 29 00:06:12 2026 -0700"
      },
      "message": "linux-gbmc: npcm7xx-lpc-bpc: Only hand out whole records\n\nIn double-word capture mode read() returns as soon as 3 bytes are queued\nand copies whatever is available, so a 4 byte record whose last byte\nhasn\u0027t been queued yet is split across two reads. Readers treating each\nread as a POST code see a torn code and every following code misaligned.\n\npoll() also disagrees with read(): it reports EPOLLIN for a single\nqueued byte while a non-blocking read() returns -EAGAIN below 3 bytes.\nOnce 1 or 2 bytes are left over, a level-triggered poller like\nphosphor-host-postd wakes up forever, pinning a CPU until the next host\nwrite. A single word write from the host is enough to trigger this.\n\nRead in units of whole records (4 bytes in double-word capture mode, 1\nbyte otherwise): reject buffers smaller than a record, only return a\nmultiple of the record size and only report EPOLLIN once a whole record\nis queued. Copy out through a local buffer so a faulting user buffer\ncan\u0027t consume part of a record.\n\nTested: Compile tested at this point in the chain (COMPILE_TEST, W\u003d1).\nThe complete chain was built for NPCM7xx (6.14) and NPCM8xx (5.15)\nkernels.\n\nGoogle-Bug-Id: 362348309\nChange-Id: I23f9561b4c447060783a607a1598caa0ea568eaf\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "cb95e1cad031a5c8210148bdb842dff4bfb5917c",
      "tree": "3b5611811385f5dfea72d379c925699304281595",
      "parents": [
        "b1113c833fc07ef110fbd0f6e1534a07da3e8469"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Wed Oct 28 22:06:48 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Thu Oct 29 00:06:12 2026 -0700"
      },
      "message": "linux-gbmc: npcm7xx-lpc-bpc: Serialize readers\n\nThe channel kfifo is accessed without locking, which is only safe with a\nsingle consumer. Concurrent read() calls on the same channel, through\nthe same or different file descriptors, race on the consumer index and\ncan hand the same data to both readers or corrupt the stream.\n\nSerialize readers with a per-channel mutex. The IRQ handler, the only\nproducer, doesn\u0027t take it. The mutex isn\u0027t held while waiting for data,\nso a non-blocking reader never waits for a blocked one.\n\nTested: Compile tested at this point in the chain (COMPILE_TEST, W\u003d1).\nThe complete chain was built for NPCM7xx (6.14) and NPCM8xx (5.15)\nkernels.\n\nGoogle-Bug-Id: 362348309\nChange-Id: I7628fbea53fed39a1d2820977457c69a43b56d49\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "b1113c833fc07ef110fbd0f6e1534a07da3e8469",
      "tree": "4d2b57e0c7f5bf39595fa6b77fb933dbc834c435",
      "parents": [
        "96531a10ef5f1c56360b0491f24b4c01ceab378a"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Wed Oct 28 22:06:47 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Thu Oct 29 00:06:12 2026 -0700"
      },
      "message": "linux-gbmc: npcm7xx-lpc-bpc: Queue whole double-word records\n\nIn double-word capture mode userspace reads the FIFO as a stream of 4\nbyte records, one per host write, but the driver doesn\u0027t keep it that\nway:\n\n- A host write of a word (outw) is queued as 2 bytes. Only byte writes\n  are padded, so every record after a word write is misaligned.\n- A full FIFO drops individual bytes, which also misaligns every\n  following record.\n- A continuation byte whose first byte was lost (e.g. to a hardware FIFO\n  overflow) is queued as if it started a record.\n\nAssemble each host write in the IRQ handler and only queue (or drop)\ncomplete records. FIFO_ADDR_DECODE marks the first byte of a host write,\nso a new first byte completes any pending short write, short writes are\nzero extended to a full record and continuation bytes without a start\nare dropped.\n\nA partial record left when the hardware FIFO runs dry is still treated\nas a short write, as before.\n\nTested: Compile tested at this point in the chain (COMPILE_TEST, W\u003d1).\nThe complete chain was built for NPCM7xx (6.14) and NPCM8xx (5.15)\nkernels.\n\nGoogle-Bug-Id: 362348309\nChange-Id: Ifb90bc01747268d50d4a5f840fccd39424d290b5\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "96531a10ef5f1c56360b0491f24b4c01ceab378a",
      "tree": "d78f216149ce92721338ef93227590e936772b8d",
      "parents": [
        "a166b120e018042439db8cae177446e4ea65de75"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Wed Oct 28 22:06:47 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Thu Oct 29 00:06:12 2026 -0700"
      },
