1 --- linux-2.6.33/scripts/mod/modpost.c~ 2010-02-24 19:52:17.000000000 +0100
2 +++ linux-2.6.33/scripts/mod/modpost.c 2010-03-07 14:26:47.242168558 +0100
7 -#include "../../include/generated/autoconf.h"
8 +// PLD architectures don't use CONFIG_SYMBOL_PREFIX
9 +//#include "../../include/generated/autoconf.h"
10 #include "../../include/linux/license.h"
12 /* Some toolchains use a `_' prefix for all user symbols. */
14 --- linux-3.0/scripts/kconfig/lxdialog/check-lxdialog.sh~ 2011-07-22 04:17:23.000000000 +0200
15 +++ linux-3.0/scripts/kconfig/lxdialog/check-lxdialog.sh 2011-08-25 21:26:04.799150642 +0200
17 $cc -print-file-name=lib${lib}.${ext} | grep -q /
20 + for libt in tinfow tinfo ; do
21 + $cc -print-file-name=lib${libt}.${ext} | grep -q /
22 + if [ $? -eq 0 ]; then
29 From e820d55cb99dd93ac2dc949cf486bb187e5cd70d Mon Sep 17 00:00:00 2001
30 From: Guoqing Jiang <gqjiang@suse.com>
31 Date: Wed, 19 Dec 2018 14:19:25 +0800
32 Subject: md: fix raid10 hang issue caused by barrier
34 Content-Type: text/plain; charset=UTF-8
35 Content-Transfer-Encoding: 8bit
37 When both regular IO and resync IO happen at the same time,
38 and if we also need to split regular. Then we can see tasks
42 [ 1463.757205] INFO: task md1_resync:5215 blocked for more than 480 seconds.
43 [ 1463.757207] Not tainted 4.19.5-1-default #1
44 [ 1463.757209] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
45 [ 1463.757212] md1_resync D 0 5215 2 0x80000000
46 [ 1463.757216] Call Trace:
47 [ 1463.757223] ? __schedule+0x29a/0x880
48 [ 1463.757231] ? raise_barrier+0x8d/0x140 [raid10]
49 [ 1463.757236] schedule+0x78/0x110
50 [ 1463.757243] raise_barrier+0x8d/0x140 [raid10]
51 [ 1463.757248] ? wait_woken+0x80/0x80
52 [ 1463.757257] raid10_sync_request+0x1f6/0x1e30 [raid10]
53 [ 1463.757265] ? _raw_spin_unlock_irq+0x22/0x40
54 [ 1463.757284] ? is_mddev_idle+0x125/0x137 [md_mod]
55 [ 1463.757302] md_do_sync.cold.78+0x404/0x969 [md_mod]
56 [ 1463.757311] ? wait_woken+0x80/0x80
57 [ 1463.757336] ? md_rdev_init+0xb0/0xb0 [md_mod]
58 [ 1463.757351] md_thread+0xe9/0x140 [md_mod]
59 [ 1463.757358] ? _raw_spin_unlock_irqrestore+0x2e/0x60
60 [ 1463.757364] ? __kthread_parkme+0x4c/0x70
61 [ 1463.757369] kthread+0x112/0x130
62 [ 1463.757374] ? kthread_create_worker_on_cpu+0x40/0x40
63 [ 1463.757380] ret_from_fork+0x3a/0x50
66 [ 1463.760679] INFO: task kworker/0:8:5367 blocked for more than 480 seconds.
67 [ 1463.760683] Not tainted 4.19.5-1-default #1
68 [ 1463.760684] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
69 [ 1463.760687] kworker/0:8 D 0 5367 2 0x80000000
70 [ 1463.760718] Workqueue: md submit_flushes [md_mod]
71 [ 1463.760721] Call Trace:
72 [ 1463.760731] ? __schedule+0x29a/0x880
73 [ 1463.760741] ? wait_barrier+0xdd/0x170 [raid10]
74 [ 1463.760746] schedule+0x78/0x110
75 [ 1463.760753] wait_barrier+0xdd/0x170 [raid10]
76 [ 1463.760761] ? wait_woken+0x80/0x80
77 [ 1463.760768] raid10_write_request+0xf2/0x900 [raid10]
78 [ 1463.760774] ? wait_woken+0x80/0x80
79 [ 1463.760778] ? mempool_alloc+0x55/0x160
80 [ 1463.760795] ? md_write_start+0xa9/0x270 [md_mod]
81 [ 1463.760801] ? try_to_wake_up+0x44/0x470
82 [ 1463.760810] raid10_make_request+0xc1/0x120 [raid10]
83 [ 1463.760816] ? wait_woken+0x80/0x80
84 [ 1463.760831] md_handle_request+0x121/0x190 [md_mod]
85 [ 1463.760851] md_make_request+0x78/0x190 [md_mod]
86 [ 1463.760860] generic_make_request+0x1c6/0x470
87 [ 1463.760870] raid10_write_request+0x77a/0x900 [raid10]
88 [ 1463.760875] ? wait_woken+0x80/0x80
89 [ 1463.760879] ? mempool_alloc+0x55/0x160
90 [ 1463.760895] ? md_write_start+0xa9/0x270 [md_mod]
91 [ 1463.760904] raid10_make_request+0xc1/0x120 [raid10]
92 [ 1463.760910] ? wait_woken+0x80/0x80
93 [ 1463.760926] md_handle_request+0x121/0x190 [md_mod]
94 [ 1463.760931] ? _raw_spin_unlock_irq+0x22/0x40
95 [ 1463.760936] ? finish_task_switch+0x74/0x260
96 [ 1463.760954] submit_flushes+0x21/0x40 [md_mod]
98 So resync io is waiting for regular write io to complete to
99 decrease nr_pending (conf->barrier++ is called before waiting).
100 The regular write io splits another bio after call wait_barrier
101 which call nr_pending++, then the splitted bio would continue
102 with raid10_write_request -> wait_barrier, so the splitted bio
103 has to wait for barrier to be zero, then deadlock happens as
113 To resolve the issue, we need to call allow_barrier to decrease
114 nr_pending before generic_make_request since regular IO is not
115 issued to underlying devices, and wait_barrier is called again
116 to ensure no internal IO happening.
118 Fixes: fc9977dd069e ("md/raid10: simplify the splitting of requests.")
119 Reported-and-tested-by: SiniĊĦa Bandin <sinisa@4net.rs>
120 Signed-off-by: Guoqing Jiang <gqjiang@suse.com>
121 Signed-off-by: Shaohua Li <shli@fb.com>
123 drivers/md/raid10.c | 4 ++++
124 1 file changed, 4 insertions(+)
126 diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
127 index 76c92e31afc0..abb5d382f64d 100644
128 --- a/drivers/md/raid10.c
129 +++ b/drivers/md/raid10.c
130 @@ -1209,7 +1209,9 @@ static void raid10_read_request(struct mddev *mddev, struct bio *bio,
131 struct bio *split = bio_split(bio, max_sectors,
132 gfp, &conf->bio_split);
133 bio_chain(split, bio);
134 + allow_barrier(conf);
135 generic_make_request(bio);
136 + wait_barrier(conf);
138 r10_bio->master_bio = bio;
139 r10_bio->sectors = max_sectors;
140 @@ -1492,7 +1494,9 @@ retry_write:
141 struct bio *split = bio_split(bio, r10_bio->sectors,
142 GFP_NOIO, &conf->bio_split);
143 bio_chain(split, bio);
144 + allow_barrier(conf);
145 generic_make_request(bio);
146 + wait_barrier(conf);
148 r10_bio->master_bio = bio;