Linux PCI subsystem development
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Leon Romanovsky" <leon@kernel.org>
Cc: media-ci@linuxtv.org, linux-pci@vger.kernel.org,
	kvm@vger.kernel.org, linux-rdma@vger.kernel.org,
	dri-devel@lists.freedesktop.org
Subject: Re: [PATCH v9 17/18] PCI/P2PDMA: Evaluate the ATS path for clients with ATS enabled
Date: Thu, 01 Oct 2026 12:07:56 +0000	[thread overview]
Message-ID: <20261001120756.A5CC41F000FF@smtp.kernel.org> (raw)
In-Reply-To: <20261001-fix-p2p-acs-v4-0-v9-17-1a8e0f50ddd9@nvidia.com>

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] pci/p2pdma: pcim_p2pdma_set_ats_per_mapping() silently fails for pure clients
--

commit 39432dea495f68fb0a21d0daa1f31ba24e130b39
Author: Leon Romanovsky <leon@kernel.org>

PCI/P2PDMA: Evaluate the ATS path for clients with ATS enabled

P2PDMA assumes every client issues Untranslated Requests. A device with
ATS enabled may translate any address it is handed, including a bus
address, and Translation Blocking can reject its Translated Requests on
a route the Untranslated answer called usable. This commit takes the
Address Type from its ATS Enable bit and avoids handing such a client
bus addresses unless the client declared per-mapping ATS.

> diff --git a/drivers/pci/p2pdma.c b/drivers/pci/p2pdma.c
> index 6a63901c65fb6..9c0943fdb3e36 100644
> --- a/drivers/pci/p2pdma.c
> +++ b/drivers/pci/p2pdma.c
[ ... ]
> @@ -1633,6 +1674,9 @@ pci_p2pdma_map_type_tlp(struct p2pdma_provider *provider, struct device *dev,
>  	cache_index = map_types_idx(client);
>  
>  	rcu_read_lock();
> +	/* The declaration belongs to the client, the cache to the provider. */
> +	p2pdma = rcu_dereference(client->p2pdma);
> +	per_mapping = p2pdma && p2pdma->ats_per_mapping;

[Severity: High]
Does this code handle pure P2PDMA clients correctly?

Since pure clients (devices initiating P2P transfers but not providing memory)
have no mechanism or reason to call pcim_p2pdma_init(), client->p2pdma will
be NULL for them.

This means when they call pcim_p2pdma_set_ats_per_mapping(), it silently returns
and drops the configuration:

drivers/pci/p2pdma.c:pcim_p2pdma_set_ats_per_mapping() {
    ...
	p2p = rcu_dereference_protected(pdev->p2pdma, 1);
	if (!p2p)
		/*
		 * ats_per_mapping is a performance optimization,
		 ...
		 */
		return;

	p2p->ats_per_mapping = true;
}

Because client->p2pdma is NULL, per_mapping will always evaluate to false
here in pci_p2pdma_map_type_tlp(). This leads to pci_p2pdma_client_map_type()
incorrectly overriding the map type, which could return
PCI_P2PDMA_MAP_NOT_SUPPORTED and break P2P transfers that should have been
permitted.

How should pure clients register their ATS per-mapping capability if they
do not initialize pdev->p2pdma?

>  	p2pdma = rcu_dereference(pdev->p2pdma);
>  
>  	if (p2pdma)

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261001-fix-p2p-acs-v4-0-v9-0-1a8e0f50ddd9@nvidia.com?part=17

  reply	other threads:[~2026-10-01 12:08 UTC|newest]

Thread overview: 59+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-10-01 11:55 [PATCH v9 00/18] PCI/P2PDMA: Route peer-to-peer DMA by TLP class Leon Romanovsky
2026-10-01 11:55 ` [PATCH v9 01/18] PCI/P2PDMA: Document the TLP attribute assumptions Leon Romanovsky
2026-10-01 11:58   ` sashiko-bot
2026-10-07 20:12   ` Bjorn Helgaas
2026-10-01 11:55 ` [PATCH v9 02/18] PCI/P2PDMA: Derive routing from directional ACS controls Leon Romanovsky
2026-10-01 12:03   ` sashiko-bot
2026-10-06 20:49   ` Bjorn Helgaas
2026-10-07  6:32     ` Leon Romanovsky
2026-10-07 20:40       ` Bjorn Helgaas
2026-10-01 11:55 ` [PATCH v9 03/18] PCI: Reject unreadable ACS controls in isolation checks Leon Romanovsky
2026-10-01 12:05   ` sashiko-bot
2026-10-01 11:55 ` [PATCH v9 04/18] PCI/P2PDMA: Evaluate ACS controls at the path divergence Leon Romanovsky
2026-10-01 12:02   ` sashiko-bot
2026-10-06 21:08   ` Bjorn Helgaas
2026-10-07 10:51     ` Leon Romanovsky
2026-10-01 11:55 ` [PATCH v9 05/18] PCI/P2PDMA: Document directional ACS routing Leon Romanovsky
2026-10-01 11:59   ` sashiko-bot
2026-10-01 15:25     ` Leon Romanovsky
2026-10-06 21:32   ` Bjorn Helgaas
2026-10-07 13:46     ` Leon Romanovsky
2026-10-07 20:10       ` Bjorn Helgaas
2026-10-01 11:55 ` [PATCH v9 06/18] PCI/P2PDMA: Collect the path's ACS controls before deciding Leon Romanovsky
2026-10-01 12:01   ` sashiko-bot
2026-10-06 21:48   ` Bjorn Helgaas
2026-10-07 13:50     ` Leon Romanovsky
2026-10-01 11:55 ` [PATCH v9 07/18] PCI/P2PDMA: Answer routing per TLP class Leon Romanovsky
2026-10-01 12:02   ` sashiko-bot
2026-10-01 11:55 ` [PATCH v9 08/18] PCI/P2PDMA: Route Relaxed Ordering Completions directly Leon Romanovsky
2026-10-01 12:04   ` sashiko-bot
2026-10-06 22:21   ` Bjorn Helgaas
2026-10-01 11:55 ` [PATCH v9 09/18] PCI/P2PDMA: Reject Translated Requests blocked by Translation Blocking Leon Romanovsky
2026-10-01 12:04   ` sashiko-bot
2026-10-01 11:55 ` [PATCH v9 10/18] PCI/P2PDMA: Route Translated Requests under Direct Translated P2P Leon Romanovsky
2026-10-01 12:02   ` sashiko-bot
2026-10-06 22:28   ` Bjorn Helgaas
2026-10-01 11:55 ` [PATCH v9 11/18] PCI/P2PDMA: Log detailed ACS routing diagnostics Leon Romanovsky
2026-10-01 12:06   ` sashiko-bot
2026-10-01 11:55 ` [PATCH v9 12/18] PCI/P2PDMA: Add KUnit tests for the ACS routing decisions Leon Romanovsky
2026-10-01 12:01   ` sashiko-bot
2026-10-01 11:55 ` [PATCH v9 13/18] PCI/P2PDMA: Test the ACS P2P routing walk Leon Romanovsky
2026-10-01 12:07   ` sashiko-bot
2026-10-01 11:55 ` [PATCH v9 14/18] PCI: Add KUnit coverage for ACS isolation checks Leon Romanovsky
2026-10-01 12:02   ` sashiko-bot
2026-10-01 11:55 ` [PATCH v9 15/18] PCI/P2PDMA: Document TLP-class routing Leon Romanovsky
2026-10-01 11:58   ` sashiko-bot
2026-10-01 11:55 ` [PATCH v9 16/18] PCI/P2PDMA: Let a client declare that it selects ATS per mapping Leon Romanovsky
2026-10-01 12:08   ` sashiko-bot
2026-10-01 12:59     ` Leon Romanovsky
2026-10-01 11:55 ` [PATCH v9 17/18] PCI/P2PDMA: Evaluate the ATS path for clients with ATS enabled Leon Romanovsky
2026-10-01 12:07   ` sashiko-bot [this message]
2026-10-01 15:23     ` Leon Romanovsky
2026-10-01 11:55 ` [PATCH v9 18/18] PCI/P2PDMA: Test the routing of " Leon Romanovsky
2026-10-01 12:04   ` sashiko-bot
2026-10-06 19:29 ` [PATCH v9 00/18] PCI/P2PDMA: Route peer-to-peer DMA by TLP class Bjorn Helgaas
2026-10-06 21:22   ` Leon Romanovsky
2026-10-06 21:36     ` Bjorn Helgaas
2026-10-07  6:56       ` Bjorn Helgaas
2026-10-07 10:54         ` Leon Romanovsky
2026-10-07 11:31           ` Bjorn Helgaas

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=20261001120756.A5CC41F000FF@smtp.kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=kvm@vger.kernel.org \
    --cc=leon@kernel.org \
    --cc=linux-pci@vger.kernel.org \
    --cc=linux-rdma@vger.kernel.org \
    --cc=media-ci@linuxtv.org \
    --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