      "message": "linux-gbmc: npcm7xx-lpc-bpc: Don\u0027t consume from the FIFO in the IRQ handler\n\nThe channel kfifo is used without locking, which is only safe with a\nsingle producer and a single consumer. When it is full, the IRQ handler\nmakes room with kfifo_skip(), advancing the consumer index while read()\nmay be copying out of the same kfifo, corrupting the data handed to\nuserspace.\n\nDrop the new data instead so the IRQ handler only ever acts as the\nproducer.\n\nTested: Compile tested at this point in the chain (COMPILE_TEST, W\u003d1).\nThe complete chain was built for NPCM7xx (6.14) and NPCM8xx (5.15)\nkernels.\n\nGoogle-Bug-Id: 362348309\nChange-Id: Id8b99c9177a70d4af79714e31f0090264ef3e053\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "a166b120e018042439db8cae177446e4ea65de75",
      "tree": "85d53246915e1a2b29911c2cf40a3b7627b451cc",
      "parents": [
        "4b97f04a1a306ced2d711f07250edd98ca6cc382"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Wed Oct 28 22:06:47 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Thu Oct 29 00:06:12 2026 -0700"
      },
      "message": "linux-gbmc: npcm7xx-lpc-bpc: Rate limit the FIFO overflow message\n\nThe IRQ handler logs a message for every byte drained while the hardware\nFIFO overflow flag is set. Under a burst of host writes this floods the\nconsole from hard IRQ context, making the overflow worse.\n\nUse dev_warn_ratelimited() so the message also identifies the device.\n\nTested: Compile tested at this point in the chain (COMPILE_TEST, W\u003d1).\nThe complete chain was built for NPCM7xx (6.14) and NPCM8xx (5.15)\nkernels.\n\nGoogle-Bug-Id: 362348309\nChange-Id: I96e65d5a01469b9a91e78220b3933602c03471ba\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "4b97f04a1a306ced2d711f07250edd98ca6cc382",
      "tree": "b7c9aa6a47eb404b5536a6b1e0e8e042b9b63117",
      "parents": [
        "047b17bfb7b1324e99c1a26b12fa6e836aaa97e4"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Wed Oct 28 22:06:47 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Thu Oct 29 00:06:12 2026 -0700"
      },
      "message": "linux-gbmc: npcm7xx-lpc-bpc: Wake every channel that received data\n\nWhen both monitored addresses are in use, a single interrupt can drain\nbytes for both channels, but only the channel of the last byte read is\nwoken, leaving readers of the other channel asleep with data queued.\nHost reset is likewise only flagged on that one channel, so pollers of\nthe other channel never see EPOLLHUP.\n\nTrack which channels received data and wake each of them, and flag\nhost reset on every registered channel.\n\nTested: Compile tested at this point in the chain (COMPILE_TEST, W\u003d1).\nThe complete chain was built for NPCM7xx (6.14) and NPCM8xx (5.15)\nkernels.\n\nGoogle-Bug-Id: 362348309\nChange-Id: I4b570e758714bef2bed371f9484e0bd571e0cd5b\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "047b17bfb7b1324e99c1a26b12fa6e836aaa97e4",
      "tree": "49f13830cda7492a394b341f9ef01ace64c7cc63",
      "parents": [
        "347e2280d32a5d167f36ec8ff92c5bd405d3f8e4"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Wed Oct 28 22:06:46 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Thu Oct 29 00:06:12 2026 -0700"
      },
      "message": "linux-gbmc: npcm7xx-lpc-bpc: Disallow unbinding through sysfs\n\nmisc_deregister() doesn\u0027t wait for open files to be closed, yet remove\nfrees the channel FIFOs and devres then frees the driver data. Unbinding\nthe driver while a reader has the device open, or is sleeping in read(),\nleads to a use after free.\n\nThe module can\u0027t be unloaded while a file is open as the file holds a\nreference to it, which leaves unbinding through sysfs as the only way\nto get there. The block is part of the SoC and there is no reason to\nunbind it, so suppress the bind attributes.\n\nTested: Compile tested at this point in the chain (COMPILE_TEST, W\u003d1).\nThe complete chain was built for NPCM7xx (6.14) and NPCM8xx (5.15)\nkernels.\n\nGoogle-Bug-Id: 362348309\nChange-Id: I233aee09c2547c14f95678bcba1f23c37d60a654\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "347e2280d32a5d167f36ec8ff92c5bd405d3f8e4",
      "tree": "11980accac37c358d3fe4865c73a87b3a91dbe05",
      "parents": [
        "2b007df7b00e9d91369c7782d191cc0b0c799db4"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Wed Oct 28 22:06:46 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Thu Oct 29 00:06:12 2026 -0700"
      },
      "message": "linux-gbmc: npcm7xx-lpc-bpc: Set up the channel before registering it\n\nmisc_register() makes the device node available to userspace, but the\nchannel\u0027s back pointer to the driver data is only set afterwards. A\nread() or poll() racing with probe dereferences the NULL pointer to look\nup the capture mode.\n\nInitialize the channel before registering the device.\n\nTested: Compile tested at this point in the chain (COMPILE_TEST, W\u003d1).\nThe complete chain was built for NPCM7xx (6.14) and NPCM8xx (5.15)\nkernels.\n\nGoogle-Bug-Id: 362348309\nChange-Id: I3e5343135596551da1e592f6882b56c78dedc9b1\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "2b007df7b00e9d91369c7782d191cc0b0c799db4",
