[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <cd2a4ea4-52d2-cf95-7769-859b0a35b564@canonical.com>
Date: Tue, 21 Jun 2022 13:27:33 -0700
From: John Johansen <john.johansen@...onical.com>
To: Matthew Wilcox <willy@...radead.org>
Cc: Linus Torvalds <torvalds@...ux-foundation.org>,
"Eric W. Biederman" <ebiederm@...ssion.com>,
Ammar Faizi <ammarfaizi2@...weeb.org>,
James Morris <jmorris@...ei.org>,
LSM List <linux-security-module@...r.kernel.org>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
Al Viro <viro@...iv.linux.org.uk>,
Kees Cook <keescook@...omium.org>,
linux-fsdevel@...r.kernel.org, Linux-MM <linux-mm@...ck.org>,
gwml@...r.gnuweeb.org
Subject: Re: Linux 5.18-rc4
On 6/13/22 15:48, Matthew Wilcox wrote:
> On Mon, Jun 06, 2022 at 02:00:33PM -0700, John Johansen wrote:
>> On 6/6/22 13:23, Matthew Wilcox wrote:
>>> On Mon, Jun 06, 2022 at 12:19:36PM -0700, John Johansen wrote:
>>>>> I suspect that part is that both Apparmor and IPC use the idr local lock.
>>>>>
>>>> bingo,
>>>>
>>>> apparmor moved its secids allocation from a custom radix tree to idr in
>>>>
>>>> 99cc45e48678 apparmor: Use an IDR to allocate apparmor secids
>>>>
>>>> and ipc is using the idr for its id allocation as well
>>>>
>>>> I can easily lift the secid() allocation out of the ctx->lock but that
>>>> would still leave it happening under the file_lock and not fix the problem.
>>>> I think the quick solution would be for apparmor to stop using idr, reverting
>>>> back at least temporarily to the custom radix tree.
>>>
>>> How about moving forward to the XArray that doesn't use that horrid
>>> prealloc gunk? Compile tested only.
>>>
>>
>> I'm not very familiar with XArray but it does seem like a good fit. We do try
>> to keep the secid allocation dense, ideally no holes. Wrt the current locking
>> issue I want to hear what Thomas has to say. Regardless I am looking into
>> whether we should just switch to XArrays going forward.
>
> Nothing from Thomas ... shall we just go with this? Do you want a
> commit message, etc for the patch?
Hey Matthew,
I have done testing with this and the patch looks good. We will certainly
go this route going forward so a commit message, would be good. As for
fixing the issue in current kernels. I am not opposed to pulling this
back as fixes for
99cc45e48678 apparmor: Use an IDR to allocate apparmor secids
but I would like some other peoples opinions on doing that, because we
don't understand why we are getting the current lock splat, and without
understanding it is a fix by avoiding the issue rather than being sure
the actual issue is fixed.
Powered by blists - more mailing lists