Re: [REGRESSION] Chrome and VSCode breakage with the commit b9b588f22a0c

[Chuck Lever <chuck.lever@oracle.com>]

On 2/26/25 4:40 PM, Eric Biggers wrote:
> On Wed, Feb 26, 2025 at 04:01:18PM -0500, Chuck Lever wrote:
>> On 2/26/25 3:42 PM, Eric Biggers wrote:
>>> On Wed, Feb 26, 2025 at 09:11:04AM -0500, Chuck Lever wrote:
>>>> On 2/26/25 3:38 AM, Takashi Iwai wrote:
>>>>> On Sun, 23 Feb 2025 16:18:41 +0100,
>>>>> Chuck Lever wrote:
>>>>>>
>>>>>> On 2/23/25 3:53 AM, Takashi Iwai wrote:
>>>>>>> [ resent due to a wrong address for regression reporting, sorry! ]
>>>>>>>
>>>>>>> Hi,
>>>>>>>
>>>>>>> we received a bug report showing the regression on 6.13.1 kernel
>>>>>>> against 6.13.0.  The symptom is that Chrome and VSCode stopped working
>>>>>>> with Gnome Scaling, as reported on openSUSE Tumbleweed bug tracker
>>>>>>>   https://bugzilla.suse.com/show_bug.cgi?id=1236943
>>>>>>>
>>>>>>> Quoting from there:
>>>>>>> """
>>>>>>> I use the latest TW on Gnome with a 4K display and 150%
>>>>>>> scaling. Everything has been working fine, but recently both Chrome
>>>>>>> and VSCode (installed from official non-openSUSE channels) stopped
>>>>>>> working with Scaling.
>>>>>>> ....
>>>>>>> I am using VSCode with:
>>>>>>> `--enable-features=UseOzonePlatform --enable-features=WaylandWindowDecorations --ozone-platform-hint=auto` and for Chrome, I select `Preferred Ozone platform` == `Wayland`.
>>>>>>> """
>>>>>>>
>>>>>>> Surprisingly, the bisection pointed to the backport of the commit
>>>>>>> b9b588f22a0c049a14885399e27625635ae6ef91 ("libfs: Use d_children list
>>>>>>> to iterate simple_offset directories").
>>>>>>>
>>>>>>> Indeed, the revert of this patch on the latest 6.13.4 was confirmed to
>>>>>>> fix the issue.  Also, the reporter verified that the latest 6.14-rc
>>>>>>> release is still affected, too.
>>>>>>>
>>>>>>> For now I have no concrete idea how the patch could break the behavior
>>>>>>> of a graphical application like the above.  Let us know if you need
>>>>>>> something for debugging.  (Or at easiest, join to the bugzilla entry
>>>>>>> and ask there; or open another bug report at whatever you like.)
>>>>>>>
>>>>>>> BTW, I'll be traveling tomorrow, so my reply will be delayed.
>>>>>>>
>>>>>>>
>>>>>>> thanks,
>>>>>>>
>>>>>>> Takashi
>>>>>>>
>>>>>>> #regzbot introduced: b9b588f22a0c049a14885399e27625635ae6ef91
>>>>>>> #regzbot monitor: https://bugzilla.suse.com/show_bug.cgi?id=1236943
>>>>>>
>>>>>> We received a similar report a few days ago, and are likewise puzzled at
>>>>>> the commit result. Please report this issue to the Chrome development
>>>>>> team and have them come up with a simple reproducer that I can try in my
>>>>>> own lab. I'm sure they can quickly get to the bottom of the application
>>>>>> stack to identify the misbehaving interaction between OS and app.
>>>>>
>>>>> Do you know where to report to?
>>>>
>>>> You'll need to drive this, since you currently have a working
>>>> reproducer. You can report the issue here:
>>>>
>>>> https://support.google.com/chrome/answer/95315?hl=en&co=GENIE.Platform%3DDesktop
>>>>
>>>>
>>>
>>> FYI this was already reported on the Chrome issue tracker 2 weeks ago:
>>> https://issuetracker.google.com/issues/396434686
>>
>> That appears to be as a response to the first report to us. Thanks for
>> finding this.
>>
>> I notice that this report indicates the problem is with a developer
>> build of Chrome, not a GA build.
>>
>> If /dev/dri is a tmpfs file system, then it would indeed be affected by
>> b9b588f22a0c. No indication yet of how.
> 
> Just to confirm, the commit did change the directory iteration order, right?

Yes, the order of entry iteration changed, but I thought it was going
back to the way tmpfs iterated directories before v6.6.

Also my impression is POSIX allows filesystems some leeway in that
order, and thus apps cannot depend on it. Makes me suspect there's some
other issue, like offset_readdir() is skipping an entry somehow.


> The theory at https://issuetracker.google.com/issues/396434686#comment4 seems
> promising.  Just the exact code hasn't been identified yet.

Yes.


-- 
Chuck Lever