      "tree": "9f288d0144bb585b244b1ac04130b3e8a69128d0",
      "parents": [
        "b2dfc20ca330f0cadc154cf75c184cf896f65b44"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Wed Oct 28 22:06:46 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Thu Oct 29 00:06:12 2026 -0700"
      },
      "message": "linux-gbmc: npcm7xx-lpc-bpc: Fix teardown\n\nremove() decides which channels to tear down from the address enable\nbits in BPCFEN, but double-word capture mode enables FIFO_DWCAPTURE\ninstead of FIFO_IOADDR1_ENABLE. In that mode nothing is torn down: the\nmisc device stays registered and the IRQ stays live, both referencing\ndevm allocated memory that is freed once remove() returns.\n\nnpcm7xx_disable_bpc() also frees the kfifo while the IRQ may still be\nwriting to it and before the misc device is deregistered, and decides\nwhether to disable the interrupts from the register value read before\nclearing the channel\u0027s enable bit, so they are never disabled.\n\nTrack which channels were registered and tear down in order: quiesce\nthe block, free the IRQ, then deregister the misc devices and free the\nFIFOs.\n\nTested: Compile tested at this point in the chain (COMPILE_TEST, W\u003d1).\nThe complete chain was built for NPCM7xx (6.14) and NPCM8xx (5.15)\nkernels.\n\nGoogle-Bug-Id: 362348309\nChange-Id: Ie59623485e2702e38b7f663632eea739c57608f0\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "b2dfc20ca330f0cadc154cf75c184cf896f65b44",
      "tree": "5a8d68c1ff2032669312109d5ce80391a69a94b4",
      "parents": [
        "65d79b74817a02697e736b346833b18222f39dbd"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Wed Oct 28 22:06:46 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Thu Oct 29 00:06:12 2026 -0700"
      },
      "message": "linux-gbmc: npcm7xx-lpc-bpc: Free the FIFO when channel setup fails\n\nIf misc_register() fails the channel\u0027s kfifo is leaked, and a failed\ndevm_kasprintf() for the device name is not detected at all. Check the\nname allocation and free the kfifo on both error paths.\n\nTested: Compile tested at this point in the chain (COMPILE_TEST, W\u003d1).\nThe complete chain was built for NPCM7xx (6.14) and NPCM8xx (5.15)\nkernels.\n\nGoogle-Bug-Id: 362348309\nChange-Id: I7fe212cbf013f3d066bc6811c5ce2906f3bfff21\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "65d79b74817a02697e736b346833b18222f39dbd",
      "tree": "f8f1c48a68594659a45cea82e4d8341cd2c595d0",
      "parents": [
        "575cf285e24e414f56a0bca1d12d9e05cb20e860"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Wed Oct 28 22:06:45 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Thu Oct 29 00:06:12 2026 -0700"
      },
      "message": "linux-gbmc: npcm7xx-lpc-bpc: Request the IRQ after the FIFOs exist\n\nThe shared IRQ is requested before the channel FIFOs are allocated, and\nthe block may still have its interrupts enabled from the bootloader or a\nprevious kernel. An interrupt arriving in that window runs the handler\nagainst a zeroed kfifo and dereferences its NULL buffer.\n\nSplit channel setup from the hardware programming and probe in order:\nquiesce the block, allocate and register the channels, request the IRQ,\nthen enable capture and interrupts. Unwind the channels if requesting\nthe IRQ fails.\n\nDisabling capture doesn\u0027t empty the hardware FIFO. As the handler runs\nfor every interrupt on the shared line, bytes left over from the old\nconfiguration, e.g. ones captured for a second address that is no\nlonger monitored, would still be routed to a channel that doesn\u0027t\nexist. Drain the hardware FIFO and acknowledge a stale host reset\nevent when quiescing.\n\nTested: Compile tested at this point in the chain (COMPILE_TEST, W\u003d1).\nThe complete chain was built for NPCM7xx (6.14) and NPCM8xx (5.15)\nkernels.\n\nGoogle-Bug-Id: 362348309\nChange-Id: Ie95e44f16f12979ee2558c899f725270c230f210\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "575cf285e24e414f56a0bca1d12d9e05cb20e860",
      "tree": "4828d3d7ec9e2e73470d955b33ccf982921fec93",
      "parents": [
        "d4fb32776574d1f67dd2c90974bb87d4d2a61646"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Thu Oct 29 05:33:09 2026 +0000"
      },
      "committer": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Thu Oct 29 05:33:09 2026 +0000"
      },
      "message": "systemd: Re-enable zstd compression on gBMC\n\nCommit da38ead66 removed zstd from systemd PACKAGECONFIG on gBMC.\nBecause meta-phosphor only enables zstd among systemd\u0027s compression\nbackends, removing it builds systemd with HAVE_COMPRESSION\u003d0\n(DEFAULT_COMPRESSION\u003dCOMPRESSION_NONE), which:\n- Leaves journal payloads (\u003e\u003d 512B) and persistent eMMC journals\n  uncompressed (and prevents reading prior zstd-compressed journals).\n- Writes uncompressed external coredumps that can exceed tmpfs/eMMC\n  MaxUse limits.\n- Breaks \u0027coredumpctl dump \u003cpid\u003e\u0027 on external coredumps with EOPNOTSUPP\n  (\u0027Cannot decompress file. Compiled without compression support.\u0027),\n  breaking off-machine coredump collection.\n\nRe-enable zstd in systemd PACKAGECONFIG on gBMC.\n\nFusion-Link: fusion2 n/a\nTested: Verified systemd PACKAGECONFIG includes zstd.\nGoogle-Bug-Id: 558894649\nChange-Id: Id965155ec4a37644c8ffeaad92c91b6c21164993\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "d4fb32776574d1f67dd2c90974bb87d4d2a61646",
      "tree": "8a7518db37f399949c075ca7497a959f3bef7f82",
      "parents": [
        "333e325133b203d43c5e9f16bf76bc0ccd178f5e"
      ],
      "author": {
        "name": "Edward Lee",
        "email": "edwarddl@google.com",
        "time": "Wed Oct 28 00:33:53 2026 +0000"
      },
      "committer": {
        "name": "Edward Lee",
        "email": "edwarddl@google.com",
        "time": "Tue Oct 27 17:41:36 2026 -0700"
      },
      "message": "gbmcweb-google-plugins master: SRCREV bump a6eb9d824b..14c9d7a16f\n\nEdward Lee (2):\n      [REDACTED]: link cards to SB_Cable_98/99\n      [REDACTED]: report drive devpath under PhysicalLocation\n\nPlatforms-Affected: All-TTF\nFusion-Link: fusion2 N/A\nTested: N/A\nGoogle-Bug-Id: 560168889\nGoogle-Bug-Id: 578544163\nChange-Id: Ia06855a575814a14779063924b6113fdd7ad87e0\nSigned-off-by: Edward Lee \u003cedwarddl@google.com\u003e\n"
    },
    {
      "commit": "333e325133b203d43c5e9f16bf76bc0ccd178f5e",
      "tree": "1a18d0ada4f2b9dc00a6edbf7c2e4dfd8d73c821",
      "parents": [
        "b9cbbf9d83f1954b214390010867640ab37b955d"
      ],
      "author": {
        "name": "Jane Kramer",
        "email": "jkrm@google.com",
        "time": "Mon Oct 26 23:49:54 2026 +0000"
      },
      "committer": {
        "name": "Edward Lee",
        "email": "edwarddl@google.com",
        "time": "Tue Oct 27 17:26:36 2026 -0700"
      },
      "message": "gbmcweb master: SRCREV bump cb8b2de80f..b48fc0f439\n\ngBMC Team (9):\n      Add support for multi-pin GPIO rules for FRU presence detection.\n      Create gemv for FRE\n      Create gerc for FRE\n      Migrate BMC interface role enum references from BMC_IR_DIRECT_ETHERNET and BMC_IR_SIDEBAND_NCSI to BMC_IR_DIRECT and BMC_IR_SIDEBAND.\n      Internal change\n      Let ADRC disturbance estimates unwind *iff* given equidirectionally saturated output\n      tlbmc: Remove stale /var/google/install/progress_status.json before update\n      Add CredentialManager::ClearAllCredentials(). Removes all credentials managed by CredentialManager from disk and memory, with the trust bundle last. Returns an error naming every item that was not removed; in that case the trust bundle is kept so the clear can be retried.\n      tlbmc: Populate RawFruTable even when FRU area checksum fails\n\nFusion-Link: fusion2 N/A\nTested: N/A\nGoogle-Bug-Id: 435693350\nGoogle-Bug-Id: 464081382,562802792\nGoogle-Bug-Id: 492137013,570840373\nGoogle-Bug-Id: 519274131\nGoogle-Bug-Id: 545329676\nGoogle-Bug-Id: 545764272\nGoogle-Bug-Id: 559754428\nGoogle-Bug-Id: 576452956\nChange-Id: I1f4c602cf1a2363527fe08d416a5c92c931cc9b3\nSigned-off-by: Jane Kramer \u003cjkrm@google.com\u003e\n"
    },
    {
      "commit": "b9cbbf9d83f1954b214390010867640ab37b955d",
      "tree": "173f49cd929c83923321e154ac321590f80fe3f8",
      "parents": [
        "dcda3885d52ae17d28dfd7b3d7a3e3977d38adc3"
      ],
      "author": {
        "name": "Edward Lee",
        "email": "edwarddl@google.com",
        "time": "Tue Oct 27 20:39:34 2026 +0000"
      },
      "committer": {
        "name": "Edward Lee",
        "email": "edwarddl@google.com",
        "time": "Tue Oct 27 20:39:35 2026 +0000"
      },
      "message": "gbmcweb-google-plugins master: SRCREV bump a72cb983dc..a6eb9d824b\n\nEdward Lee (1):\n      iliad: report DRAM vendor and DIMM type errors\n\nFusion-Link: fusion2 N/A\nTested: N/A\nGoogle-Bug-Id: 578156064\nChange-Id: I1d8b3270a44b5831065bc8fd56b789408059ed32\nSigned-off-by: Edward Lee \u003cedwarddl@google.com\u003e\n"
    },
    {
      "commit": "dcda3885d52ae17d28dfd7b3d7a3e3977d38adc3",
      "tree": "fc56d66e90d3668d5b21f58896c8736bebbd9a0f",
      "parents": [
        "d9d4ed5b2cfb7b36f927715930cfd4204d2f7223"
      ],
      "author": {
        "name": "Siddharth Golechha",
        "email": "siddharthpg@google.com",
        "time": "Mon Oct 26 23:48:13 2026 +0000"
      },
      "committer": {
        "name": "Siddharth Golechha",
        "email": "siddharthpg@google.com",
        "time": "Tue Oct 27 11:58:24 2026 -0700"
      },
      "message": "gbmcweb-google-plugins master: SRCREV bump 38453ca582..a72cb983dc\n\nSiddharth Golechha (1):\n      iliad: support MemTestBlob PPR on Tubul Chassis\n\nFusion-Link: fusion2 N/A\nTested: N/A\nGoogle-Bug-Id: 577189056\nChange-Id: If9831ad4d8c1dadf1436355a9949c98a692f6d12\nSigned-off-by: Siddharth Golechha \u003csiddharthpg@google.com\u003e\n"
    },
    {
      "commit": "d9d4ed5b2cfb7b36f927715930cfd4204d2f7223",
      "tree": "d101b6dbff0e15ba825f7a6653c099c58ca3cbf5",
      "parents": [
        "18c0df03f07801509d01b1b19a31ae228bd2c5be"
      ],
      "author": {
        "name": "Joseph CT Chan",
        "email": "josephctchan@google.com",
        "time": "Fri Oct 23 23:02:51 2026 +0000"
      },
      "committer": {
        "name": "Joseph Chan",
        "email": "josephctchan@google.com",
        "time": "Tue Oct 27 10:50:06 2026 -0700"
      },
      "message": "linux-gbmc: npcm7xx-lpc-bpc: fix wait condition\n\nIn npcm7xx_bpc_read, the wait condition used strict inequality\nkfifo_len(\u0026chan-\u003efifo) \u003e cond_size instead of \u003e\u003d cond_size.\nWhen a reader blocks on an empty FIFO, arrival of cond_size bytes\nfailed to satisfy the condition, causing the process to sleep\nindefinitely until additional bytes arrived.\n\nTested: Built obmc-phosphor-image with linux-gbmc 5.15, signed and\nflashed onto hardware testbed (lcatlc-ep1), and confirmed normal boot\nand character device node operation.\nFusion-Link: http://fusion2/ci/kokoro/prod%3Agbmc%2Ffirmware-build%2Ffirmware%2Fredacted_gbmc_platforms/activity/0806e662-d9ba-4d64-a4f0-78a7d4cd2584\nGoogle-Bug-Id: 552611529\nChange-Id: I4b41915f63f64c7b00edf9f8b6357e11457a08aa\nSigned-off-by: Joseph Chan \u003cjosephctchan@google.com\u003e\n"
    },
    {
      "commit": "18c0df03f07801509d01b1b19a31ae228bd2c5be",
      "tree": "e5e0b17c375134191a008dcd88d6a99811d49c71",
      "parents": [
        "0ef7cb12ec98a576f6b974cbaf844114dbba19bb"
      ],
      "author": {
        "name": "Vivekanand Veeracholan",
        "email": "vveerach@google.com",
        "time": "Fri Oct 23 14:12:37 2026 -0700"
      },
      "committer": {
        "name": "Vivekanand Veeracholan",
        "email": "vveerach@google.com",
        "time": "Tue Oct 27 10:37:43 2026 -0700"
      },
      "message": "google/libhoth main: SRCREV bump bf14409af2..9d5f1a561d\n\nMichael Field (4):\n      Move the CLI framework from examples/ to tools/common/\n      Move htool from examples/ to tools/htool/\n      Rename header guards to match the new paths\n      Add a //examples:htool compat target\n\nRoyce (1):\n      htool: parse `disarm_after_seconds` from the 16-byte TWCH blob (#311)\n\nesnguyen (7):\n      [examples] Convert htool_security_v2 to use the new compact error format (#312)\n      [htool] Add htool interface for payload effective + bug fix (#310)\n      [examples] Convert htool_target_control to use the new compact error format (#316)\n      [examples] Convert htool_security_version to use the new compact error format (#314)\n      [examples] Convert htool_raw_host_command to use the new compact error format (#317)\n      [examples] Convert htool_tpm to use the new compact error format (#315)\n      [protocol] Convert provisioning log commands to use the new compact error format (#319)\n\nFusion-Link: fusion2 N/A\nGoogle-Bug-Id: 573734175\nTested: N/A\nChange-Id: Ib780feacb3f6d8fd1fe66a9520e70d8e4bb1bf2e\nSigned-off-by: Vivekanand Veeracholan \u003cvveerach@google.com\u003e\n"
    },
    {
      "commit": "0ef7cb12ec98a576f6b974cbaf844114dbba19bb",
      "tree": "da8dd6302ef4987cf2e119b9b919a40080d338a1",
      "parents": [
        "bc86e9f0837b0e39741d2c4a2a41c9eb66774b3d"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Sun Oct 25 19:48:58 2026 +0000"
      },
      "committer": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Tue Oct 27 05:45:51 2026 +0000"
      },
      "message": "gbmc-staging: Avoid BCJ filter on true-hitless-image\n\nCommit 32a0fa1fb enabled -b 256k -Xbcj arm / -Xbcj arm64 globally for\nsquashfs-xz images, which requires CONFIG_XZ_DEC_ARM /\nCONFIG_XZ_DEC_ARM64 in the kernel (added in 134e9ab18).\n\nUnlike obmc-phosphor-image, true-hitless-image is mounted at runtime on\nolder base firmware kernels during cross-version upgrades and testing.\nOlder kernels without CONFIG_XZ_DEC_ARM fail to decompress BCJ-filtered\nSquashFS blocks with EIO.\n\nClear EXTRA_IMAGECMD:squashfs-xz:pn-true-hitless-image:* so\ntrue-hitless-image remains mountable on older base firmware kernels.\n\nTested:\n- Mounted True Hitless 26.40.901.8 SquashFS image on hardware running\n  older base firmware 26.31.19.0 without SquashFS decompression errors.\nFusion-Link: fusion2 n/a\nGoogle-Bug-Id: 575520741\nGoogle-Bug-Id: 577493452\nChange-Id: I4ceb6512e3a6e683596aa138f6cbdc73e1b74c4b\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "bc86e9f0837b0e39741d2c4a2a41c9eb66774b3d",
