diff options
| author | Christoph Hellwig <hch@lst.de> | 2016-04-18 23:06:28 +0200 |
|---|---|---|
| committer | Doug Ledford <dledford@redhat.com> | 2016-05-05 18:41:24 +0200 |
| commit | 9c674815d346305068b27bf03b5e86b659a1b111 (patch) | |
| tree | b512f798dbdfb99a3cfb7ca971b04fc5c52a9000 /drivers | |
| parent | RDMA/nes: don't leak skb if carrier down (diff) | |
| download | linux-9c674815d346305068b27bf03b5e86b659a1b111.tar.xz linux-9c674815d346305068b27bf03b5e86b659a1b111.zip | |
IB/iser: Fix max_sectors calculation
iSER currently has a couple places that set max_sectors in either the host
template or SCSI host, and all of them get it wrong.
This patch instead uses a single assignment that (hopefully) gets it right:
the max_sectors value must be derived from the number of segments in the
FR or FMR structure, but actually be one lower than the page size multiplied
by the number of sectors, as it has to handle the case of non-aligned I/O.
Without this I get trivial to reproduce hangs when running xfstests
(on XFS) over iSER to Linux targets.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Max Gurtovoy <maxg@mellanox.com>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Diffstat (limited to 'drivers')
| -rw-r--r-- | drivers/infiniband/ulp/iser/iscsi_iser.c | 14 |
1 files changed, 10 insertions, 4 deletions
diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c b/drivers/infiniband/ulp/iser/iscsi_iser.c index 80b6bedc172f..64b3d11dcf1e 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.c +++ b/drivers/infiniband/ulp/iser/iscsi_iser.c @@ -612,6 +612,7 @@ iscsi_iser_session_create(struct iscsi_endpoint *ep, struct Scsi_Host *shost; struct iser_conn *iser_conn = NULL; struct ib_conn *ib_conn; + u32 max_fr_sectors; u16 max_cmds; shost = iscsi_host_alloc(&iscsi_iser_sht, 0, 0); @@ -632,7 +633,6 @@ iscsi_iser_session_create(struct iscsi_endpoint *ep, iser_conn = ep->dd_data; max_cmds = iser_conn->max_cmds; shost->sg_tablesize = iser_conn->scsi_sg_tablesize; - shost->max_sectors = iser_conn->scsi_max_sectors; mutex_lock(&iser_conn->state_mutex); if (iser_conn->state != ISER_CONN_UP) { @@ -657,8 +657,6 @@ iscsi_iser_session_create(struct iscsi_endpoint *ep, */ shost->sg_tablesize = min_t(unsigned short, shost->sg_tablesize, ib_conn->device->ib_device->attrs.max_fast_reg_page_list_len); - shost->max_sectors = min_t(unsigned int, - 1024, (shost->sg_tablesize * PAGE_SIZE) >> 9); if (iscsi_host_add(shost, ib_conn->device->ib_device->dma_device)) { @@ -672,6 +670,15 @@ iscsi_iser_session_create(struct iscsi_endpoint *ep, goto free_host; } + /* + * FRs or FMRs can only map up to a (device) page per entry, but if the + * first entry is misaligned we'll end up using using two entries + * (head and tail) for a single page worth data, so we have to drop + * one segment from the calculation. + */ + max_fr_sectors = ((shost->sg_tablesize - 1) * PAGE_SIZE) >> 9; + shost->max_sectors = min(iser_max_sectors, max_fr_sectors); + if (cmds_max > max_cmds) { iser_info("cmds_max changed from %u to %u\n", cmds_max, max_cmds); @@ -989,7 +996,6 @@ static struct scsi_host_template iscsi_iser_sht = { .queuecommand = iscsi_queuecommand, .change_queue_depth = scsi_change_queue_depth, .sg_tablesize = ISCSI_ISER_DEF_SG_TABLESIZE, - .max_sectors = ISER_DEF_MAX_SECTORS, .cmd_per_lun = ISER_DEF_CMD_PER_LUN, .eh_abort_handler = iscsi_eh_abort, .eh_device_reset_handler= iscsi_eh_device_reset, |