[Kernel-packages] [Bug 2109804] [NEW] mlxbf_bootctl: bring in latest upstream commits

Public bug reported: SRU Justification: [Impact] The mlxbf-bootctl driver is not using the latest upstream commits. There is a SAUCE patch in Jammy that can be reverted and replaced with upstream. Also, there are three upstream patches that are missing from the mlxbf-bootctl driver in Jammy. [F

[Kernel-packages] [Bug 2109748] [NEW] bluefield_edac: uses uninitialized var in error paths

Public bug reported: SRU Justification: [Impact] The 'smatch' static analyzer tool raised 5 warnings against the EDAC driver. There is one warning for each call to "bluefield_edac_readl()" since the code uses the result of the function call even if there is an error (leaving resulting value uni

[Kernel-packages] [Bug 2101193] [NEW] mlxbf_gige: msglvl infrastructure

Public bug reported: SRU Justification: [Impact] The mlxbf_gige driver does not utilize the msglvl infrastructure, so there is no way to dynamically enable/disable certain debug log entries from the driver. [Fix] The fix is to add logic for setting a default msglvl, providing support for get/s

[Kernel-packages] [Bug 2100785] [NEW] mlxbf_gige: move BF2 autoneg logic out of open()

Public bug reported: SRU Justification: [Impact] The mlxbf_gige driver open() method includes a segment of BF2-specific code that waits for autoneg completion before proceeding. The open() method should not pend waiting for this link up event. [Fix] The fix is to move BF2 autoneg logic out of

[Kernel-packages] [Bug 2097022] [NEW] mlxbf_gige: emit messages during open and probe failures

Public bug reported: SRU Justification: [Impact] There are paths in the mlxbf_gige driver open() and probe() functions where a failure happens without a dmesg log entry. [Fix] The fix is to add dmesg log entries for all open() and probe() failures [Test Case] Unit tested only, since these ar

[Kernel-packages] [Bug 2085561] [NEW] bluefield_edac: replace SAUCE patches with upstream content

Public bug reported: SRU Justification: [Impact] The BlueField-1/BlueField-2 EDAC driver in Jammy repo consists of several SAUCE patches. There are now equivalent upstream patches that should be used instead. [Fix] 1) Revert the three SAUCE patches for bluefield_edac driver 2) Cherry pick fo

[Kernel-packages] [Bug 2077976] [NEW] mlxbf_livefish: driver build fails if CONFIG_ACPI not set

Public bug reported: SRU Justification: [Impact] The mlxbf_livefish driver depends on CONFIG_ACPI but that requirement is not specified in the driver's Kconfig. So, any attempt to build a kernel with CONFIG_ACPI not set and CONFIG_MLXBF_LIVEFISH set will result in a build error. [Fix] * Upda

[Kernel-packages] [Bug 2074123] [NEW] mlxbf_gige: packets received before RX path fully init

Public bug reported: SRU Justification: [Impact] During the reboot and power cycle testing, QA found an intermittent issue where "oob_net0" interface would experience RX MAC errors and fail to get a DHCP-assigned IP. There is a small window where packets can be received before the RX path is fu

[Kernel-packages] [Bug 2059951] [NEW] mlxbf_gige: stop interface during shutdown

Public bug reported: SRU Justification: [Impact] The mlxbf_gige driver shutdown() is invoked during reboot processing, but the stop() logic is not guaranteed to execute on all distros. If stop() does not execute NAPI remains enabled during system shutdown and can cause an exception if NAPI is sch

[Kernel-packages] [Bug 2059310] [NEW] mlxbf_gige: call request_irq() after NAPI initialized

Public bug reported: SRU Justification: [Impact] The mlxbf_gige driver encounters a NULL pointer exception in mlxbf_gige_open() when kdump is enabled. The exception happens because there is a pending RX interrupt before the call to request_irq(RX IRQ) executes. Then, the RX IRQ handler fires im

[Kernel-packages] [Bug 2059279] [NEW] mlxbf_gige: replace SAUCE patch for pause frame counters

Public bug reported: SRU Justification: [Impact] The support for mlxbf_gige pause frame counters was added to Jammy via a SAUCE patch. After upstream review, a small issue was fixed that makes the SAUCE patch different from the upstream commit. [Fix] The fix is to replace the SAUCE patch in the

[Kernel-packages] [Bug 2055086] [NEW] mlxbf_gige: add support for pause frame counters

Public bug reported: SRU Justification: [Impact] The BlueField-2 and BlueField-3 silicon designs include a provision for reporting pause frame counters, but the driver (mlxbf_gige) does not have software support yet. [Fix] The fix is to update the mlxbf_gige driver to support the "get_pause_st

[Kernel-packages] [Bug 2049930] [NEW] mlxbf_gige: disable RX_DMA during NAPI poll

Public bug reported: SRU Justification: [Impact] A regression has been introduced in the latest 22.04 kernel by including the following two commits: revert "UBUNTU: SAUCE: Fix OOB handling RX packets in heavy traffic" addition of upstream "mlxbf_gige: fix receive packet race condition" The

[Kernel-packages] [Bug 2047853] [NEW] mlxbf_gige: replace SAUCE patch for RX race condition

Public bug reported: SRU Justification: [Impact] There is a SAUCE patch that addresses a mlxbf-gige RX race condition "UBUNTU: SAUCE: Fix OOB handling RX packets in heavy traffic" This SAUCE patch should be reverted and replaced with upstream content. [Fix] Bring in the following upstream c

[Kernel-packages] [Bug 2045919] [NEW] mlxbf-bootctl: fails to report info about cards using dev keys

Public bug reported: SRU Justification: [Impact] There is a gap in the mlxbf-bootctl driver logic, resulting in failure to report when a secure boot enabled card is using development keys. The program will instead report lifecycle state as "GA Secured" which is misleading. [Fix] Bring in the f

[Kernel-packages] [Bug 2035374] [NEW] pwr-mlxbf: update Kconfig with upstream changes

Public bug reported: SRU Justification: [Impact] The Mellanox BlueField power driver (pwr-mlxbf) has a dependency on either the BlueField-2 or BlueField-3 GPIO driver. The current Kconfig does not have the BlueField-3 dependency. [Fix] There is an upstream commit that extends the Kconfig to

[Kernel-packages] [Bug 2035123] [NEW] scripts/pahole-flags.sh needs upstream fix

Public bug reported: When doing a kernel make in our Jammy repo we notice the follow error being emitted: ./scripts/pahole-flags.sh: line 7: return: can only `return' from a function or sourced script It appears that while the Jammy baseline is 5.15.99 (as per the top-level Makefile) the file s

[Kernel-packages] [Bug 2033551] [NEW] mlxbf_bootctl: replace SAUCE patches with upstream

Public bug reported: SRU Justification: [Impact] The mlxbf_bootctl BlueField platform driver in the Jammy repo consists of some SAUCE patches. These need to be replaced. [Fix] The fix is to revert the four SAUCE patches, replacing them with upstream commits for the same functionality. One patc

[Kernel-packages] [Bug 2012649] [NEW] mlxbf_gige: need to replace SAUCE patches with upstream commits

Public bug reported: SRU Justification: [Impact] The GIGE driver ("mlxbf_gige") in the Jammy repo contains SAUCE patches that enable BlueField-3 support. This same functionality is contained in upstream commits and Jammy repo should utilize them. [Fix] The BlueField-3 support will be provided

[Kernel-packages] [Bug 1995148] [NEW] mlxbf_gige: need to add BlueField-3 support

Public bug reported: SRU Justification: [Impact] The GIGE driver ("mlxbf_gige") in the Jammy repo does not yet support the BlueField-3 SoC. The "mlxbf_gige" driver which supports BlueField-2 SoC currently will be extended to include the BlueField-3 support. [Fix] The BlueField-3 support will b

[Kernel-packages] [Bug 1991403] [NEW] mlxbf_gige: fix devm_ioremap error handling

Public bug reported: SRU Justification: [Impact] The following upstream commit was recently added to Jammy: 3a1a274e933f mlxbf_gige: compute MDIO period based on i1clk This patch added driver logic that includes a "devm_ioremap" call, but the error handling logic pertaining to this call is in

[Kernel-packages] [Bug 1991151] Re: mlxbf_gige: need to revert SAUCE patch for "clear MDIO gateway lock"

Sorry as this has become confusing. This bug does still apply to Jammy linux-bluefield, and the fix is contained in a patch series that I sent in yesterday: v3--cover-letter.patch v3-0001-UBUNTU-SAUCE-Revert-UBUNTU-SAUCE-mlxbf_gige-clear.patch v3-0002-mlxbf_gige-clear-MDIO-gateway-lock-af

[Kernel-packages] [Bug 1991151] Re: mlxbf_gige: need to revert SAUCE patch for "clear MDIO gateway lock"

This bug is within the Jammy repo of the BlueField project (22.04). I sent in the patch with a subject line containing this prefix: "[SRU][F:linux-bluefield]" Is that relevant for 22.04 patches, or should the "F" be replaced with something to signify Jammy? -- You received this bug notification

[Kernel-packages] [Bug 1991151] [NEW] mlxbf_gige: need to revert SAUCE patch for "clear MDIO gateway lock"

Public bug reported: SRU Justification: [Impact] The Jammy repo contains the following SAUCE patch: UBUNTU: SAUCE: mlxbf_gige: clear MDIO gateway lock after read This patch is now in mainline Linux kernel, so the SAUCE patch can be reverted and replaced with the upstream commit SHA. [Fix] The

[Kernel-packages] [Bug 1989495] [NEW] mlxbf_gige: need to clear MDIO gateway lock after read

Public bug reported: SRU Justification: [Impact] The BlueField-2 GIGE logic accesses the MDIO device via reads/writes to a gateway (GW) register. The MDIO GW lock is set after read. Make sure to always clear to indicate that the GW register is not being used. If the lock is mistakenly interpreted

[Kernel-packages] [Bug 1975749] [NEW] mlxbf_gige: remove driver-managed interrupt counts

Public bug reported: SRU Justification: [Impact] The mlxbf_gige driver currently has three driver-managed interrupt counts that increment each time every interrupt handler executes. This has been reviewed and found to be unnecessary given core logic in Linux kernel manages the same information.

[Kernel-packages] [Bug 1974246] [NEW] mlxbf_gige: increase MDIO polling rate

Public bug reported: SRU Justification: [Impact] There is MDIO logic within the mlxbf_gige driver that currently polls for completed transactions every 100us. This has been reviewed and tested in the lab and found to be inefficient. [Fix] The fix is to modify the driver logic to increase the

[Kernel-packages] [Bug 1974241] [NEW] mlxbf_gige: revert "add interrupt counts" commit

Public bug reported: SRU Justification: [Impact] There is a mlxbf_gige driver commit (ed0dd97e64a5) which pertains to "UBUNTU: SAUCE: mlxbf_gige: add interrupt counts to "ethtool -S". This logic is modified in upstream Linux, so this SAUCE commit will need to be reverted. [Fix] The fix is to

[Kernel-packages] [Bug 1943065] [NEW] mlxbf_gige: update driver version to 1.25

Public bug reported: SRU Justification: [Impact] * The mlxbf_gige driver needs a version increase to identify a driver containing recent important bug fixes. [Fix] * The driver version implemented by the DRV_VERSION define and MODULE_VERSION macro will be incremented from 1.24 to 1.25 [Test

[Kernel-packages] [Bug 1942932] [NEW] oob_net0 up but stops processing Rx packets

Public bug reported: SRU Justification: [Impact] * The oob_net0 interface, managed by the mlxbf_gige driver, will be up and enabled but can stop processing received packets if three specific conditions occur: 1) driver has received more than (N * RxRingSize) packets but less than (N+1

[Kernel-packages] [Bug 1942931] [NEW] "ethtool -S oob_net0" does not report interrupt counts

Public bug reported: SRU Justification: [Impact] * The command "ethtool -S oob_net0" displays a set of statistics related to mlxbf_gige driver. This set of statistics does not include the driver's interrupt counts. [Fix] * The fix is to extend the statistics list supported by "ethtool -S"