Re: [BUG] rbtree bug with mmotm 2009-04-14-17-24

From: Steve Wise
Date: Wed Apr 22 2009 - 14:16:32 EST


Jens Axboe wrote:

Don't bother, I see what the bug is now. It's caused by commit
a36e71f996e25d6213f57951f7ae1874086ec57e and it's due to ->ioprio being
changed without the prio rb root being adjusted.

I'll provide a fix shortly.

Quick'n dirty, I think this should fix the issue.

diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c
index 7e13f04..79ebb4c 100644
--- a/block/cfq-iosched.c
+++ b/block/cfq-iosched.c
@@ -594,7 +594,7 @@ cfq_prio_tree_lookup(struct cfq_data *cfqd, int ioprio, sector_t sector,
static void cfq_prio_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq)
{
- struct rb_root *root = &cfqd->prio_trees[cfqq->ioprio];
+ struct rb_root *root = &cfqd->prio_trees[cfqq->org_ioprio];
struct rb_node **p, *parent;
struct cfq_queue *__cfqq;
@@ -606,8 +606,8 @@ static void cfq_prio_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq)
if (!cfqq->next_rq)
return;
- __cfqq = cfq_prio_tree_lookup(cfqd, cfqq->ioprio, cfqq->next_rq->sector,
- &parent, &p);
+ __cfqq = cfq_prio_tree_lookup(cfqd, cfqq->org_ioprio,
+ cfqq->next_rq->sector, &parent, &p);
BUG_ON(__cfqq);
rb_link_node(&cfqq->p_node, parent, p);
@@ -656,8 +656,10 @@ static void cfq_del_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
if (!RB_EMPTY_NODE(&cfqq->rb_node))
cfq_rb_erase(&cfqq->rb_node, &cfqd->service_tree);
- if (!RB_EMPTY_NODE(&cfqq->p_node))
- rb_erase_init(&cfqq->p_node, &cfqd->prio_trees[cfqq->ioprio]);
+ if (!RB_EMPTY_NODE(&cfqq->p_node)) {
+ rb_erase_init(&cfqq->p_node,
+ &cfqd->prio_trees[cfqq->org_ioprio]);
+ }
BUG_ON(!cfqd->busy_queues);
cfqd->busy_queues--;
@@ -976,7 +978,7 @@ static struct cfq_queue *cfqq_close(struct cfq_data *cfqd,
* First, if we find a request starting at the end of the last
* request, choose it.
*/
- __cfqq = cfq_prio_tree_lookup(cfqd, cur_cfqq->ioprio,
+ __cfqq = cfq_prio_tree_lookup(cfqd, cur_cfqq->org_ioprio,
sector, &parent, NULL);
if (__cfqq)
return __cfqq;
@@ -1559,8 +1561,9 @@ cfq_alloc_io_context(struct cfq_data *cfqd, gfp_t gfp_mask)
static void cfq_init_prio_data(struct cfq_queue *cfqq, struct io_context *ioc)
{
+ struct cfq_data *cfqd = cfqq->cfqd;
struct task_struct *tsk = current;
- int ioprio_class;
+ int ioprio_class, prio_readd = 0;
if (!cfq_cfqq_prio_changed(cfqq))
return;
@@ -1592,12 +1595,24 @@ static void cfq_init_prio_data(struct cfq_queue *cfqq, struct io_context *ioc)
}
/*
+ * Remove us from the prio_tree if we are present, since we index
+ * by ->org_ioprio
+ */
+ if (!RB_EMPTY_NODE(&cfqq->p_node)) {
+ rb_erase(&cfqq->p_node, &cfqd->prio_trees[cfqq->org_ioprio]);
+ prio_readd = 1;
+ }
+
+ /*
* keep track of original prio settings in case we have to temporarily
* elevate the priority of this queue
*/
cfqq->org_ioprio = cfqq->ioprio;
cfqq->org_ioprio_class = cfqq->ioprio_class;
cfq_clear_cfqq_prio_changed(cfqq);
+
+ if (prio_readd)
+ cfq_prio_tree_add(cfqd, cfqq);
}
static void changed_ioprio(struct io_context *ioc, struct cfq_io_context *cic)


Still crashes:

