linux-kernel - Re: [REGRESSION] Long boot times due to USB enumeration

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <2025030559-radiated-reviver-eebb@gregkh>
Date: Wed, 5 Mar 2025 19:09:49 +0100
From: Greg KH <gregkh@...uxfoundation.org>
To: Seïfane Idouchach <seifane53@...il.com>
Cc: dirk.behme@...bosch.com, rafael@...nel.org, dakr@...nel.org,
	linux-kernel@...r.kernel.org, regressions@...ts.linux.dev,
	stable@...r.kernel.org
Subject: Re: [REGRESSION] Long boot times due to USB enumeration

On Thu, Mar 06, 2025 at 12:32:59AM +0800, Seïfane Idouchach wrote:
> Dear all,
> 
> I am reporting what I believe to be regression due to
> c0a40097f0bc81deafc15f9195d1fb54595cd6d0.
> 
> After this change I am experiencing long boot times on a setup that
> has what seems like a bad usb.
> The progress of the boot gets halted while retrying (and ultimately
> failing) to enumerate the USB device and is only allowed to continue
> after giving up enumerating the USB device.
> On Arch Linux this manifests itself by a message from SystemD having a
> wait job on journald. Journald starts just after the enumeration fails
> with "unable to enumerate USB device".
> This results in longer boot times on average 1 minute longer than
> usual (usually around 10s).
> No stable kernel before this change exhibits the issue all stable
> kernels after this change exhibit the issue.
> 
> See the related USB messages attached below (these messages are
> continuous and have not been snipped) :
> 
> [...]
> [    9.640854] usb 1-9: device descriptor read/64, error -110
> [   25.147505] usb 1-9: device descriptor read/64, error -110
> [   25.650779] usb 1-9: new high-speed USB device number 5 using xhci_hcd
> [   30.907482] usb 1-9: device descriptor read/64, error -110
> [   46.480900] usb 1-9: device descriptor read/64, error -110
> [   46.589883] usb usb1-port9: attempt power cycle
> [   46.990815] usb 1-9: new high-speed USB device number 6 using xhci_hcd
> [   51.791571] usb 1-9: Device not responding to setup address.
> [   56.801594] usb 1-9: Device not responding to setup address.
> [   57.010803] usb 1-9: device not accepting address 6, error -71
> [   57.137485] usb 1-9: new high-speed USB device number 7 using xhci_hcd
> [   61.937624] usb 1-9: Device not responding to setup address.
> [   66.947485] usb 1-9: Device not responding to setup address.
> [   67.154086] usb 1-9: device not accepting address 7, error -71
> [   67.156426] usb usb1-port9: unable to enumerate USB device

That's a real issue, but should not be due to the commit id you
referenced.

> [...]
> 
> This issue does not manifest in 44a45be57f85.

What does that commit have to do with this?  That's just a build break
fix.

> I am available to test any patches to address this on my system since
> I understand this could be quite hard to replicate on any system.
> I am available to provide more information if I am able or with
> guidance to help troubleshoot the issue further.
> 
> Wishing you all a good day.
> 
> #regzbot introduced: c0a40097f0bc81deafc15f9195d1fb54595cd6d0
> 

We know there are issues here.  That commit was "fixed" by commit
15fffc6a5624 ("driver core: Fix uevent_show() vs driver detach race"),
but then that caused a different problem, so it was reverted by commit
9a71892cbcdb ("Revert "driver core: Fix uevent_show() vs driver detach
race"").

There are many discussions about this on the mailing list, with a
proposal to add Dan's "fix" back.  If you could try that, it would be
great to see.

I think your USB problem is different here, but if you add 15fffc6a5624
("driver core: Fix uevent_show() vs driver detach race") to your kernel,
that would be great to see.

thanks,

greg k-h