nvme qemu cleanups and fixes
cover nvme qemu cleanups and fixes
After going through the zns enabling, I notice the controller enabling is not correct. Then I just continued maked more stuff. The series, I think, contains some of the less controversial patches from the two conflicting zns series, preceeded by some cleanups and fixes from me.
If this is all fine, I took the liberty of porting the zns enabling to it and made a public branch for consideration here:
http://git.infradead.org/qemu-nvme.git/shortlog/refs/heads/kb-zns
Dmitry Fomichev (1):
hw/block/nvme: report actual LBA data shift in LBAF
Keith Busch (5):
hw/block/nvme: remove pointless rw indirection hw/block/nvme: fix log page offset check hw/block/nvme: support per-namespace smart log hw/block/nvme: validate command set selected hw/block/nvme: support for admin-only command set
Klaus Jensen (3):
hw/block/nvme: reject io commands if only admin command set selected hw/block/nvme: add nsid to get/setfeat trace events hw/block/nvme: add trace event for requests with non-zero status code
hw/block/nvme-ns.c | 5 ++ hw/block/nvme.c | 194 ++++++++++++++++++++---------------------- hw/block/trace-events | 6 +- include/block/nvme.h | 11 +++ 4 files changed, 114 insertions(+), 102 deletions(-)
Signature · 1 line
2.24.1
PATCH 1/9 hw/block/nvme: remove pointless rw indirection
The code switches on the opcode to invoke a function specific to that opcode. There's no point in consolidating back to a common function that just switches on that same opcode without any actual common code. Restore the opcode specific behavior without going back through another level of switches.
--- hw/block/nvme.c | 91 ++++++++++++++++--------------------------------- 1 file changed, 29 insertions(+), 62 deletions(-)
hw/block/nvme.c +29 -62
@@ -927,68 +927,12 @@ static void nvme_rw_cb(void *opaque, int ret)nvme_enqueue_req_completion (nvme_cq (req), req); } -static uint16_t nvme_do_aio (BlockBackend *blk,int64_t offset,size_t len, -NvmeRequest *req) -{ -BlockAcctCookie *acct = &req->acct; -BlockAcctStats *stats =blk_get_stats (blk); - -bool is_write = false; - -trace_pci_nvme_do_aio (nvme_cid (req), req->cmd.opcode, -nvme_io_opc_str (req->cmd.opcode),blk_name (blk), - offset, len); - -switch (req->cmd.opcode) { -case NVME_CMD_FLUSH : -block_acct_start (stats, acct,0 ,BLOCK_ACCT_FLUSH ); - req->aiocb =blk_aio_flush (blk, nvme_rw_cb, req); -break ; - -case NVME_CMD_WRITE_ZEROES : -block_acct_start (stats, acct, len,BLOCK_ACCT_WRITE ); - req->aiocb =blk_aio_pwrite_zeroes (blk, offset, len, -BDRV_REQ_MAY_UNMAP , nvme_rw_cb, - req); -break ; - -case NVME_CMD_WRITE : - is_write = true; - -/* fallthrough */ - -case NVME_CMD_READ : -block_acct_start (stats, acct, len, - is_write ?BLOCK_ACCT_WRITE :BLOCK_ACCT_READ ); - -if (req->qsg.sg) { -if (is_write) { - req->aiocb =dma_blk_write (blk, &req->qsg, offset, -BDRV_SECTOR_SIZE , nvme_rw_cb, req); - }else { - req->aiocb =dma_blk_read (blk, &req->qsg, offset, -BDRV_SECTOR_SIZE , nvme_rw_cb, req); - } - }else { -if (is_write) { - req->aiocb =blk_aio_pwritev (blk, offset, &req->iov,0 , - nvme_rw_cb, req); - }else { - req->aiocb =blk_aio_preadv (blk, offset, &req->iov,0 , - nvme_rw_cb, req); - } - } - -break ; - } - -return NVME_NO_COMPLETE ; -} -static uint16_t nvme_flush (NvmeCtrl *n,NvmeRequest *req) { -NvmeNamespace *ns = req->ns; -return nvme_do_aio (ns->blkconf.blk,0 ,0 , req); +block_acct_start (blk_get_stats (n->conf.blk), &req->acct,0 , +BLOCK_ACCT_FLUSH ); + req->aiocb =blk_aio_flush (n->conf.blk, nvme_rw_cb, req); +return NVME_NO_COMPLETE ; }static uint16_t nvme_write_zeroes (NvmeCtrl *n,NvmeRequest *req)