      "tree": "2d35fd5e3615cb92eb456d98c7ea321ed2707ffe",
      "parents": [
        "c05fe689590225171dfaedf716234d26143166a3"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Mon Oct 26 05:19:04 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Mon Oct 26 22:03:08 2026 -0700"
      },
      "message": "recipes: Prune unused target perl, python3, cracklib, and shared-mime-info\n\nNeither target perl nor target python3 is installed in any gBMC image,\nyet uninstalled auxiliary subpackages (i2c-tools-misc, makedumpfile-tools,\nndisc6-misc, openssl-misc, safec-check, glib-2.0-codegen, iproute2-routel,\nnftables-python, nghttp2-proxy) and default PACKAGECONFIGs (libxml2,\nopkg-utils) triggered full cross-compilation of target perl and target\npython3 along with 20+ supporting target libraries (sqlite3, gdbm, libffi,\nbzip2, etc.).\n\nStrip these unused runtime/build dependencies in their respective\nbbappends, along with cracklib from libpam and shared-mime-info from\nsystemd and glib-2.0, eliminating 38+ target and native recipes from the\nbuild graph.\n\nFusion-Link: fusion2 n/a\nTested: Built obmc-phosphor-image across gBMC platforms.\nGoogle-Bug-Id: 577771936\nChange-Id: I336953066766ebe02e109a222aa397d7b6e12cee\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "c05fe689590225171dfaedf716234d26143166a3",
      "tree": "7ad94266fe0cd53108c8f67c1a5b97cf2918e067",
      "parents": [
        "d442e145f8797c39d87e3f3ef23cc63cc02cd837"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Mon Oct 26 05:19:03 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Mon Oct 26 22:03:08 2026 -0700"
      },
      "message": "gbmc-staging: Disable unused locales, gconv, cpio.u-boot, and Boost libs\n\n- Disable glibc binary locale and gconv generation (gBMC uses the C\n  locale exclusively).\n- Remove cpio.${INITRAMFS_CTYPE}.u-boot from IMAGE_FSTYPES so\n  obmc-phosphor-image does not XZ-compress the entire rootfs into an\n  unused ~45 MB CPIO archive alongside squashfs-xz while keeping\n  cpio.xz for obmc-phosphor-initramfs.\n- Scope out unused compiled Boost libraries (iostreams, program_options,\n  regex, system) on gbmc targets in boost_%.bbappend.\n\nFusion-Link: fusion2 n/a\nTested: Built obmc-phosphor-image across gBMC platforms.\nGoogle-Bug-Id: 577771936\nChange-Id: I76d0cd764a258313636a4a649dbc49f9dcd87519\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "d442e145f8797c39d87e3f3ef23cc63cc02cd837",
      "tree": "0bfd2ff804c62693013fdf090b1f31972ec9e8f7",
      "parents": [
        "c2db6f46c006df1bb3d544beb33eef751da2ec5e"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Mon Oct 26 05:19:03 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Mon Oct 26 22:03:08 2026 -0700"
      },
      "message": "gbmc-staging: Remove QEMU from the build dependency graph\n\nOpt out of qemu-usermode and gobject-introspection-data, disable\nEXEWRAPPER_ENABLED, and remove qemu-native / qemuwrapper-cross /\ncross-localedef-native from build dependencies so that qemu-native and\nits host graphics/audio dependencies (glib-2.0-native, pixman-native,\nlibslirp-native, alsa-lib-native) are never built.\n\nFusion-Link: fusion2 n/a\nTested: Built obmc-phosphor-image across gBMC platforms.\nGoogle-Bug-Id: 577771936\nChange-Id: I169e78333c954700dc2975010884caa10decd0fe\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "c2db6f46c006df1bb3d544beb33eef751da2ec5e",
      "tree": "217d74d649ece4fecce6d38611c249c894492f94",
      "parents": [
        "ce24a73602b3dcc488b32db16275cbb4a2cba65b"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Mon Oct 26 05:19:02 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Mon Oct 26 22:03:08 2026 -0700"
      },
      "message": "gbmc-staging: Only run create-spdx on compliance builds\n\ncreate-spdx adds over 1,100 do_create_spdx and do_create_runtime_spdx\ntasks per build (~2,443s cumulative task time), which are only consumed\nwhen generating compliance distribution artifacts.\n\nGate create-spdx on GBMC_COMPLIANCE_DIST \u003d\u003d \u00271\u0027.\n\nFusion-Link: fusion2 n/a\nTested: Built obmc-phosphor-image across gBMC platforms.\nGoogle-Bug-Id: 577771936\nChange-Id: I0714c76e655629b617e8643e0f38cddb5c2ff0f3\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "ce24a73602b3dcc488b32db16275cbb4a2cba65b",
      "tree": "433b53ade22c96edeb0f25fa6737783cace7cbb1",
      "parents": [
        "cddf39df4d07e475ea2da6ab508a39e995dffa94"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Mon Oct 26 05:19:01 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Mon Oct 26 22:03:08 2026 -0700"
      },
      "message": "gbmc-staging: Switch package backend from RPM to zstd-compressed IPK\n\npackage_rpm defaults to zstd level 19 (w19T.zstdio) for intermediate\npackage archives that are immediately unpacked during do_rootfs and\nre-compressed into squashfs-xz / cpio.xz. Switching to package_ipk with\nzstd level 3 reduces do_package_write_* elapsed time by 84% (6.35x\nspeedup) and do_rootfs elapsed time by 70% (3.38x speedup), while also\navoiding building rpm-native, dnf-native, libdnf-native, librepo-native,\nand createrepo-c-native. Also pass -1 to pbzip2 for fast .bz2 image\nconversions.\n\nFusion-Link: fusion2 n/a\nTested: Built obmc-phosphor-image across gBMC platforms.\nGoogle-Bug-Id: 577771936\nChange-Id: I3a59c57c81a90fe0e96a91cb34e8494bc31c56e0\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "cddf39df4d07e475ea2da6ab508a39e995dffa94",
      "tree": "5a3a873faa136fc4f401223e05a3392fb1433703",
      "parents": [
        "1b96be6051766a119a5be7f1a963b550d56bb6a4"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Mon Oct 26 04:50:55 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Mon Oct 26 22:03:08 2026 -0700"
      },
      "message": "gbmc_extendable_image: Avoid building unused container tar.bz2\n\nAdding \u0027container\u0027 to IMAGE_FSTYPES inherits image-container.bbclass,\nwhich sets IMAGE_TYPEDEP:container \u003d \"tar.bz2\" and forces emmc-image to\nbuild and bzip2-compress an unused rootfs tarball alongside the\nsquashfs-xz.verity image.\n\nSet ROOTFS_BOOTSTRAP_INSTALL \u003d \"\" directly and drop \u0027container\u0027 from\nIMAGE_FSTYPES to skip kernel dependency injection without generating a\ntar.bz2 image.\n\nFusion-Link: fusion2 n/a\nTested: Built obmc-phosphor-image across gBMC platforms.\nGoogle-Bug-Id: 577771936\nChange-Id: If21d23cc935aa4db9f9488a32222a968c1440961\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "1b96be6051766a119a5be7f1a963b550d56bb6a4",
      "tree": "c18dec8257c71741fe013d0ea859f5c82f8656d0",
      "parents": [
        "f80e7274990766aa45b2b5ed0a24ae230372343d"
      ],
      "author": {
        "name": "Jane Kramer",
        "email": "jkrm@google.com",
        "time": "Thu Oct 22 22:49:43 2026 +0000"
      },
      "committer": {
        "name": "Jane Kramer",
        "email": "jkrm@google.com",
        "time": "Mon Oct 26 16:43:22 2026 -0700"
      },
      "message": "entity-manager: Publish FruDevice D-Bus object even on checksum failure\n\nContinue publishing the xyz.openbmc_project.FruDevice D-Bus object when\nvalidateFRUAreas() fails, so dependent daemons (such as entity-manager\nand mb-info) continue to work on machines with corrupted FRU checksums.\n\nThe I2CDevice D-Bus object (/xyz/openbmc_project/FruDevice/\u003cbus\u003e_\u003caddr\u003e)\nstill sets Corrupted \u003d true and retains raw EEPROM bytes in busMap, so\n/redfish/v1/Managers/bmc/Oem/Google/FruDevice/RawEeproms continues to\nreport corrupted FRUs.\n\nTested: https://paste.googleplex.com/6755399436243019\nFusion-Link: fusion2 N/A\nGoogle-Bug-Id: 492137013\nGoogle-Bug-Id: 570840373\nChange-Id: I9e9833261401bf4f4fc67c798a1ff62fe28c0e12\nSigned-off-by: Jane Kramer \u003cjkrm@google.com\u003e\n"
    },
    {
      "commit": "f80e7274990766aa45b2b5ed0a24ae230372343d",
      "tree": "19287b81bd78b89d3a8d000f98d5c1d403b91bec",
      "parents": [
        "fa3670bf378edc25c59701630fbaa54ef8143dc9"
      ],
      "author": {
        "name": "Rudransh Agrawal",
        "email": "rudyagrawal@google.com",
        "time": "Mon Oct 26 17:40:15 2026 +0000"
      },
      "committer": {
        "name": "Rudransh Agrawal",
        "email": "rudyagrawal@google.com",
        "time": "Mon Oct 26 13:30:18 2026 -0700"
      },
      "message": "pcie-bifurcation: SRCREV bump 5a1090e3da..4fb137de57\n\nMunawar Hussain (1):\n      pcie-bifurcation: Add unit test for new cables configs and format file\n\nRudransh Agrawal (1):\n      fru: Apply default_config after slot duplication\n\nTested: Built pcie-bifurcation for yusaku and verified on lcsqlp-d4\nFusion-Link: fusion2 N/A\nGoogle-Bug-Id: 559811656\nChange-Id: Id8caa9c72c4b4ecaa8b31fc5307b45289f903b43\nSigned-off-by: Rudransh Agrawal \u003crudyagrawal@google.com\u003e\n"
    },
    {
      "commit": "fa3670bf378edc25c59701630fbaa54ef8143dc9",
      "tree": "4abb9af1dee5115b4786de3714cffdc712e849c1",
      "parents": [
        "9497da1d49a405fa4aafcb8fc6cc6e32f18b5102"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Sun Oct 25 23:42:09 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Mon Oct 26 13:00:11 2026 -0700"
      },