Starting udev: BUG: unable to handle kernel NULL pointer dereference at (null)
IP: [<ffffffff8034ee05>] rb_erase+0x130/0x2a7
PGD 12cd05067 PUD 12e09f067 PMD 0
Oops: 0000 [#1] SMP
last sysfs file: /sys/block/sda/sda3/dev
CPU 0
Modules linked in: snd_hda_codec_intelhdmi snd_hda_codec_realtek snd_hda_intel snd_hda_codec snd_seq_dummy snd_seq_oss snd_seq_midi_event snd_seq snd_seq_device snd_pcm_oss snd_mixer_oss snd_pcm cxgb3 r8169 snd_timer snd sg sr_mod cdrom rtc_cmos rtc_core button i2c_i801 serio_raw floppy soundcore shpchp mii rtc_lib i2c_core snd_page_alloc pcspkr dm_snapshot dm_zero dm_mirror dm_region_hash dm_log dm_mod ata_piix libata sd_mod scsi_mod ext3 jbd uhci_hcd ohci_hcd ehci_hcd
Pid: 2381, comm: vol_id Not tainted 2.6.30-rc2-jens #11 P5E-VM HDMI
RIP: 0010:[<ffffffff8034ee05>] [<ffffffff8034ee05>] rb_erase+0x130/0x2a7
RSP: 0018:ffff88012cc7f7f8 EFLAGS: 00010046
RAX: ffff88012cde2a21 RBX: ffff88012cde2a20 RCX: 0000000000000000
RDX: ffff88012cde2a20 RSI: ffff88012d421e30 RDI: 0000000000000000
RBP: ffff88012cc7f808 R08: 0000000000000000 R09: ffff88012d421e30
R10: ffff88012fb8a140 R11: ffff88012f03e3c0 R12: ffff88012d421e30
R13: ffff88012d421e00 R14: ffff88012cde2630 R15: 0000000000000000
FS: 00000000006e3880(0063) GS:ffff88002804b000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b
CR2: 0000000000000000 CR3: 000000012b454000 CR4: 00000000000006e0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
Process vol_id (pid: 2381, threadinfo ffff88012cc7e000, task ffff88012a901530)
Stack:
ffff88012cde2660 ffff88012d5a1788 ffff88012cc7f828 ffffffff8034690d
ffff88012cc7f828 ffff88012cde2630 ffff88012cc7f868 ffffffff803478e5
ffff88012d421e00 ffff88012cde2630 ffff88012d5a1788 ffff88012d421e00
Call Trace:
[<ffffffff8034690d>] rb_erase_init+0x11/0x21
[<ffffffff803478e5>] cfq_remove_request+0x184/0x1e6
[<ffffffff80347993>] cfq_dispatch_insert+0x4c/0x70
[<ffffffff80348618>] cfq_dispatch_requests+0x2fb/0x40a
[<ffffffff8033b344>] elv_next_request+0x193/0x1a8
[<ffffffff8034c1e1>] ? kobject_get+0x1a/0x22
[<ffffffffa005fe52>] scsi_request_fn+0x7a/0x496 [scsi_mod]
[<ffffffff8033d720>] blk_start_queueing+0x1a/0x23
[<ffffffff80347d2a>] cfq_insert_request+0x244/0x385
[<ffffffff8033b46a>] elv_insert+0x111/0x1bd
[<ffffffff8033b5ac>] __elv_add_request+0x96/0x9e
[<ffffffff8033e306>] __make_request+0x3c2/0x400
[<ffffffff802efb2e>] ? proc_pid_instantiate+0x87/0x9c
[<ffffffff8033ca9b>] generic_make_request+0x27f/0x319
[<ffffffff802cd91e>] ? bio_init+0x18/0x32
[<ffffffff8033df3b>] submit_bio+0xb4/0xbd
[<ffffffff802c9970>] submit_bh+0xe5/0x109
[<ffffffff802cc659>] block_read_full_page+0x261/0x27f
[<ffffffff802cfd5e>] ? blkdev_get_block+0x0/0x4e
[<ffffffff802cefc3>] blkdev_readpage+0x13/0x15
[<ffffffff80285329>] __do_page_cache_readahead+0x134/0x16c
[<ffffffff8028552d>] ondemand_readahead+0x143/0x155
[<ffffffff802855d7>] page_cache_sync_readahead+0x17/0x19
[<ffffffff8027ecac>] generic_file_aio_read+0x245/0x594
[<ffffffff802abff4>] do_sync_read+0xe2/0x126
[<ffffffff8028f553>] ? __do_fault+0x362/0x3ac
[<ffffffff8024c30c>] ? autoremove_wake_function+0x0/0x38
[<ffffffff8029104c>] ? handle_mm_fault+0x1d9/0x6d1
[<ffffffff8031a144>] ? security_file_permission+0x11/0x13
[<ffffffff802ac748>] vfs_read+0xab/0x134
[<ffffffff802acae9>] sys_read+0x47/0x70
[<ffffffff8020ba2b>] system_call_fastpath+0x16/0x1b
Code: eb 0a 48 89 4a 08 eb 04 49 89 0c 24 41 ff c8 0f 85 88 01 00 00 e9 4b 01 00 00 48 8b 7b 10 48 39 cf 0f 85 a1 00 00 00 48 8b 7b 08 <48> 8b 07 a8 01 75 1a 48 83 c8 01 4c 89 e6 48 89 07 48 83 23 fe
RIP [<ffffffff8034ee05>] rb_erase+0x130/0x2a7
RSP <ffff88012cc7f7f8>
CR2: 0000000000000000
---[ end trace 98881b739b7ec57a ]---

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/