This patchset include two patchsets as follows:
1. Fix hungtask when nbd_config_put
https://patchwork.kernel.org/project/linux-block/list/?series=573381
2. nbd: fix sanity check for first_minor
https://lore.kernel.org/linux-block/[email protected]/
I have consulted with Yu Kuai, and his modification has also been confirmed by him.
Ye Bin (2):
nbd: Fix incorrect error handle when first_minor is illegal in
nbd_dev_add
nbd: Fix hungtask when nbd_config_put
Yu Kuai (2):
nbd: fix max value for 'first_minor'
nbd: fix possible overflow for 'first_minor' in nbd_dev_add()
drivers/block/nbd.c | 42 +++++++++++++++++++-----------------------
1 file changed, 19 insertions(+), 23 deletions(-)
--
2.31.1
From: Yu Kuai <[email protected]>
If 'part_shift' is not zero, then 'index << part_shift' might
overflow to a value that is not greater than '0xfffff', then sysfs
might complains about duplicate creation.
Fixes: b0d9111a2d53 ("nbd: use an idr to keep track of nbd devices")
Signed-off-by: Yu Kuai <[email protected]>
---
drivers/block/nbd.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/drivers/block/nbd.c b/drivers/block/nbd.c
index 678ee2688f63..cc32b5bc0f49 100644
--- a/drivers/block/nbd.c
+++ b/drivers/block/nbd.c
@@ -1803,11 +1803,11 @@ static struct nbd_device *nbd_dev_add(int index, unsigned int refs)
disk->major = NBD_MAJOR;
/* Too big first_minor can cause duplicate creation of
- * sysfs files/links, since MKDEV() expect that the max bits of
- * first_minor is 20.
+ * sysfs files/links, since index << part_shift might overflow, or
+ * MKDEV() expect that the max bits of first_minor is 20.
*/
disk->first_minor = index << part_shift;
- if (disk->first_minor > MINORMASK) {
+ if (disk->first_minor < index || disk->first_minor > MINORMASK) {
err = -EINVAL;
goto out_free_idr;
}
--
2.31.1
If first_minor is illegal will goto out_free_idr label, this will miss
cleanup disk.
Fixes: b1a811633f73 ("block: nbd: add sanity check for first_minor")
Signed-off-by: Ye Bin <[email protected]>
---
drivers/block/nbd.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/block/nbd.c b/drivers/block/nbd.c
index cc32b5bc0f49..88dc0c49631c 100644
--- a/drivers/block/nbd.c
+++ b/drivers/block/nbd.c
@@ -1809,7 +1809,7 @@ static struct nbd_device *nbd_dev_add(int index, unsigned int refs)
disk->first_minor = index << part_shift;
if (disk->first_minor < index || disk->first_minor > MINORMASK) {
err = -EINVAL;
- goto out_free_idr;
+ goto out_err_disk;
}
disk->minors = 1 << part_shift;
--
2.31.1
I got follow issue:
[ 247.381177] INFO: task kworker/u10:0:47 blocked for more than 120 seconds.
[ 247.382644] Not tainted 4.19.90-dirty #140
[ 247.383502] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 247.385027] Call Trace:
[ 247.388384] schedule+0xb8/0x3c0
[ 247.388966] schedule_timeout+0x2b4/0x380
[ 247.392815] wait_for_completion+0x367/0x510
[ 247.397713] flush_workqueue+0x32b/0x1340
[ 247.402700] drain_workqueue+0xda/0x3c0
[ 247.403442] destroy_workqueue+0x7b/0x690
[ 247.405014] nbd_config_put.cold+0x2f9/0x5b6
[ 247.405823] recv_work+0x1fd/0x2b0
[ 247.406485] process_one_work+0x70b/0x1610
[ 247.407262] worker_thread+0x5a9/0x1060
[ 247.408699] kthread+0x35e/0x430
[ 247.410918] ret_from_fork+0x1f/0x30
We can reproduce issue as follows:
1. Inject memory fault in nbd_start_device
-1244,10 +1248,18 @@ static int nbd_start_device(struct nbd_device *nbd)
nbd_dev_dbg_init(nbd);
for (i = 0; i < num_connections; i++) {
struct recv_thread_args *args;
-
- args = kzalloc(sizeof(*args), GFP_KERNEL);
+
+ if (i == 1) {
+ args = NULL;
+ printk("%s: inject malloc error\n", __func__);
+ }
+ else
+ args = kzalloc(sizeof(*args), GFP_KERNEL);
2. Inject delay in recv_work
-757,6 +760,8 @@ static void recv_work(struct work_struct *work)
blk_mq_complete_request(blk_mq_rq_from_pdu(cmd));
}
+ printk("%s: comm=%s pid=%d\n", __func__, current->comm, current->pid);
+ mdelay(5 * 1000);
nbd_config_put(nbd);
atomic_dec(&config->recv_threads);
wake_up(&config->recv_wq);
3. Create nbd server
nbd-server 8000 /tmp/disk
4. Create nbd client
nbd-client localhost 8000 /dev/nbd1
Then will trigger above issue.
Reason is when add delay in recv_work, lead to release the last reference
of 'nbd->config_refs'. nbd_config_put will call flush_workqueue to make
all work finish. Obviously, it will lead to deadloop.
To solve this issue, according to Josef's suggestion move 'recv_work'
init from start device to nbd_dev_add, then destroy 'recv_work'when
nbd device teardown.
Signed-off-by: Ye Bin <[email protected]>
---
drivers/block/nbd.c | 36 ++++++++++++++++--------------------
1 file changed, 16 insertions(+), 20 deletions(-)
diff --git a/drivers/block/nbd.c b/drivers/block/nbd.c
index 88dc0c49631c..ef31f81d7b31 100644
--- a/drivers/block/nbd.c
+++ b/drivers/block/nbd.c
@@ -260,7 +260,7 @@ static void nbd_dev_remove(struct nbd_device *nbd)
mutex_lock(&nbd_index_mutex);
idr_remove(&nbd_index_idr, nbd->index);
mutex_unlock(&nbd_index_mutex);
-
+ destroy_workqueue(nbd->recv_workq);
kfree(nbd);
}
@@ -1314,10 +1314,6 @@ static void nbd_config_put(struct nbd_device *nbd)
kfree(nbd->config);
nbd->config = NULL;
- if (nbd->recv_workq)
- destroy_workqueue(nbd->recv_workq);
- nbd->recv_workq = NULL;
-
nbd->tag_set.timeout = 0;
nbd->disk->queue->limits.discard_granularity = 0;
nbd->disk->queue->limits.discard_alignment = 0;
@@ -1346,14 +1342,6 @@ static int nbd_start_device(struct nbd_device *nbd)
return -EINVAL;
}
- nbd->recv_workq = alloc_workqueue("knbd%d-recv",
- WQ_MEM_RECLAIM | WQ_HIGHPRI |
- WQ_UNBOUND, 0, nbd->index);
- if (!nbd->recv_workq) {
- dev_err(disk_to_dev(nbd->disk), "Could not allocate knbd recv work queue.\n");
- return -ENOMEM;
- }
-
blk_mq_update_nr_hw_queues(&nbd->tag_set, config->num_connections);
nbd->pid = task_pid_nr(current);
@@ -1779,6 +1767,15 @@ static struct nbd_device *nbd_dev_add(int index, unsigned int refs)
}
nbd->disk = disk;
+ nbd->recv_workq = alloc_workqueue("nbd%d-recv",
+ WQ_MEM_RECLAIM | WQ_HIGHPRI |
+ WQ_UNBOUND, 0, nbd->index);
+ if (!nbd->recv_workq) {
+ dev_err(disk_to_dev(nbd->disk), "Could not allocate knbd recv work queue.\n");
+ err = -ENOMEM;
+ goto out_err_disk;
+ }
+
/*
* Tell the block layer that we are not a rotational device
*/
@@ -1809,7 +1806,7 @@ static struct nbd_device *nbd_dev_add(int index, unsigned int refs)
disk->first_minor = index << part_shift;
if (disk->first_minor < index || disk->first_minor > MINORMASK) {
err = -EINVAL;
- goto out_err_disk;
+ goto out_free_work;
}
disk->minors = 1 << part_shift;
@@ -1818,7 +1815,7 @@ static struct nbd_device *nbd_dev_add(int index, unsigned int refs)
sprintf(disk->disk_name, "nbd%d", index);
err = add_disk(disk);
if (err)
- goto out_err_disk;
+ goto out_free_work;
/*
* Now publish the device.
@@ -1827,6 +1824,8 @@ static struct nbd_device *nbd_dev_add(int index, unsigned int refs)
nbd_total_devices++;
return nbd;
+out_free_work:
+ destroy_workqueue(nbd->recv_workq);
out_err_disk:
blk_cleanup_disk(disk);
out_free_idr:
@@ -2082,13 +2081,10 @@ static void nbd_disconnect_and_put(struct nbd_device *nbd)
nbd_disconnect(nbd);
sock_shutdown(nbd);
/*
- * Make sure recv thread has finished, so it does not drop the last
- * config ref and try to destroy the workqueue from inside the work
- * queue. And this also ensure that we can safely call nbd_clear_que()
+ * Make sure recv thread has finished, we can safely call nbd_clear_que()
* to cancel the inflight I/Os.
*/
- if (nbd->recv_workq)
- flush_workqueue(nbd->recv_workq);
+ flush_workqueue(nbd->recv_workq);
nbd_clear_que(nbd);
nbd->task_setup = NULL;
mutex_unlock(&nbd->config_lock);
--
2.31.1
From: Yu Kuai <[email protected]>
commit b1a811633f73 ("block: nbd: add sanity check for first_minor")
checks that 'first_minor' should not be greater than 0xff, which is
wrong. Whitout the commit, the details that when user pass 0x100000,
it ends up create sysfs dir "/sys/block/43:0" are as follows:
nbd_dev_add
disk->first_minor = index << part_shift
-> default part_shift is 5, first_minor is 0x2000000
device_add_disk
ddev->devt = MKDEV(disk->major, disk->first_minor)
-> (0x2b << 20) | (0x2000000) = 0x2b00000
device_add
device_create_sys_dev_entry
format_dev_t
sprintf(buffer, "%u:%u", MAJOR(dev), MINOR(dev));
-> got 43:0
sysfs_create_link -> /sys/block/43:0
By the way, with the wrong fix, when part_shift is the default value,
only 8 ndb devices can be created since 8 << 5 is greater than 0xff.
Since the max bits for 'first_minor' should be the same as what
MKDEV() does, which is 20. Change the upper bound of 'first_minor'
from 0xff to 0xfffff.
Fixes: b1a811633f73 ("block: nbd: add sanity check for first_minor")
Signed-off-by: Yu Kuai <[email protected]>
---
drivers/block/nbd.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/drivers/block/nbd.c b/drivers/block/nbd.c
index b47b2a87ae8f..678ee2688f63 100644
--- a/drivers/block/nbd.c
+++ b/drivers/block/nbd.c
@@ -1803,11 +1803,11 @@ static struct nbd_device *nbd_dev_add(int index, unsigned int refs)
disk->major = NBD_MAJOR;
/* Too big first_minor can cause duplicate creation of
- * sysfs files/links, since first_minor will be truncated to
- * byte in __device_add_disk().
+ * sysfs files/links, since MKDEV() expect that the max bits of
+ * first_minor is 20.
*/
disk->first_minor = index << part_shift;
- if (disk->first_minor > 0xff) {
+ if (disk->first_minor > MINORMASK) {
err = -EINVAL;
goto out_free_idr;
}
--
2.31.1
On Tue, Nov 02, 2021 at 09:52:33AM +0800, Ye Bin wrote:
> This patchset include two patchsets as follows:
> 1. Fix hungtask when nbd_config_put
> https://patchwork.kernel.org/project/linux-block/list/?series=573381
> 2. nbd: fix sanity check for first_minor
> https://lore.kernel.org/linux-block/[email protected]/
>
> I have consulted with Yu Kuai, and his modification has also been confirmed by him.
>
> Ye Bin (2):
> nbd: Fix incorrect error handle when first_minor is illegal in
> nbd_dev_add
> nbd: Fix hungtask when nbd_config_put
>
> Yu Kuai (2):
> nbd: fix max value for 'first_minor'
> nbd: fix possible overflow for 'first_minor' in nbd_dev_add()
>
> drivers/block/nbd.c | 42 +++++++++++++++++++-----------------------
> 1 file changed, 19 insertions(+), 23 deletions(-)
>
Perfect, thanks guys, you can add
Reviewed-by: Josef Bacik <[email protected]>
to the series,
Josef
On Tue, 2 Nov 2021 09:52:33 +0800, Ye Bin wrote:
> This patchset include two patchsets as follows:
> 1. Fix hungtask when nbd_config_put
> https://patchwork.kernel.org/project/linux-block/list/?series=573381
> 2. nbd: fix sanity check for first_minor
> https://lore.kernel.org/linux-block/[email protected]/
>
> I have consulted with Yu Kuai, and his modification has also been confirmed by him.
>
> [...]
Applied, thanks!
[1/4] nbd: fix max value for 'first_minor'
commit: e4c4871a73944353ea23e319de27ef73ce546623
[2/4] nbd: fix possible overflow for 'first_minor' in nbd_dev_add()
commit: 940c264984fd1457918393c49674f6b39ee16506
[3/4] nbd: Fix incorrect error handle when first_minor is illegal in nbd_dev_add
commit: 69beb62ff0d1723a750eebe1c4d01da573d7cd19
[4/4] nbd: Fix hungtask when nbd_config_put
commit: e2daec488c57069a4a431d5b752f50294c4bf273
Best regards,
--
Jens Axboe