From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752401AbaGBFcN (ORCPT ); Wed, 2 Jul 2014 01:32:13 -0400 Received: from userp1040.oracle.com ([156.151.31.81]:33649 "EHLO userp1040.oracle.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750888AbaGBFcL (ORCPT ); Wed, 2 Jul 2014 01:32:11 -0400 From: Junxiao Bi To: linux-kernel@vger.kernel.org Cc: axboe@kernel.dk, joe.jin@oracle.com Subject: [PATCH] block: fix reqeust->__data_len overflow Date: Wed, 2 Jul 2014 13:32:30 +0800 Message-Id: <1404279150-9372-1-git-send-email-junxiao.bi@oracle.com> X-Mailer: git-send-email 1.7.9.5 X-Source-IP: ucsinet21.oracle.com [156.151.31.93] Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org blk_rq_sectors(req) + bio_sectors(bio) > blk_rq_get_max_sectors(req) is used to check whether a bio can be merged into an exist request. If can, req->__data_len += bio->bio_size. Since req->__data_len is a 32bit uint, if blk_rq_get_max_sectors(req) > (UINT_MAX >> 9), req->__date_len may overflow when merging a new bio. This probably happen for discard request. In xen blkfront driver, its max_discard_sectors is set to the whole disk sector size, see xlvbd_init_blk_queue(). So issuing discrad requests to a xen virtual disk with a size over 4G is very possible to trigger the overflow. This overflow will cause kernel panic in blk_end_request_all() due to BUG() triggered. The following is a call trace we saw in 3.0.69. Upstream kernel also suffer this issue. @ __end_that: dev xvdg: type=1, flags=2224441 @ sector 0, nr/cnr 8378368/4294959104 @ bio ffff8803d8cf3080, biotail ffff8803d8cf32c0, buffer (null), @ len 4289724416 @ blk_update_request: bio idx 0 >= vcnt 0 @ request botched: dev xvdg: type=1, flags=2224441 @ sector 0, nr/cnr 8378368/4294959104 @ bio ffff8803d8cf3080, biotail ffff8803d8cf32c0, buffer (null), @ len 4289724416 @ ------------[ cut here ]------------ @ kernel BUG at block/blk-core.c:2394! @ invalid opcode: 0000 [#1] SMP @ CPU 0 @ Modules linked in: nfs fscache auth_rpcgss nfs_acl autofs4 i2c_dev i2c_core @ lockd sunrpc(U) ksplice_x773z34q_vmlinux_new(U) ksplice_x773z34q(U) @ ksplice_bj7y22gc_vmlinux_new(U) ksplice_bj7y22gc_ipv6_new(U) @ ksplice_bj7y22gc(U) @ . @ ksplice_i1o46065(U) ksplice_5gqtkuvt_vmlinux_new(U) ksplice_5gqtkuvt(U) @ ksplice_2bcv8td6(U) ksplice_v5bs54bz_vmlinux_new(U) ksplice_v5bs54bz(U) @ ksplice_l7s0dhx6(U) ksplice_aur7sgvi(U) ksplice_ckie4cpv(U) @ nf_conntrack_netbios_ns @ . @ nf_conntrack_broadcast ipt_REJECT nf_conntrack_ipv4 nf_defrag_ipv4 xt_state @ nf_conntrack xt_comment iptable_filter ip_tables be2iscsi iscsi_boot_sysfs @ ib_iser rdma_cm ib_cm iw_cm ib_sa ib_mad ib_core ib_addr iscsi_tcp bnx2i cnic @ uio ipv6 @ . @ cxgb3i libcxgbi cxgb3 mdio libiscsi_tcp libiscsi scsi_transport_iscsi @ parport_pc lp parport snd_seq_dummy snd_seq_oss snd_seq_midi_event snd_seq @ snd_seq_device snd_pcm_oss snd_mixer_oss snd_pcm snd_timer snd soundcore @ snd_page_alloc @ . @ pcspkr xen_netfront dm_snapshot dm_zero dm_mirror dm_region_hash dm_log @ dm_mod xen_blkfront ext3 jbd mbcache sd_mod crc_t10dif [last unloaded: @ ksplice_x773z34q_vmlinux_old] @ . @ Pid: 0, comm: swapper Not tainted 2.6.39-400.212.1.el5uek #1 @ RIP: e030:[] [] @ __blk_end_request_all+0x2a/0x40 @ RSP: e02b:ffff8803ffc03df8 EFLAGS: 00010002 @ RAX: 0000000000000001 RBX: ffff8803db3c8000 RCX: ffff8803d8cf32c0 @ RDX: 0000000000000001 RSI: ffff8803d8cf3080 RDI: ffff8803daed08d8 @ RBP: ffff8803ffc03df8 R08: 