Re: [REGRESSION] 527139d738d7 ("PCI/sysfs: Convert "rom" to static attribute")

From: Thorsten Leemhuis <regressions@leemhuis.info>
To: Bjorn Helgaas <helgaas@kernel.org>
Cc: "Krzysztof Wilczyński" <kw@linux.com>,
	"Ville Syrjälä" <ville.syrjala@linux.intel.com>,
	"Bjorn Helgaas" <bhelgaas@google.com>,
	"Oliver O'Halloran" <oohall@gmail.com>,
	linux-pci@vger.kernel.org
Subject: Re: [REGRESSION] 527139d738d7 ("PCI/sysfs: Convert "rom" to static attribute")
Date: Thu, 20 Jan 2022 16:05:19 +0100	[thread overview]
Message-ID: <5c4982af-dc61-1578-ad47-aa5abc0efbe1@leemhuis.info> (raw)
In-Reply-To: <20220120150051.GA1015185@bhelgaas>

On 20.01.22 16:00, Bjorn Helgaas wrote:
> On Thu, Jan 20, 2022 at 02:19:05PM +0100, Thorsten Leemhuis wrote:
>> On 17.12.21 23:49, Krzysztof Wilczyński wrote:
>>> Hi Ville,
>>>
>>> Thank you for letting us know, and sincere apologies for troubles!
>>>
>>> [...]
>>>>>> The pci sysfs "rom" file has disappeared for VGA devices.
>>>>>> Looks to be a regression from commit 527139d738d7 ("PCI/sysfs:
>>>>>> Convert "rom" to static attribute").
>>>>>>
>>>>>> Some kind of ordering issue between the sysfs file creation 
>>>>>> vs. pci_fixup_video() perhaps?
>>>>>
>>>>> Can you attach your complete "lspci -vv" output?  Also, which is the
>>>>> default device?  I think there's a "boot_vga" sysfs file that shows
>>>>> this.  "find /sys -name boot_vga | xargs grep ."
>>>>
>>>> All I have is Intel iGPUs so it's always 00:02.0. 
>>>>
>>>> $ cat /sys/bus/pci/devices/0000\:00\:02.0/boot_vga 
>>>> 1
>>>> $ cat /sys/bus/pci/devices/0000\:00\:02.0/rom
>>>> cat: '/sys/bus/pci/devices/0000:00:02.0/rom': No such file or directory
>>>>
>>>> I've attached the full lspci from my IVB laptop, but the problem
>>>> happens on every machine (with an iGPU at least).
>>>>
>>>> I presume with a discrete GPU it might not happen since they
>>>> actually have a real ROM.
>>>
>>> Admittedly, the automated testing I was running before the patch was released
>>> didn't catch this.  I primarily focused on trying to catch the race condition
>>> related to the ROM attribute creation.
>>>
>>> I need to look into how to properly address this problem as if we were to
>>> revert the ROM attribute changes, then we would introduce the race condition
>>> we've had back.
>>>
>>> Again, apologies for troubles this caused!
>>
>> What's the status of this regression and getting it fixed? It looks like
>> there was no progress for quite a while. Could anyone please provide a
>> status update?
> 
> What a coincidence.  Krzysztof and I chatted about this yesterday.  No
> progress to report yet, but we are working on it.

Many thx for the update! Ciao, Thorsten