Re: [Bug 215925] New: PCIe regression on Raspberry Pi Compute Module 4 (CM4) breaks booting

[Bjorn Helgaas <helgaas@kernel.org>]

On Mon, May 09, 2022 at 09:44:29AM +0200, Thorsten Leemhuis wrote:
> Hi, this is your Linux kernel regression tracker. Partly top-posting to
> mnake this easily accessible.
> 
> Jim, what's up here? The regression was reported more than a week ago
> and it seems nothing happened since then. Or was there progress and I
> just missed it?
> 
> Anyway:
> 
> [TLDR: I'm adding this regression report to the list of tracked
> regressions; all text from me you find below is based on a few templates
> paragraphs you might have encountered already already in similar form.]
> 
> On 02.05.22 20:38, Bjorn Helgaas wrote:
> > On Sat, Apr 30, 2022 at 2:53 PM <bugzilla-daemon@kernel.org> wrote:
> >>
> >> https://bugzilla.kernel.org/show_bug.cgi?id=215925
> >>
> >>             Bug ID: 215925
> >>            Summary: PCIe regression on Raspberry Pi Compute Module 4 (CM4)
> >>                     breaks booting
> >>            Product: Drivers
> >>            Version: 2.5
> >>     Kernel Version: v5.17-rc1
> >>           Hardware: ARM
> >>                 OS: Linux
> >>               Tree: Mainline
> >>             Status: NEW
> >>           Severity: normal
> >>           Priority: P1
> >>          Component: PCI
> >>           Assignee: drivers_pci@kernel-bugs.osdl.org
> >>           Reporter: kibi@debian.org
> >>         Regression: No
> >>
> >> Catching up with latest kernel releases in Debian, it turned out that my
> >> Raspberry Pi Compute Module 4, mounted on an official Compute Module 4 IO
> >> Board,
> >> and booting from an SD card, no longer boots: this means a black screen on the
> >> HDMI output, and no output on the serial console.
> >>
> >> Trying various releases, I confirmed that v5.16 was fine, and v5.17-rc1 was the
> >> first (pre)release that wasn't.
> >>
> >> After some git bisect, it turns out the cause seems to be the following commit
> >> (https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=830aa6f29f07a4e2f1a947dfa72b3ccddb46dd21):
> >>
> >> ```
> >> commit 830aa6f29f07a4e2f1a947dfa72b3ccddb46dd21
> >> Author: Jim Quinlan <jim2101024@gmail.com>
> >> Date:   Thu Jan 6 11:03:27 2022 -0500
> >>
> >>     PCI: brcmstb: Split brcm_pcie_setup() into two funcs
> >> ```
> >>
> >> Starting with this commit, the kernel panics early (before 0.30 seconds), with
> >> an `Asynchronous SError Interrupt`. The backtrace references various
> >> `brcm_pcie_*` functions; I can share a picture or try and transcribe it
> >> manually if that helps (nothing on the serial console…).
> >>
> >> This commit is part of a branch that was ultimately merged as
> >> d0a231f01e5b25bacd23e6edc7c979a18a517b2b; starting with this commit, there's
> >> not even a backtrace anymore, the screen stays black after the usual “boot-up
> >> rainbow”, and there's still nothing on the serial console.
> >>
> >> I confirmed that 88db8458086b1dcf20b56682504bdb34d2bca0e2 (on the master side)
> >> was still booting properly, and that 87c71931633bd15e9cfd51d4a4d9cd685e8cdb55
> >> (from the branch being merged into master) is the last commit showing the
> >> panic.
> >>
> >> Since d0a231f01e5b25bacd23e6edc7c979a18a517b2b is a merge commit that includes
> >> conflict resolutions in drivers/pci/controller/pcie-brcmstb.c, I suppose this
> >> could be consistent with the initial panic being “upgraded” into an even more
> >> serious issue.
> >>
> >> I've also verified that latest master (v5.18-rc4-396-g57ae8a492116) is still
> >> affected by this issue.
> >>
> >> The regular Raspberry Pi 4 B doesn't seem to be affected by this issue: the
> >> exact same image on the same SD card (with latest master) boots fine on it.

Cyril, 830aa6f29f07 ("PCI: brcmstb: Split brcm_pcie_setup() into two
funcs") reverts cleanly as of 57ae8a492116.  Does reverting it avoid
the regression?