Linux CXL
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Li Ming" <ming.li@zohomail.com>
Cc: linux-cxl@vger.kernel.org
Subject: Re: [PATCH] cxl/region: Fix NULL pointer within p->targets[]
Date: Sat, 30 May 2026 05:06:49 +0000	[thread overview]
Message-ID: <20260530050649.EBFCD1F00893@smtp.kernel.org> (raw)
In-Reply-To: <20260530-fix_null_in_targets_array-v1-1-312c3bf1fe0f@zohomail.com>

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:

Pre-existing issues:
- [High] __cxl_decoder_detach() suffers from the exact same p->targets hole vulnerability for fully assembled auto-discovered regions, which will cause a NULL pointer dereference upon re-attachment.
- [High] Failed auto-discovery decoders are left in the region targets array if failure occurs after endpoint port attachment, leading to a Use-After-Free (UAF).
--

commit 8beb15c55def2c0ba7cda46a708f6b3d2ab5e643
Author: Li Ming <ming.li@zohomail.com>

cxl/region: Fix NULL pointer within p->targets[]

This patch fixes a NULL pointer issue in p->targets[] during endpoint
decoder removal. It ensures the last valid endpoint decoder pointer is
swapped into the removed slot, avoiding gaps in the valid targets array.

> diff --git a/drivers/cxl/core/region.c b/drivers/cxl/core/region.c
> index e90c024c80365..54018db87a4c9 100644
> --- a/drivers/cxl/core/region.c
> +++ b/drivers/cxl/core/region.c
> @@ -2220,7 +2220,15 @@ static int cxl_region_remove_target(struct device *dev, void *data)
>  			p->nr_targets--;
>  			cxled->state = CXL_DECODER_STATE_AUTO;
>  			cxled->pos = -1;
> -			p->targets[i] = NULL;
> +
> +			/*
> +			 * Swap the last valid target into the slot to
> +			 * ensure no invalid target in p->nr_targets range.
> +			 * The targets array will be re-sorted during the
> +			 * last endpoint decoder attaching again.
> +			 */
> +			p->targets[i] = p->targets[p->nr_targets];
> +			p->targets[p->nr_targets] = NULL;

[Severity: High]
This isn't a bug introduced by this patch, but does __cxl_decoder_detach()
suffer from the exact same p->targets hole vulnerability for fully assembled
auto-discovered regions?

When a target is detached from an auto-discovered region:

drivers/cxl/core/region.c:__cxl_decoder_detach() {
    ...
	p->targets[cxled->pos] = NULL;
	p->nr_targets--;
    ...
}

For auto-discovered regions, new targets are appended at
p->targets[p->nr_targets] by cxl_region_attach_auto(). This leaves the NULL
hole intact while overwriting a valid endpoint pointer that was previously
positioned at the end of the array. Once p->nr_targets reaches
p->interleave_ways again, cxl_region_sort_targets() iterates through the
array and attempts to dereference the NULL pointer, which could cause a
kernel panic.


[Severity: High]
This isn't a bug introduced by this patch, but are failed auto-discovery
decoders left in the region targets array if failure occurs after endpoint
port attachment, leading to a Use-After-Free?

cxl_rr_ep_add() correctly upgrades an endpoint's staging state to
CXL_DECODER_STATE_AUTO:

drivers/cxl/core/region.c:cxl_rr_ep_add() {
    ...
		if (cxld == &cxled->cxld &&
		    cxled->state == CXL_DECODER_STATE_AUTO_STAGED)
			cxled->state = CXL_DECODER_STATE_AUTO;
    ...
}

However, if subsequent topology validation fails, the error path calls
cxl_port_detach_region(), which clears the region pointer but neglects to
restore the endpoint's state back to CXL_DECODER_STATE_AUTO_STAGED. Because
of this, when __cxl_decoder_detach() invokes cxl_cancel_auto_attach(), it
silently skips removal due to the state mismatch:

drivers/cxl/core/region.c:cxl_cancel_auto_attach() {
    ...
	if (cxled->state != CXL_DECODER_STATE_AUTO_STAGED)
		return;
    ...
}

The endpoint decoder remains in the p->targets array and is subsequently
freed upon removal, resulting in a persistent Use-After-Free vulnerability
when the sysfs region targets are read or the array is otherwise accessed.

>  
>  			return 1;
>  		}

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260530-fix_null_in_targets_array-v1-1-312c3bf1fe0f@zohomail.com?part=1

  reply	other threads:[~2026-05-30  5:06 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-05-30  4:24 [PATCH] cxl/region: Fix NULL pointer within p->targets[] Li Ming
2026-05-30  5:06 ` sashiko-bot [this message]
2026-05-30 14:54   ` Li Ming
2026-06-03 22:40 ` Alison Schofield
2026-06-04 13:28   ` Li Ming
2026-06-04 15:45     ` Dave Jiang

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=20260530050649.EBFCD1F00893@smtp.kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=linux-cxl@vger.kernel.org \
    --cc=ming.li@zohomail.com \
    --cc=sashiko-reviews@lists.linux.dev \
    /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