mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: netdev-bot+sashiko@kernel.org
To: lcy8047@gmail.com
Cc: michael.chan@broadcom.com, pavan.chebbi@broadcom.com,
	andrew+netdev@lunn.ch, davem@davemloft.net, edumazet@kernel.org,
	kuba@kernel.org, pabeni@redhat.com, vikas.gupta@broadcom.com,
	andrew.gospodarek@broadcom.com, jacob.e.keller@intel.com,
	netdev@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH net] bnxt_en: Fix auxiliary device leak on early probe failures
Date: Fri, 20 Nov 2026 05:07:17 +0000	[thread overview]
Message-ID: <179515123717.3136.1891710789877841672@kernel.org> (raw)
In-Reply-To: <20261118050716.3873167-1-lcy8047@gmail.com>

Thank you for your contribution! Sashiko AI review found 3 potential
issue(s) to consider.

Critical: 0 · High: 0 · Medium: 2 · Low: 1

If you believe none of the issues are legitimate, please reply to each
issue of Medium severity or higher and explain why the AI is wrong.
If any of them are legitimate and you plan to respin, please let patchwork
know by including "pw-bot: cr" as a separate line at the end of your reply
(one such reply per series is enough).

- [Medium] In bnxt_init_one()
  (drivers/net/ethernet/broadcom/bnxt/bnxt.c), a bnxt_pf_wq allocation…
- [Medium] bnxt_en: skipped bnxt_clear_int_mode() on bnxt_pf_wq error in
  bnxt_init_one()
- [Low] The commit message lists the leaking failure points as
  bnxt_set_dflt_rings(), bnxt_init_int_mode() and the bnxt_pf_wq…

--

bnxt_en: Fix auxiliary device leak on early probe failures

This patch adds an init_err_auxdev label to bnxt_init_one(). Failures
after bnxt_aux_devices_init() now uninitialize the auxiliary devices and
free the auxdev ID. The bnxt_set_dflt_rings(), bnxt_init_int_mode() and
bnxt_pf_wq failure paths jump to the new label. On the register_netdev()
failure path, the aux cleanup now runs after bnxt_clear_int_mode().

> bnxt_init_one() allocates the auxiliary device ID and initializes the
> auxiliary devices before bnxt_set_dflt_rings(). If that call,
> bnxt_init_int_mode() or the bnxt_pf_wq allocation fails, probe jumps to
> init_err_pci_clean, which does not uninitialize the devices or free the
> ID. Only init_err_cleanup does.

[Severity: Low]

This list leaves out one path whose behaviour the patch also changes:
the bnxt_dl_register() failure that jumps to init_err_dl.

Before the patch, init_err_dl ran bnxt_shutdown_tc() and
bnxt_clear_int_mode() and then fell straight into init_err_pci_clean.
That skipped bnxt_aux_devices_uninit() and bnxt_auxdev_id_free(), so
aux_priv, edev, ulp_tbl and the bnxt_aux_dev_ids entry leaked on that
path too.

Now init_err_dl falls through into init_err_auxdev, so that path gets the
aux cleanup. The code change is fine. Could the commit message name this
path as well, so the scope is clear to backporters?

> diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
> index e78cf7ad3cbc8..ba0e458a467f5 100644
> --- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
> +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c

[ ... ]

> @@ -17449,7 +17449,7 @@ static int bnxt_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>  			if (!bnxt_pf_wq) {
>  				dev_err(&pdev->dev, "Unable to create workqueue.\n");
>  				rc = -ENOMEM;
> -				goto init_err_pci_clean;
> +				goto init_err_auxdev;

[Severity: Medium]
Does this path still leak the MSI-X vectors and bp->irq_tbl?

By the time create_singlethread_workqueue() runs, bnxt_init_int_mode()
has already succeeded. It has allocated the MSI-X vectors with
pci_alloc_irq_vectors() and allocated bp->irq_tbl. The new
init_err_auxdev label sits below bnxt_clear_int_mode() under
init_err_dl, so jumping to it skips the only call that releases them.
Nothing under init_err_pci_clean or init_err_free calls
pci_free_irq_vectors() or frees the table either. bnxt_cleanup_pci()
only unmaps the BARs, releases the regions and disables the device.

This can happen on the first PF probe, while bnxt_pf_wq is still NULL,
if the workqueue allocation fails. Once free_netdev() frees bp, the
irq_tbl pointer is lost. The vectors also stay allocated on bp->pdev
with msix_enabled still set. A later probe of the same function would
then trip the msix_enabled WARN_ON_ONCE() in __pci_enable_msix_range()
and fail to set up interrupts.

The old init_err_pci_clean target had the same gap. The commit message
says the unwind now follows the reverse order of setup, so should this
goto jump to init_err_dl instead?

At that point bp->tc_info is still NULL, so bnxt_shutdown_tc() returns
early. bnxt_clear_int_mode() then frees the vectors and irq_tbl, and
execution falls through into the new aux cleanup.

[Severity: Medium]

Does this path still leak the MSI-X vectors and bp->irq_tbl?

By this point bnxt_init_int_mode() has already succeeded. It has called
pci_alloc_irq_vectors() and allocated bp->irq_tbl. The new label sits
below bnxt_clear_int_mode():

init_err_dl:
	bnxt_shutdown_tc(bp);
	bnxt_clear_int_mode(bp);
init_err_auxdev:
	bnxt_aux_devices_uninit(bp);
	bnxt_auxdev_id_free(bp, bp->auxdev_id);

Nothing under init_err_pci_clean frees the vectors or the table either.
bnxt_cleanup_pci() only disables PTM, unmaps the BARs, releases the
regions and disables the device.

This can happen on the first PF probe, while bnxt_pf_wq is still NULL, if
create_singlethread_workqueue() fails. After free_netdev() the irq_tbl
pointer is lost and pdev->msix_enabled stays set.
msi_device_data_release() would then hit its WARN_ON_ONCE(!xa_empty(...))
on devres release. A later probe of the same function would also fail in
__pci_enable_msix_range():

	if (WARN_ON_ONCE(dev->msix_enabled))
		return -EINVAL;

The old init_err_pci_clean target had the same gap. Since this goto is
being rewritten to unwind in reverse order of setup, should it jump to
init_err_dl instead?

At that point bnxt_shutdown_tc() is a no-op because bp->tc_info is still
NULL. bnxt_clear_int_mode() would release the vectors and irq_tbl, and
execution would then fall through into the new aux cleanup.

[ ... ]

-- 
Sashiko AI review · https://netdev-ai.bots.linux.dev/sashiko/#/patchset/20261118050716.3873167-1-lcy8047%40gmail.com

      parent reply	other threads:[~2026-11-20  5:07 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-18  5:07 Changyul Lee
2026-11-18 17:23 ` Joe Damato
2026-11-20  5:07 ` netdev-bot+sashiko [this message]

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=179515123717.3136.1891710789877841672@kernel.org \
    --to=netdev-bot+sashiko@kernel.org \
    --cc=andrew+netdev@lunn.ch \
    --cc=andrew.gospodarek@broadcom.com \
    --cc=davem@davemloft.net \
    --cc=edumazet@kernel.org \
    --cc=jacob.e.keller@intel.com \
    --cc=kuba@kernel.org \
    --cc=lcy8047@gmail.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=michael.chan@broadcom.com \
    --cc=netdev@vger.kernel.org \
    --cc=pabeni@redhat.com \
    --cc=pavan.chebbi@broadcom.com \
    --cc=vikas.gupta@broadcom.com \
    /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

all inboxes | Powered by JetHome®