Please find attached a pcap file showing TGTD with this patch applied
correctly responding to the different over/under-flow conditions.
On Sat, Mar 23, 2013 at 2:16 PM, Ronnie Sahlberg
<ronniesahlberg@xxxxxxxxx> wrote:
> This patch adds support to return overflow/underflow residuals
> for the cases where the transport layer (iSCSI, ...) specifies
> an EDTL that is not matching the SCSI transdfer length.
>
> Additionally, in case of residual underflow it will also clamp the size of
> returned DATA-IN to the amount of data that was actually read and not the
> EDTL that was requested.
>
> This is from RFC 3720, 10.4.1 :
> bit 6 - (U) set for Residual Underflow. In this case, the Residual
> Count indicates the number of bytes that were not transferred out
> of the number of bytes that were expected to be transferred. For
> a bidirectional operation, the Residual Count contains the
> residual for the write operation.
>
> I have added a specific test for this condition to the testsute in
> libiscsi :
>
> ./bin/iscsi-test-cu -t "TestRead10.testRead10Residuals" \
> iscsi://127.0.0.1/iqn.ronnie.test/1 -V
>
> Suite: TestRead10
> Test: testRead10Residuals ...
> Test invalid READ10 commands
> Block size is 512
> Try reading one block but with iSCSI expected transfer length==0
> Verify that the target returned SUCCESS
> Verify residual overflow flag is set
> Verify we got 512 bytes of residual overflow
> Try reading one block but with iSCSI expected transfer length==10000
> Verify that the target returned SUCCESS
> Verify we got one whole block back from the target
> Verify residual underflow flag is set
> Verify we got 9488 bytes of residual underflow
> Try reading one block but with iSCSI expected transfer length==200
> Verify that the target returned SUCCESS
> Verify we got 200 bytes back from the target
> Verify residual overflow flag is set
> Verify we got 312 bytes of residual overflow
> Try reading two blocks but iSCSI expected transfer length==512 (==one block)
> Verify that the target returned SUCCESS
> Verify we got one whole block back from the target
> Verify residual overflow flag is set
> Verify we got one block of residual overflow
> passed
>
> --Run Summary: Type Total Ran Passed Failed
> suites 1 1 n/a 0
> tests 1 1 1 0
> asserts 24 24 24 0
>
> Signed-off-by: Ronnie Sahlberg <ronniesahlberg@xxxxxxxxx>
> ---
> usr/sbc.c | 30 ++++++++++++++++++++++++++++--
> 1 files changed, 28 insertions(+), 2 deletions(-)
>
> diff --git a/usr/sbc.c b/usr/sbc.c
> index 48dab04..26cceb1 100644
> --- a/usr/sbc.c
> +++ b/usr/sbc.c
> @@ -226,6 +226,8 @@ static int sbc_rw(int host_no, struct scsi_cmd *cmd)
> int ret;
> uint64_t lba;
> uint32_t tl;
> + long residual = 0;
> + long length;
> unsigned char key = ILLEGAL_REQUEST;
> uint16_t asc = ASC_LUN_NOT_SUPPORTED;
> struct scsi_lu *lu = cmd->dev;
> @@ -335,12 +337,36 @@ static int sbc_rw(int host_no, struct scsi_cmd *cmd)
> cmd->offset = lba << cmd->dev->blk_shift;
> cmd->tl = tl << cmd->dev->blk_shift;
>
> + /* Handle residuals */
> + switch (cmd->scb[0]) {
> + case READ_6:
> + case READ_10:
> + case READ_12:
> + case READ_16:
> + length = scsi_get_in_length(cmd);
> + if (length != cmd->tl) {
> + residual = length - cmd->tl;
> +
> + if (length > cmd->tl) {
> + length = cmd->tl;
> + scsi_set_in_transfer_len(cmd, cmd->tl);
> + scsi_set_in_length(cmd, cmd->tl);
> + }
> + }
> + break;
> + }
> +
> ret = cmd->dev->bst->bs_cmd_submit(cmd);
> if (ret) {
> key = HARDWARE_ERROR;
> asc = ASC_INTERNAL_TGT_FAILURE;
> - } else
> - return SAM_STAT_GOOD;
> + goto sense;
> + }
> +
> + if (residual)
> + scsi_set_in_resid(cmd, residual);
> +
> + return SAM_STAT_GOOD;
>
> sense:
> cmd->offset = 0;
> --
> 1.7.3.1
>
Attachment:
read10-residuals-test.pcapng
Description: Binary data
