lightnvm: pblk: remove internal IO timeout
authorIgor Konopko <igor.j.konopko@intel.com>
Sat, 4 May 2019 18:38:04 +0000 (20:38 +0200)
committerJens Axboe <axboe@kernel.dk>
Mon, 6 May 2019 16:19:18 +0000 (10:19 -0600)
Currently during pblk padding, there is internal IO timeout introduced,
which is smaller than default NVMe timeout. This can lead to various
use-after-free issues. Since in case of any IO timeouts NVMe and block
layer will handle timeout by themselves and report it back to use,
there is no need to keep this internal timeout in pblk.

Signed-off-by: Igor Konopko <igor.j.konopko@intel.com>
Signed-off-by: Matias Bjørling <mb@lightnvm.io>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
drivers/lightnvm/pblk-recovery.c
drivers/lightnvm/pblk.h

index 137e963..865fe31 100644 (file)
@@ -290,12 +290,7 @@ next_pad_rq:
 
 fail_complete:
        kref_put(&pad_rq->ref, pblk_recov_complete);
-
-       if (!wait_for_completion_io_timeout(&pad_rq->wait,
-                               msecs_to_jiffies(PBLK_COMMAND_TIMEOUT_MS))) {
-               pblk_err(pblk, "pad write timed out\n");
-               ret = -ETIME;
-       }
+       wait_for_completion(&pad_rq->wait);
 
        if (!pblk_line_is_full(line))
                pblk_err(pblk, "corrupted padded line: %d\n", line->id);
index 381f074..90c703d 100644 (file)
@@ -43,8 +43,6 @@
 
 #define PBLK_CACHE_NAME_LEN (DISK_NAME_LEN + 16)
 
-#define PBLK_COMMAND_TIMEOUT_MS 30000
-
 /* Max 512 LUNs per device */
 #define PBLK_MAX_LUNS_BITMAP (4)