public inbox for linux-usb@vger.kernel.org
 help / color / mirror / Atom feed
From: bugzilla-daemon@kernel.org
To: linux-usb@vger.kernel.org
Subject: [Bug 215740] kernel warning: DMA-API: xhci_hcd: cacheline tracking EEXIST, overlapping mappings aren't supported
Date: Thu, 26 Mar 2026 19:46:32 +0000	[thread overview]
Message-ID: <bug-215740-208809-diamE8VkAv@https.bugzilla.kernel.org/> (raw)
In-Reply-To: <bug-215740-208809@https.bugzilla.kernel.org/>

https://bugzilla.kernel.org/show_bug.cgi?id=215740

Mikhail (mikhail.v.gavrilov@gmail.com) changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mikhail.v.gavrilov@gmail.co
                   |                            |m

--- Comment #51 from Mikhail (mikhail.v.gavrilov@gmail.com) ---
Created attachment 309769
  --> https://bugzilla.kernel.org/attachment.cgi?id=309769&action=edit
dma_cacheline_eexist_repro.mod.c

I hit this exact warning on 7.0-rc5 with an ASUS USB Audio device
(0b05:1a5c) on AMD B650E (ROG STRIX B650E-I) + xhci_hcd 0000:0c:00.0.
Same stack trace through snd_usb_ctl_msg -> usb_control_msg ->
usb_hcd_map_urb_for_dma -> dma_map_phys -> add_dma_entry -> EEXIST.

I tested Guenter's patch from comment #43 (ARCH_KMALLOC_MINALIGN =
L1_CACHE_BYTES when CONFIG_DMA_API_DEBUG) and can confirm it fixes
the issue.

I also wrote a kernel module reproducer that triggers the bug reliably
on demand, without needing to wait for a specific USB device to be
plugged in during boot. The module does exactly what hub_configure()
does: two kmalloc(8) allocations back-to-back, then DMA-maps both.
On x86_64 with ARCH_KMALLOC_MINALIGN=8, they land in the same
64-byte cacheline within ~50 attempts, triggering the EEXIST warning.

Before fix:
  dma_repro: pair 53: buf_a=ffff8881c8f9eb80 buf_b=ffff8881c8f9eba0
             -- SAME cacheline 119793582!
  DMA-API: cacheline tracking EEXIST, overlapping mappings aren't
           supported

After fix (ARCH_KMALLOC_MINALIGN >= L1_CACHE_BYTES):
  dma_repro: no same-cacheline pair found in 64 attempts
             (ARCH_KMALLOC_MINALIGN may already be >= cache_line_size)

I can submit a formal patch based on Guenter's fix if nobody objects.

Reproducer module source attached.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching the assignee of the bug.

  parent reply	other threads:[~2026-03-26 19:46 UTC|newest]

Thread overview: 56+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-03-24 15:18 [Bug 215740] New: kernel warning: DMA-API: xhci_hcd: cacheline tracking EEXIST, overlapping mappings aren't supported bugzilla-daemon
2022-03-31  6:52 ` [Bug 215740] " bugzilla-daemon
2022-03-31  6:54 ` bugzilla-daemon
2022-03-31  7:41 ` [Bug 215740] New: " Thorsten Leemhuis
2022-04-05  5:19   ` Christoph Hellwig
2022-03-31  7:41 ` [Bug 215740] " bugzilla-daemon
2022-03-31  7:54 ` bugzilla-daemon
2022-04-05  5:19 ` bugzilla-daemon
2022-05-04  8:10 ` bugzilla-daemon
2022-05-04 10:17 ` bugzilla-daemon
2023-10-10 20:19 ` bugzilla-daemon
2023-11-12 20:46 ` bugzilla-daemon
2023-11-13 22:11 ` bugzilla-daemon
2023-11-14  3:25 ` bugzilla-daemon
2023-11-15 20:24 ` bugzilla-daemon
2023-11-16  2:19 ` bugzilla-daemon
2023-11-17 22:13 ` bugzilla-daemon
2023-11-19 16:47 ` bugzilla-daemon
2023-11-19 21:12 ` bugzilla-daemon
2023-11-20 19:43 ` bugzilla-daemon
2023-11-21 19:31 ` bugzilla-daemon
2023-11-21 20:51 ` bugzilla-daemon
2023-11-23 21:27 ` bugzilla-daemon
2023-11-24 12:02 ` bugzilla-daemon
2023-11-24 12:07 ` bugzilla-daemon
2023-11-24 15:27 ` bugzilla-daemon
2023-11-24 15:30 ` bugzilla-daemon
2023-11-24 15:40 ` bugzilla-daemon
2023-11-24 15:50 ` bugzilla-daemon
2023-11-24 16:09 ` bugzilla-daemon
2023-11-24 16:21 ` bugzilla-daemon
2023-11-24 21:05 ` bugzilla-daemon
2023-11-24 22:22 ` bugzilla-daemon
2023-11-25  2:03 ` bugzilla-daemon
2023-11-25 20:50 ` bugzilla-daemon
2023-11-25 21:17 ` bugzilla-daemon
2023-11-25 22:15 ` bugzilla-daemon
2023-11-28 19:04 ` bugzilla-daemon
2023-11-28 21:28 ` bugzilla-daemon
2023-11-29 15:44 ` bugzilla-daemon
2023-11-30 19:35 ` bugzilla-daemon
2024-09-24 16:50 ` bugzilla-daemon
2024-09-24 16:56 ` bugzilla-daemon
2025-06-03  4:08 ` bugzilla-daemon
2025-06-03 14:28 ` bugzilla-daemon
2025-06-03 18:31 ` bugzilla-daemon
2025-06-03 19:09 ` bugzilla-daemon
2026-01-22  4:53 ` bugzilla-daemon
2026-01-22  7:27 ` bugzilla-daemon
2026-01-22 19:38 ` bugzilla-daemon
2026-03-17 14:52 ` bugzilla-daemon
2026-03-17 16:18 ` bugzilla-daemon
2026-03-17 17:51 ` bugzilla-daemon
2026-03-26 19:46 ` bugzilla-daemon [this message]
2026-03-27  0:04 ` bugzilla-daemon
2026-03-27  7:32 ` bugzilla-daemon

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=bug-215740-208809-diamE8VkAv@https.bugzilla.kernel.org/ \
    --to=bugzilla-daemon@kernel.org \
    --cc=linux-usb@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox