linux-kernel - Re: [PATCH] bsg: call idr_pre

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]

Message-Id: <20100918123721B.fujita.tomonori@lab.ntt.co.jp>
Date:	Sat, 18 Sep 2010 12:40:14 +0900
From:	FUJITA Tomonori <fujita.tomonori@....ntt.co.jp>
To:	akpm@...ux-foundation.org
Cc:	naota@...sp.net, fujita.tomonori@....ntt.co.jp,
	linux-kernel@...r.kernel.org, axboe@...nel.dk, jkosina@...e.cz,
	gregkh@...e.de, daniel@...aq.de, linux-scsi@...r.kernel.org
Subject: Re: [PATCH] bsg: call idr_pre_get() outside the lock.

On Thu, 16 Sep 2010 16:37:35 -0700
Andrew Morton <akpm@...ux-foundation.org> wrote:

> On Wed, 01 Sep 2010 23:26:01 +0900
> Naohiro Aota <naota@...sp.net> wrote:
> 
> > The idr_pre_get() kernel-doc says "This function should be called
> > prior to locking and calling the idr_get_new* functions.", but the
> > idr_pre_get() calling in bsg_register_queue() is put inside
> > mutex_lock(). Let's fix it.
> > 
> 
> The idr_pre_get() kerneldoc is wrong.  Or at least, misleading.

Ah, thanks a lot. As usual, seems that there are few reliable
documents in kernel.


> The way this all works is that we precharge the tree via idr_pre_get()
> and we do this outside locks so we can use GFP_KERNEL.  Then we take
> the lock (a spinlock!) and then try to add an element to the tree,
> which will consume objects from the pool which was prefilled by
> idr_pre_get().
> 
> There's an obvious race here where someone else can get in and steal
> objects from the prefilled pool.  We can fix that with a retry loop:
> 
> 
> again:
> 	if (idr_pre_get(..., GFP_KERNEL) == NULL)
> 		return -ENOMEM;		/* We're really out-of-memory */
> 	spin_lock(lock);
> 	if (idr_get_new(...) == -EAGAIN) {
> 		spin_unlock(lock);
> 		goto again;		/* Someone stole our preallocation! */
> 	}
> 	...
> 
> this way we avoid the false -ENOMEM which the race would have caused. 
> We only declare -ENOMEM when we're REALLY out of memory.

Looks like there are many misuses idr_pre_get and idr_get_new; some
even calls idr_get_new without any lock (e.g. blk_alloc_devt).
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/