0000000000000000 R09: ffff8803daed08d8 @ R10: 0000000000000000 R11: 000000000000000a R12: 0000000000000000 @ R13: ffff8803dad5e3c0 R14: 0000000000000001 R15: 0000000000000029 @ FS: 00007f1f34a32940(0000) GS:ffff8803ffc00000(0000) knlGS:0000000000000000 @ CS: e033 DS: 0000 ES: 0000 CR0: 000000008005003b @ CR2: 00000000020c6148 CR3: 00000003c6492000 CR4: 0000000000002660 @ DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 @ DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 @ Process swapper (pid: 0, threadinfo ffffffff81794000, task ffffffff8179f020) @ Stack: @ ffff8803ffc03e48 ffffffffa005c56a ffff8803da57a8d0 00000028810d99ee @ 0000000000000000 ffff8803db1ea7c0 ffff8803db1beec0 000000000000005e @ 0000000000000001 0000000000000000 ffff8803ffc03e98 ffffffff810d735d @ Call Trace: @ @ [] blkif_interrupt+0x20a/0x3a0 [xen_blkfront] @ [] handle_irq_event_percpu+0x5d/0x1a0 @ [] handle_irq_event+0x4f/0x80 @ [] handle_edge_irq+0xa5/0x100 @ [] __xen_evtchn_do_upcall+0x218/0x310 @ [] xen_evtchn_do_upcall+0x2f/0x50 @ [] xen_do_hypervisor_callback+0x1e/0x30 @ @ [] ? xen_hypercall_sched_op+0xa/0x20 @ [] ? xen_hypercall_sched_op+0xa/0x20 @ [] ? xen_safe_halt+0x10/0x20 @ [] ? default_idle+0x5b/0x170 @ [] ? cpu_idle+0xc6/0xf0 @ [] ? rest_init+0x72/0x80 @ [] ? start_kernel+0x2aa/0x390 @ [] ? x86_64_start_reservations+0x6a/0xa0 @ [] ? xen_start_kernel+0x315/0x440 @ Code: 00 55 48 89 e5 0f 1f 44 00 00 48 8b 87 60 01 00 00 31 c9 48 85 c0 75 0e @ 8b 57 54 e8 91 ff ff ff 84 c0 75 07 c9 c3 8b 48 54 eb ed <0f> 0b 0f 1f 40 00 @ eb fa 0f 1f 80 00 00 00 00 0f 1f 80 00 00 00 @ RIP [] __blk_end_request_all+0x2a/0x40 @ RSP @ ---[ end trace b09ff97496363201 ]--- Signed-off-by: Junxiao Bi --- block/blk-merge.c | 29 +++++++++++++++++++++++------ 1 file changed, 23 insertions(+), 6 deletions(-) diff --git a/block/blk-merge.c b/block/blk-merge.c index b3bf0df..ae4f4c8 100644 --- a/block/blk-merge.c +++ b/block/blk-merge.c @@ -325,11 +325,30 @@ no_merge: return 0; } -int ll_back_merge_fn(struct request_queue *q, struct request *req, +static inline bool ll_allow_merge_bio(struct request *req, struct bio *bio) { if (blk_rq_sectors(req) + bio_sectors(bio) > - blk_rq_get_max_sectors(req)) { + min(blk_rq_get_max_sectors(req), UINT_MAX >> 9)) + return false; + + return true; +} + +static inline bool ll_allow_merge_req(struct request *req, + struct request *next) +{ + if (blk_rq_sectors(req) + blk_rq_sectors(next) > + min(blk_rq_get_max_sectors(req), UINT_MAX >> 9)) + return false; + + return true; +} + +int ll_back_merge_fn(struct request_queue *q, struct request *req, + struct bio *bio) +{ + if (!ll_allow_merge_bio(req, bio)) { req->cmd_flags |= REQ_NOMERGE; if (req == q->last_merge) q->last_merge = NULL; @@ -346,8 +365,7 @@ int ll_back_merge_fn(struct request_queue *q, struct request *req, int ll_front_merge_fn(struct request_queue *q, struct request *req, struct bio *bio) { - if (blk_rq_sectors(req) + bio_sectors(bio) > - blk_rq_get_max_sectors(req)) { + if (!ll_allow_merge_bio(req, bio)) { req->cmd_flags |= REQ_NOMERGE; if (req == q->last_merge) q->last_merge = NULL; @@ -389,8 +407,7 @@ static int ll_merge_requests_fn(struct request_queue *q, struct request *req, /* * Will it become too large? */ - if ((blk_rq_sectors(req) + blk_rq_sectors(next)) > - blk_rq_get_max_sectors(req)) + if (!ll_allow_merge_req(req, next)) return 0; total_phys_segments = req->nr_phys_segments + next->nr_phys_segments; -- 1.7.9.5