Bug#1055161: linux-image-6.1.55-1-amd64: Dell R650xs hangs on reboot if ipxe.efi initialized network (tg3 driver bug)
Hi,
On Wed, Nov 01, 2023 at 02:16:14PM +0200, George Shuklin wrote:
> Package: linux-image-6.1.55-1-amd64
> Version: 6.1.38-4
> Severity: normal
> Tags: patch upstream
> X-Debbugs-Cc: george.shuklin@gmail.com
>
> When Dell R660xs server is booting using ipxe.efi (in UEFI mode), reboots are
> hanging because of tg3 driver.
>
> Steps to reproduce:
>
> 1. Switch server to EFI mode
> 2. Enable PXE on any interfaces
> 3. Enable PXE as first in boot order
> 4. Configure external dhcp/tftp server with ipxe.efi (of any version)
> 5. Boot. ipxe.efi will try to get loading parameters and exit, a normal linux
> is loaded
> 6. Assure tg3 driver is loaded (rmmod tg3; modprobe tg3)
> 7. Reboot
>
> Expected behaivor: reboot
>
> Actual behavior: system hanging after message `ACPI: PM: Preparing to enter
> system sleep state S5`.
>
> I belive this problem was present present for long time, but in 6.0 there was
> introduced commit 2ca1c94ce0b which changed the tg3_shutdown function.
>
> - if (system_state == SYSTEM_POWER_OFF)
> - tg3_power_down(tp);
> + tg3_power_down(tp);
>
> Because of that tg3_power_down now is called even when SYSTEM_RESTART. (I
> confirmed that it causes hangs on older kernels during power off, but it's less
> of the problem, because for servers it's expected to get server working back
> after reboot, not so much for poweroff).
>
> (ipxe.efi is important to reproduce this bug). Any network card can be used,
> just a presence of tg3 driver and embedded Broadcom Inc. and subsidiaries
> BCM57416 NetXtreme-E Dual-Media 10G RDMA Ethernet Controller is enough to
> trigger it.
>
> Restoring back this behavior is solving reboot problem. I've reported it to the
> Dell community forum:
> https://www.dell.com/community/en/users/647c085ec3b1ef2541ca0c64, but it's
> unknown if they fix EFI network driver or not.
>
> Patch to fix hanging is attached.
The issue is still present as well in Linux mainline?
Can you please report the regression to upstream (including the
regressions list) and from the get_maintainers.pl output:
Siva Reddy Kallam <siva.kallam@broadcom.com> (supporter:BROADCOM TG3 GIGABIT ETHERNET DRIVER)
Prashant Sreedharan <prashant@broadcom.com> (supporter:BROADCOM TG3 GIGABIT ETHERNET DRIVER)
Michael Chan <mchan@broadcom.com> (supporter:BROADCOM TG3 GIGABIT ETHERNET DRIVER)
"David S. Miller" <davem@davemloft.net> (maintainer:NETWORKING DRIVERS)
Eric Dumazet <edumazet@google.com> (maintainer:NETWORKING DRIVERS)
Jakub Kicinski <kuba@kernel.org> (maintainer:NETWORKING DRIVERS)
Paolo Abeni <pabeni@redhat.com> (maintainer:NETWORKING DRIVERS)
netdev@vger.kernel.org (open list:BROADCOM TG3 GIGABIT ETHERNET DRIVER)
linux-kernel@vger.kernel.org (open list)
Additionally the people who are from 2ca1c94ce0b6 ("tg3: Disable tg3
device on system reboot to avoid triggering AER"), at least Kai-Heng
Feng <kai.heng.feng@canonical.com>.
Regards,
Salvatore
Reply to: