rq->raw_data_len introduced for block layer padding and draining (commit 6b00769fe1502b4ad97bb327ef7ac971b208bfb5) broke residual byte count handling. Block drivers modify rq->data_len to notify residual byte count to the block layer which blindly reported unmodified rq->raw_data_len to userland. To keep block drivers dealing only with rq->data_len, this should be handled inside block layer. However, how much extra buffer was appened is lost after rq->data_len is modified. This patch replaces rq->raw_data_len with rq->extra_len and add blk_rq_raw_data_len() helper to calculate raw data size from rq->data_len and rq->extra_len. The helper returns correct raw residual byte count when called on a rq whose data_len is modified to carry residual byte count. This problem was reported and diagnosed by Mike Galbraith. Signed-off-by: Tejun Heo <htejun@xxxxxxxxx> Cc: Mike Galbraith <efault@xxxxxx> --- Comments updated compared to the previous version. block/blk-core.c | 3 +-- block/blk-map.c | 2 +- block/blk-merge.c | 1 + block/blk-settings.c | 4 ++++ block/bsg.c | 8 ++++---- block/scsi_ioctl.c | 4 ++-- drivers/ata/libata-scsi.c | 3 ++- include/linux/blkdev.h | 8 +++++++- 8 files changed, 22 insertions(+), 11 deletions(-) Index: work/block/blk-core.c =================================================================== --- work.orig/block/blk-core.c +++ work/block/blk-core.c @@ -127,7 +127,7 @@ void rq_init(struct request_queue *q, st rq->nr_hw_segments = 0; rq->ioprio = 0; rq->special = NULL; - rq->raw_data_len = 0; + rq->extra_len = 0; rq->buffer = NULL; rq->tag = -1; rq->errors = 0; @@ -2016,7 +2016,6 @@ void blk_rq_bio_prep(struct request_queu rq->hard_cur_sectors = rq->current_nr_sectors; rq->hard_nr_sectors = rq->nr_sectors = bio_sectors(bio); rq->buffer = bio_data(bio); - rq->raw_data_len = bio->bi_size; rq->data_len = bio->bi_size; rq->bio = rq->biotail = bio; Index: work/block/blk-map.c =================================================================== --- work.orig/block/blk-map.c +++ work/block/blk-map.c @@ -19,7 +19,6 @@ int blk_rq_append_bio(struct request_que rq->biotail->bi_next = bio; rq->biotail = bio; - rq->raw_data_len += bio->bi_size; rq->data_len += bio->bi_size; } return 0; @@ -156,6 +155,7 @@ int blk_rq_map_user(struct request_queue bio->bi_io_vec[bio->bi_vcnt - 1].bv_len += pad_len; bio->bi_size += pad_len; rq->data_len += pad_len; + rq->extra_len += pad_len; } rq->buffer = rq->data = NULL; Index: work/block/blk-merge.c =================================================================== --- work.orig/block/blk-merge.c +++ work/block/blk-merge.c @@ -232,6 +232,7 @@ new_segment: (PAGE_SIZE - 1)); nsegs++; rq->data_len += q->dma_drain_size; + rq->extra_len += q->dma_drain_size; } if (sg) Index: work/block/bsg.c =================================================================== --- work.orig/block/bsg.c +++ work/block/bsg.c @@ -437,14 +437,14 @@ static int blk_complete_sgv4_hdr_rq(stru } if (rq->next_rq) { - hdr->dout_resid = rq->raw_data_len; - hdr->din_resid = rq->next_rq->raw_data_len; + hdr->dout_resid = blk_rq_raw_data_len(rq); + hdr->din_resid = blk_rq_raw_data_len(rq->next_rq); blk_rq_unmap_user(bidi_bio); blk_put_request(rq->next_rq); } else if (rq_data_dir(rq) == READ) - hdr->din_resid = rq->raw_data_len; + hdr->din_resid = blk_rq_raw_data_len(rq); else - hdr->dout_resid = rq->raw_data_len; + hdr->dout_resid = blk_rq_raw_data_len(rq); /* * If the request generated a negative error number, return it Index: work/block/scsi_ioctl.c =================================================================== --- work.orig/block/scsi_ioctl.c +++ work/block/scsi_ioctl.c @@ -266,7 +266,7 @@ static int blk_complete_sghdr_rq(struct hdr->info = 0; if (hdr->masked_status || hdr->host_status || hdr->driver_status) hdr->info |= SG_INFO_CHECK; - hdr->resid = rq->raw_data_len; + hdr->resid = blk_rq_raw_data_len(rq); hdr->sb_len_wr = 0; if (rq->sense_len && hdr->sbp) { @@ -528,8 +528,8 @@ static int __blk_send_generic(struct req rq = blk_get_request(q, WRITE, __GFP_WAIT); rq->cmd_type = REQ_TYPE_BLOCK_PC; rq->data = NULL; - rq->raw_data_len = 0; rq->data_len = 0; + rq->extra_len = 0; rq->timeout = BLK_DEFAULT_SG_TIMEOUT; memset(rq->cmd, 0, sizeof(rq->cmd)); rq->cmd[0] = cmd; Index: work/drivers/ata/libata-scsi.c =================================================================== --- work.orig/drivers/ata/libata-scsi.c +++ work/drivers/ata/libata-scsi.c @@ -2549,7 +2549,8 @@ static unsigned int atapi_xlat(struct at * want to set it properly, and for DMA where it is * effectively meaningless. */ - nbytes = min(scmd->request->raw_data_len, (unsigned int)63 * 1024); + nbytes = min(blk_rq_raw_data_len(scmd->request), + (unsigned int)63 * 1024); /* Most ATAPI devices which honor transfer chunk size don't * behave according to the spec when odd chunk size which Index: work/include/linux/blkdev.h =================================================================== --- work.orig/include/linux/blkdev.h +++ work/include/linux/blkdev.h @@ -216,8 +216,8 @@ struct request { unsigned int cmd_len; unsigned char cmd[BLK_MAX_CDB]; - unsigned int raw_data_len; unsigned int data_len; + unsigned int extra_len; /* length of padding and draining buffers */ unsigned int sense_len; void *data; void *sense; @@ -477,6 +477,12 @@ enum { #define rq_data_dir(rq) ((rq)->cmd_flags & 1) +/* data_len of the request sans extra stuff for padding and draining */ +static inline unsigned int blk_rq_raw_data_len(struct request *rq) +{ + return rq->data_len - min(rq->extra_len, rq->data_len); +} + /* * We regard a request as sync, if it's a READ or a SYNC write. */ Index: work/block/blk-settings.c =================================================================== --- work.orig/block/blk-settings.c +++ work/block/blk-settings.c @@ -309,6 +309,10 @@ EXPORT_SYMBOL(blk_queue_stack_limits); * does is adjust the queue so that the buf is always appended * silently to the scatterlist. * + * Appending draining buffer to a request modifies ->data_len such + * that it includes the drain buffer. The original requested data + * length can be obtained using blk_rq_raw_data_len(). + * * Note: This routine adjusts max_hw_segments to make room for * appending the drain buffer. If you call * blk_queue_max_hw_segments() or blk_queue_max_phys_segments() after -- To unsubscribe from this list: send the line "unsubscribe linux-ide" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html