From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from smtp.nokia.com ([131.228.20.173] helo=mgw-ext14.nokia.com) by canuck.infradead.org with esmtps (Exim 4.63 #1 (Red Hat Linux)) id 1IQ3GG-0002pz-7L for linux-mtd@lists.infradead.org; Tue, 28 Aug 2007 11:45:51 -0400 Subject: [PATCH] MTD: nandsim: avoid deadlocking FS From: Artem Bityutskiy To: David Woodhouse Content-Type: text/plain; charset=UTF-8 Date: Tue, 28 Aug 2007 18:45:40 +0300 Message-Id: <1188315941.3741.33.camel@sauron> Mime-Version: 1.0 Content-Transfer-Encoding: quoted-printable Cc: linux-mtd Reply-To: dedekind@infradead.org List-Id: Linux MTD discussion mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , From: Artem Bityutskiy Date: Tue, 28 Aug 2007 20:33:32 +0300 Subject: [PATCH] MTD: nandsim: avoid deadlocking FS Make nandsim use GFP_NOFS when allocating memory, because it might be used by a file-system (e.g. UBIFS2) which means, if we are short of memory, we may deadlock. Indee, UBIFS is holding a lock, writes to the media, reaches this place in NANDsim, kmalloc does not find the requested amount of RAM, calls memory shrinker, which decides to writeback inodes, calls FS, and it deadlocks on the lock which is already being held. Below is the UBIFS backtrace which demonstrates that: [] __mutex_lock_slowpath+0xc8/0x2e6 [] mutex_lock+0x1c/0x1f [] reserve_space+0x3d/0xa9 [ubifs] [] make_one_reservation+0x2b/0x86 [ubifs] [] ubifs_jrn_write_block+0xda/0x12f [ubifs] [] ubifs_writepage+0x11d/0x1ec [ubifs] [] shrink_inactive_list+0x7fa/0x969 [] shrink_zone+0xae/0x10c [] try_to_free_pages+0x159/0x251 [] __alloc_pages+0x125/0x2f0 [] cache_alloc_refill+0x380/0x6ba [] __kmalloc+0x14f/0x157 [] do_state_action+0xab7/0xc74 [nandsim] [] switch_state+0x225/0x402 [nandsim] [] ns_hwcontrol+0x3e2/0x620 [nandsim] [] nand_command+0x2e/0x1a5 [nand] [] nand_write_page+0x4a/0x9a [nand] [] nand_do_write_ops+0x1cf/0x343 [nand] [] nand_write+0x88/0xa6 [nand] [] part_write+0x72/0x8b [mtd] [] ubi_io_write+0x189/0x29c [ubi] [] ubi_eba_write_leb+0xb6/0x699 [ubi] [] ubi_leb_write+0xe4/0xe9 [ubi] [] ubifs_wbuf_write_nolock+0x333/0x4c9 [ubifs] [] write_node+0x74/0x8e [ubifs] [] ubifs_jrn_write_block+0x100/0x12f [ubifs] [] ubifs_writepage+0x11d/0x1ec [ubifs] [] __writepage+0xb/0x26 [] write_cache_pages+0x203/0x2d9 [] generic_writepages+0x23/0x2d [] do_writepages+0x37/0x39 [] __writeback_single_inode+0x96/0x399 [] sync_sb_inodes+0x1a3/0x274 [] writeback_inodes+0xa6/0xd8 [] background_writeout+0x86/0x9e [] pdflush+0xfb/0x1b6 [] kthread+0x37/0x59 [] kernel_thread_helper+0x7/0x14 The deadlock is funny because it starts in pdflush/writeback, and comes back to writeback, then deadlocks. It seems we should look carefully for other places in UBI and MTD and use GFP_NOFS instead of GFP_KERNEL. Caught-by: Adrian Hunter Signed-off-by: Artem Bityutskiy --- drivers/mtd/nand/nandsim.c | 8 +++++++- 1 files changed, 7 insertions(+), 1 deletions(-) diff --git a/drivers/mtd/nand/nandsim.c b/drivers/mtd/nand/nandsim.c index 205df0f..0b1823c 100644 --- a/drivers/mtd/nand/nandsim.c +++ b/drivers/mtd/nand/nandsim.c @@ -1272,7 +1272,13 @@ static int prog_page(struct nandsim *ns, int num) mypage =3D NS_GET_PAGE(ns); if (mypage->byte =3D=3D NULL) { NS_DBG("prog_page: allocating page %d\n", ns->regs.row); - mypage->byte =3D kmalloc(ns->geom.pgszoob, GFP_KERNEL); + /* + * We allocate memory with GFP_NOFS because a flash FS may + * utilize this. If it is holding an FS lock, then gets here, + * then kmalloc runs writeback which goes to the FS again + * and deadlocks. This was seen in practice. + */ + mypage->byte =3D kmalloc(ns->geom.pgszoob, GFP_NOFS); if (mypage->byte =3D=3D NULL) { NS_ERR("prog_page: error allocating memory for page %d\n", ns->regs.row= ); return -1; --=20 1.5.0.6 --=20 Best regards, Artem Bityutskiy (=D0=91=D0=B8=D1=82=D1=8E=D1=86=D0=BA=D0=B8=D0=B9 =D0=90= =D1=80=D1=82=D1=91=D0=BC)