| Message ID | 20230928020907.5730-1-michael.christie@oracle.com (mailing list archive) |
|---|---|
| State | Accepted |
| Headers | show |
| Series | [v2,1/8] scsi: target: Make write_pending_must_be_called a bit field | expand |
The following patches were made over Linus's tree but apply over Martin's branches. They allow userspace to configure how fabric drivers submit cmds to backend drivers. Right now loop and vhost use a worker thread, and the other drivers submit from the contexts they receive/process the cmd from. For multiple LUN cases where the target can queue more cmds than the backend can handle then deferring to a worker thread is safest because the backend driver can block when doing things like waiting for a free request/tag. Deferring also helps when the target has to handle transport level requests from the recv context. For cases where the backend devices can queue everything the target sends, then there is no need to defer to a workqueue and you can see a perf boost of up to 26% for small IO workloads. For a nvme device and vhost-scsi I can see with 4K IOs: fio jobs 1 2 4 8 10 -------------------------------------------------- workqueue submit 94K 190K 394K 770K 890K direct submit 128K 252K 488K 950K - v2: - Use sysfs_emit. - Add iSCSI target support so now all targets are supported.
Mike, > The following patches were made over Linus's tree but apply over Martin's > branches. They allow userspace to configure how fabric drivers submit cmds > to backend drivers. Applied to 6.7/scsi-staging, thanks!
On Wed, 27 Sep 2023 21:31:32 -0500, Mike Christie wrote: > The following patches were made over Linus's tree but apply over Martin's > branches. They allow userspace to configure how fabric drivers submit cmds > to backend drivers. > > Right now loop and vhost use a worker thread, and the other drivers submit > from the contexts they receive/process the cmd from. For multiple LUN > cases where the target can queue more cmds than the backend can handle > then deferring to a worker thread is safest because the backend driver can > block when doing things like waiting for a free request/tag. Deferring also > helps when the target has to handle transport level requests from the > recv context. > > [...] Applied to 6.7/scsi-queue, thanks! [1/8] scsi: target: Make write_pending_must_be_called a bit field https://git.kernel.org/mkp/scsi/c/40ddd6df93a3 [2/8] scsi: target: Have drivers report if they support direct submissions https://git.kernel.org/mkp/scsi/c/194605d45dcb [3/8] target: Move core_alua_check_nonop_delay call https://git.kernel.org/mkp/scsi/c/ee48345e1cca [4/8] target: Move buffer clearing hack https://git.kernel.org/mkp/scsi/c/5c48a4ea3280 [5/8] target: Kill transport_handle_cdb_direct https://git.kernel.org/mkp/scsi/c/428926796e7f [6/8] scsi: target: Allow userspace to request direct submissions https://git.kernel.org/mkp/scsi/c/e2f4ea40138e [7/8] scsi: target: Unexport target_queue_submission https://git.kernel.org/mkp/scsi/c/e344c00e7ccd [8/8] scsi: target: Export fabric driver direct submit settings https://git.kernel.org/mkp/scsi/c/6dbc829d101d
diff --git a/drivers/target/iscsi/iscsi_target_configfs.c b/drivers/target/iscsi/iscsi_target_configfs.c index 1cff6052e820..bf190dcb9eee 100644 --- a/drivers/target/iscsi/iscsi_target_configfs.c +++ b/drivers/target/iscsi/iscsi_target_configfs.c @@ -1589,5 +1589,5 @@ const struct target_core_fabric_ops iscsi_ops = { .tfc_tpg_nacl_auth_attrs = lio_target_nacl_auth_attrs, .tfc_tpg_nacl_param_attrs = lio_target_nacl_param_attrs, - .write_pending_must_be_called = true, + .write_pending_must_be_called = 1, }; diff --git a/include/target/target_core_fabric.h b/include/target/target_core_fabric.h index b188b1e90e1e..2a6c4c935666 100644 --- a/include/target/target_core_fabric.h +++ b/include/target/target_core_fabric.h @@ -113,11 +113,11 @@ struct target_core_fabric_ops { struct configfs_attribute **tfc_tpg_nacl_param_attrs; /* - * Set this member variable to true if the SCSI transport protocol + * Set this member variable if the SCSI transport protocol * (e.g. iSCSI) requires that the Data-Out buffer is transferred in * its entirety before a command is aborted. */ - bool write_pending_must_be_called; + unsigned int write_pending_must_be_called:1; }; int target_register_template(const struct target_core_fabric_ops *fo);
The next patches add more on/off type of settings to the target_core_fabric_ops struct so this makes write_pending_must_be_called a bit field instead of a bool to better organize the settings. Signed-off-by: Mike Christie <michael.christie@oracle.com> --- drivers/target/iscsi/iscsi_target_configfs.c | 2 +- include/target/target_core_fabric.h | 4 ++-- 2 files changed, 3 insertions(+), 3 deletions(-)