      "message": "packagegroup-optee: Remove optee-examples\n\nRemove optee-examples sample applications and trusted applications from\npackagegroup-optee-apps.\n\nFusion-Link: fusion2 n/a\nTested: Built obmc-phosphor-image across AST2600 platforms; verified optee-examples is removed.\nGoogle-Bug-Id: 558894649\nChange-Id: I3a9720b04e4f029639eaf54bca8beaa6f3ead5f4\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "9497da1d49a405fa4aafcb8fc6cc6e32f18b5102",
      "tree": "98f3825a8b8aff18053cb6f0bdb95902b9826fa3",
      "parents": [
        "9accbe0373b9b5f361ba8e8d776732f5206de291"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Sun Oct 25 23:42:08 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Mon Oct 26 13:00:11 2026 -0700"
      },
      "message": "openssh: Enable function/data sections and GC\n\nCompile openssh with -ffunction-sections -fdata-sections and link with\n-Wl,--gc-sections so openssh binaries and static libopenssh.a consumers\n(hiba-id) garbage-collect unused functions.\n\nFusion-Link: fusion2 n/a\nTested: Built obmc-phosphor-image across AST2600 and NPCM845 platforms; verified openssh and hiba-id size reduction.\nGoogle-Bug-Id: 558894649\nChange-Id: Iac5dffdbf82937f59e37980bfbe7f79c70747815\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "9accbe0373b9b5f361ba8e8d776732f5206de291",
      "tree": "551414b291a676fe0bdb2c7e35fa62ca671eac58",
      "parents": [
        "69817cdc849a772ea9f1dd5bf03f5e2e66df4829"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Sun Oct 25 23:42:08 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Mon Oct 26 13:00:11 2026 -0700"
      },
      "message": "sysstat: Remove unused sadf, cifsiostat, and tapestat\n\nRemove unused sadf, cifsiostat, and tapestat binaries from sysstat on\ngBMC.\n\nFusion-Link: fusion2 n/a\nTested: Built obmc-phosphor-image across AST2600 and NPCM845 platforms; verified sysstat size reduction.\nGoogle-Bug-Id: 558894649\nChange-Id: I5ed2e62a45348f9092fbd23d8aa06ff47ed10588\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "69817cdc849a772ea9f1dd5bf03f5e2e66df4829",
      "tree": "0c5abb6147045074e3df4f09f9201d74332df177",
      "parents": [
        "1e45739e86cfa917994d5ce75846983427907067"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Sun Oct 25 23:42:07 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Mon Oct 26 13:00:11 2026 -0700"
      },
      "message": "shadow: Disable subordinate IDs and remove unused binaries\n\nDisable --enable-subordinate-ids on gBMC (dropping libsubid.so.4,\ngetsubids, newuidmap, and newgidmap) and remove interactive/batch\nutilities not used by phosphor-user-manager.\n\nFusion-Link: fusion2 n/a\nTested: Built obmc-phosphor-image across AST2600 and NPCM845 platforms; verified shadow size reduction.\nGoogle-Bug-Id: 558894649\nChange-Id: I32f40e547c613aa2b2da58d16971ea25856afe7e\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "1e45739e86cfa917994d5ce75846983427907067",
      "tree": "84519ea6764cd73a141860f45de22f35203d32b6",
      "parents": [
        "82ac70e7cff5d8469c0b0a4b1463b2e526af138f"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Sun Oct 25 23:42:07 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Mon Oct 26 13:00:11 2026 -0700"
      },
      "message": "mbedtls: Build static-only with function/data sections\n\nBuild mbedtls static-only with -fPIC -ffunction-sections -fdata-sections\nand omit sample/test programs so libmbedcrypto, libmbedx509, and\nlibmbedtls shared libraries are omitted from the target rootfs.\n\nFusion-Link: fusion2 n/a\nTested: Built obmc-phosphor-image across AST2600 and NPCM845 platforms; verified mbedtls shared libraries are removed.\nGoogle-Bug-Id: 558894649\nChange-Id: I5570d60a750e58d4fd5ba3457b76abba7cf27fc4\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    },
    {
      "commit": "82ac70e7cff5d8469c0b0a4b1463b2e526af138f",
      "tree": "30ae09a88696937d69f4b455e5acdcd9ee5d1a15",
      "parents": [
        "36d638deb1bdacbeebf804b3a3ec73a11b698be6"
      ],
      "author": {
        "name": "William A. Kennington III",
        "email": "wak@google.com",
        "time": "Sun Oct 25 23:42:06 2026 +0000"
      },
      "committer": {
        "name": "William Kennington",
        "email": "wak@google.com",
        "time": "Mon Oct 26 13:00:11 2026 -0700"
      },
      "message": "iperf3: Build static-only with GC\n\nBuild iperf3 static-only with -ffunction-sections/-fdata-sections and\n-Wl,--gc-sections so libiperf.so.0 is omitted from the target rootfs.\n\nFusion-Link: fusion2 n/a\nTested: Built obmc-phosphor-image across AST2600 and NPCM845 platforms; verified libiperf.so.0 is removed.\nGoogle-Bug-Id: 558894649\nChange-Id: I522402b849edf4892a9365e52c154ee314a86f16\nSigned-off-by: William A. Kennington III \u003cwak@google.com\u003e\n"
    }
  ],
  "next": "36d638deb1bdacbeebf804b3a3ec73a11b698be6"
}
