linux-kernel - Re: [regression v6.5-rc1] PCI: comm "swapper/0" leaking memory

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]

Date:   Tue, 5 Sep 2023 17:28:38 +0100
From:   Ross Lagerwall <ross.lagerwall@...ud.com>
To:     Kalle Valo <kvalo@...nel.org>
Cc:     Bjorn Helgaas <bhelgaas@...gle.com>, linux-pci@...r.kernel.org,
        linux-kernel@...r.kernel.org, ath11k@...ts.infradead.org,
        regressions@...ts.linux.dev
Subject: Re: [regression v6.5-rc1] PCI: comm "swapper/0" leaking memory

On Wed, Aug 30, 2023 at 10:21 AM Kalle Valo <kvalo@...nel.org> wrote:
>
> [CAUTION - EXTERNAL EMAIL] DO NOT reply, click links, or open attachments unless you have verified the sender and know the content is safe.
>
> Hi,
>
> I noticed that starting from v6.5-rc1 my ath11k tests reported several
> memory leaks from swapper/0:
>
> unreferenced object 0xffff88810a02b7a8 (size 96):
>   comm "swapper/0", pid 1, jiffies 4294671838 (age 98.120s)
>   hex dump (first 32 bytes):
>     80 b8 02 0a 81 88 ff ff b8 72 07 00 00 c9 ff ff  .........r......
>     c8 b7 02 0a 81 88 ff ff 00 00 00 00 00 00 00 00  ................
>   backtrace:
> unreferenced object 0xffff88810a02b880 (size 96):
>   comm "swapper/0", pid 1, jiffies 4294671838 (age 98.120s)
>   hex dump (first 32 bytes):
>     58 b9 02 0a 81 88 ff ff a8 b7 02 0a 81 88 ff ff  X...............
>     a0 b8 02 0a 81 88 ff ff 00 00 00 00 00 00 00 00  ................
>   backtrace:
> unreferenced object 0xffff88810a02b958 (size 96):
>   comm "swapper/0", pid 1, jiffies 4294671838 (age 98.120s)
>   hex dump (first 32 bytes):
>     30 ba 02 0a 81 88 ff ff 80 b8 02 0a 81 88 ff ff  0...............
>     78 b9 02 0a 81 88 ff ff 00 00 00 00 00 00 00 00  x...............
>   backtrace:
> unreferenced object 0xffff88810a02ba30 (size 96):
>   comm "swapper/0", pid 1, jiffies 4294671838 (age 98.120s)
>   hex dump (first 32 bytes):
>     08 bb 02 0a 81 88 ff ff 58 b9 02 0a 81 88 ff ff  ........X.......
>     50 ba 02 0a 81 88 ff ff 00 00 00 00 00 00 00 00  P...............
>   backtrace:
> unreferenced object 0xffff88810a02bb08 (size 96):
>   comm "swapper/0", pid 1, jiffies 4294671838 (age 98.120s)
>   hex dump (first 32 bytes):
>     e0 bb 02 0a 81 88 ff ff 30 ba 02 0a 81 88 ff ff  ........0.......
>     28 bb 02 0a 81 88 ff ff 00 00 00 00 00 00 00 00  (...............
>   backtrace:
>
> I can easily reproduce this by doing a simple insmod and rmmod of ath11k
> and it's dependencies (mac80211, MHI etc). I can reliability reproduce
> the leaks but I only see them once after a boot, I need to reboot the
> host to see the leaks again. v6.4 has no leaks.
>
> I did a bisect and found the commit below. I verified reverting the
> commit makes the leaks go away.
>
> commit e54223275ba1bc6f704a6bab015fcd2ae4f72572
> Author:     Ross Lagerwall <ross.lagerwall@...rix.com>
> AuthorDate: Thu May 25 16:32:48 2023 +0100
> Commit:     Bjorn Helgaas <bhelgaas@...gle.com>
> CommitDate: Fri Jun 9 15:06:16 2023 -0500
>
>     PCI: Release resource invalidated by coalescing
>
> Kalle
>

Hi Kalle,

I can't reproduce the leak by loading/unloading the ath11k module. I suspect
that the leak is always there when PCI resources are coalesced but
kmemleak doesn't notice until ath11k is loaded.

Can you please try the following to confirm it fixes it?

Ross

8<-----------------------

diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c
index 8bac3ce02609..907c873473e2 100644
--- a/drivers/pci/probe.c
+++ b/drivers/pci/probe.c
@@ -998,6 +998,7 @@ static int pci_register_host_bridge(struct
pci_host_bridge *bridge)
         res = window->res;
         if (!res->flags && !res->start && !res->end) {
             release_resource(res);
+            resource_list_destroy_entry(window);
             continue;
         }