[PATCH] nvmet: fix data_len to 0 for bdev-backed write_zeroes

[PATCH] nvmet: fix data_len to 0 for bdev-backed write_zeroes

[Minwoo Im]

The WRITE ZEROES command has no data transfer so that we need to
initialize the struct (nvmet_req *req)->data_len to 0x0.  While
(nvmet_req *req)->transfer_len is initialized in nvmet_req_init(),
data_len will be initialized by nowhere which might cause the failure
with status code NVME_SC_SGL_INVALID_DATA | NVME_SC_DNR randomly.  It's
because nvmet_req_execute() checks like:

	if (unlikely(req->data_len != req->transfer_len)) {
		req->error_loc = offsetof(struct nvme_common_command, dptr);
		nvmet_req_complete(req, NVME_SC_SGL_INVALID_DATA | NVME_SC_DNR);
	} else
		req->execute(req);

This patch fixes req->data_len not to be a randomly assigned by
initializing it to 0x0 when preparing the command in
nvmet_bdev_parse_io_cmd().

nvmet_file_parse_io_cmd() which is for file-backed I/O has already
initialized the data_len field to 0x0, though.

Cc: Christoph Hellwig <hch at lst.de>
Cc: Sagi Grimberg <sagi at grimberg.me>
Cc: Chaitanya Kulkarni <Chaitanya.Kulkarni at wdc.com>
Signed-off-by: Minwoo Im <minwoo.im.dev at gmail.com>
---
 drivers/nvme/target/io-cmd-bdev.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/nvme/target/io-cmd-bdev.c b/drivers/nvme/target/io-cmd-bdev.c
index 3efc52f9c309..7a1cf6437a6a 100644
--- a/drivers/nvme/target/io-cmd-bdev.c
+++ b/drivers/nvme/target/io-cmd-bdev.c
@@ -293,6 +293,7 @@ u16 nvmet_bdev_parse_io_cmd(struct nvmet_req *req)
 		return 0;
 	case nvme_cmd_write_zeroes:
 		req->execute = nvmet_bdev_execute_write_zeroes;
+		req->data_len = 0;
 		return 0;
 	default:
 		pr_err("unhandled cmd %d on qid %d\n", cmd->common.opcode,
-- 
2.21.0

[PATCH] nvmet: fix data_len to 0 for bdev-backed write_zeroes

[Sagi Grimberg]

Is this a regression? IIRC write-zeros used to work...

[PATCH] nvmet: fix data_len to 0 for bdev-backed write_zeroes

[Minwoo Im]

On 19-06-02 00:32:43, Sagi Grimberg wrote:
> Is this a regression? IIRC write-zeros used to work...

Hi Sagi,

The following outputs are a looping basic tests with a single command:
(I think you can find the commands will easily be failed.)

root at target:~# nvme list-subsys
nvme-subsys0 - NQN=nqn.2014.08.org.nvmexpress:80861af4foo
QEMU NVMe Ctrl
\
 +- nvme0 pcie 0000:00:04.0 live
 nvme-subsys1 - NQN=test
 \
  +- nvme1 loop  live
root at target:~# nvme write-zeroes /dev/nvme1n1
NVME Write Zeroes Success
root at target:~# nvme write-zeroes /dev/nvme1n1
NVME Write Zeroes Success
root at target:~# nvme write-zeroes /dev/nvme1n1
NVME Write Zeroes Success
root at target:~# nvme write-zeroes /dev/nvme1n1
NVMe status: Unknown(0x600f)
root at target:~# nvme write-zeroes /dev/nvme1n1
NVME Write Zeroes Success
root at target:~#

The write-zeroes command has been requested with SLBA 0 and NLB 0 as a
default options.  This might happen with status code 0x600f which is
NVME_SC_SGL_INVALID_DATA | NVME_SC_DNR which happened due to:
	req->data_len != req->transfer_len

The transfer_len is always 0x0 for write-zeroes, but data-len is never
initialized in this case.

Please correct me if I'm wrong here.

Thanks,

[PATCH] nvmet: fix data_len to 0 for bdev-backed write_zeroes

[Chaitanya Kulkarni]

Thanks for the fix.

Reviewed-by: Chaitanya Kulkarni <chaitanya.kulkarni at wdc.com>

On 6/1/19 8:44 PM, Minwoo Im wrote:
> The WRITE ZEROES command has no data transfer so that we need to
> initialize the struct (nvmet_req *req)->data_len to 0x0.  While
> (nvmet_req *req)->transfer_len is initialized in nvmet_req_init(),
> data_len will be initialized by nowhere which might cause the failure
> with status code NVME_SC_SGL_INVALID_DATA | NVME_SC_DNR randomly.  It's
> because nvmet_req_execute() checks like:
>
> 	if (unlikely(req->data_len != req->transfer_len)) {
> 		req->error_loc = offsetof(struct nvme_common_command, dptr);
> 		nvmet_req_complete(req, NVME_SC_SGL_INVALID_DATA | NVME_SC_DNR);
> 	} else
> 		req->execute(req);
>
> This patch fixes req->data_len not to be a randomly assigned by
> initializing it to 0x0 when preparing the command in
> nvmet_bdev_parse_io_cmd().
>
> nvmet_file_parse_io_cmd() which is for file-backed I/O has already
> initialized the data_len field to 0x0, though.
>
> Cc: Christoph Hellwig <hch at lst.de>
> Cc: Sagi Grimberg <sagi at grimberg.me>
> Cc: Chaitanya Kulkarni <Chaitanya.Kulkarni at wdc.com>
> Signed-off-by: Minwoo Im <minwoo.im.dev at gmail.com>
> ---
>  drivers/nvme/target/io-cmd-bdev.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/drivers/nvme/target/io-cmd-bdev.c b/drivers/nvme/target/io-cmd-bdev.c
> index 3efc52f9c309..7a1cf6437a6a 100644
> --- a/drivers/nvme/target/io-cmd-bdev.c
> +++ b/drivers/nvme/target/io-cmd-bdev.c
> @@ -293,6 +293,7 @@ u16 nvmet_bdev_parse_io_cmd(struct nvmet_req *req)
>  		return 0;
>  	case nvme_cmd_write_zeroes:
>  		req->execute = nvmet_bdev_execute_write_zeroes;
> +		req->data_len = 0;
>  		return 0;
>  	default:
>  		pr_err("unhandled cmd %d on qid %d\n", cmd->common.opcode,

[PATCH] nvmet: fix data_len to 0 for bdev-backed write_zeroes

[Christoph Hellwig]

Looks good,

Reviewed-by: Christoph Hellwig <hch at lst.de>