@@ -1009,7 +953,11 @@ static uint16_t nvme_write_zeroes(NvmeCtrl *n, NvmeRequest *req)return status; } -return nvme_do_aio (ns->blkconf.blk, offset, count, req); +block_acct_start (blk_get_stats (n->conf.blk), &req->acct,0 , +BLOCK_ACCT_WRITE ); + req->aiocb =blk_aio_pwrite_zeroes (n->conf.blk, offset, count, +BDRV_REQ_MAY_UNMAP , nvme_rw_cb, req); +return NVME_NO_COMPLETE ; }static uint16_t nvme_rw (NvmeCtrl *n,NvmeRequest *req)
@@ -1023,6 +971,7 @@ static uint16_t nvme_rw(NvmeCtrl *n, NvmeRequest *req)uint64_t data_offset =nvme_l2b (ns, slba);enum BlockAcctType acct = req->cmd.opcode ==NVME_CMD_WRITE ?BLOCK_ACCT_WRITE :BLOCK_ACCT_READ ; +BlockBackend *blk = ns->blkconf.blk;uint16_t status; trace_pci_nvme_rw(nvme_cid(req), nvme_io_opc_str(rw->opcode),
@@ -1045,7 +994,25 @@ static uint16_t nvme_rw(NvmeCtrl *n, NvmeRequest *req) goto invalid; } -return nvme_do_aio (ns->blkconf.blk, data_offset, data_size, req); +block_acct_start (blk_get_stats (blk), &req->acct, data_size, acct); +if (req->qsg.sg) { +if (acct ==BLOCK_ACCT_WRITE ) { + req->aiocb =dma_blk_write (blk, &req->qsg, data_offset, +BDRV_SECTOR_SIZE , nvme_rw_cb, req); + }else { + req->aiocb =dma_blk_read (blk, &req->qsg, data_offset, +BDRV_SECTOR_SIZE , nvme_rw_cb, req); + } + }else { +if (acct ==BLOCK_ACCT_WRITE ) { + req->aiocb =blk_aio_pwritev (blk, data_offset, &req->iov,0 , + nvme_rw_cb, req); + }else { + req->aiocb =blk_aio_preadv (blk, data_offset, &req->iov,0 , + nvme_rw_cb, req); + } + } +return NVME_NO_COMPLETE ; invalid:block_acct_invalid (blk_get_stats (ns->blkconf.blk), acct);
Signature · 1 line
2.24.1
Re: [PATCH 1/9] hw/block/nvme: remove pointless rw indirection
Quoted 7 lines · Keith Busch
The code switches on the opcode to invoke a function specific to that opcode. There's no point in consolidating back to a common function that just switches on that same opcode without any actual common code. Restore the opcode specific behavior without going back through another level of switches.
Signed-off-by Keith Busch <[email protected]>
Point taken. I could've sweared I had a better reason for this.
Quoted 133 lines
--- hw/block/nvme.c | 91 ++++++++++++++++--------------------------------- 1 file changed, 29 insertions(+), 62 deletions(-)hw/block/nvme.c +29
-62@@ -927,68 +927,12 @@ static void nvme_rw_cb(void *opaque, int ret)nvme_enqueue_req_completion (nvme_cq (req), req); } -static uint16_t nvme_do_aio (BlockBackend *blk,int64_t offset,size_t len, -NvmeRequest *req) -{ -BlockAcctCookie *acct = &req->acct; -BlockAcctStats *stats =blk_get_stats (blk); - -bool is_write = false; - -trace_pci_nvme_do_aio (nvme_cid (req), req->cmd.opcode, -nvme_io_opc_str (req->cmd.opcode),blk_name (blk), - offset, len); - -switch (req->cmd.opcode) { -case NVME_CMD_FLUSH : -block_acct_start (stats, acct,0 ,BLOCK_ACCT_FLUSH ); - req->aiocb =blk_aio_flush (blk, nvme_rw_cb, req); -break ; - -case NVME_CMD_WRITE_ZEROES : -block_acct_start (stats, acct, len,BLOCK_ACCT_WRITE ); - req->aiocb =blk_aio_pwrite_zeroes (blk, offset, len, -BDRV_REQ_MAY_UNMAP , nvme_rw_cb, - req); -break ; - -case NVME_CMD_WRITE : - is_write = true; - -/* fallthrough */ - -case NVME_CMD_READ : -block_acct_start (stats, acct, len, - is_write ?BLOCK_ACCT_WRITE :BLOCK_ACCT_READ ); - -if (req->qsg.sg) { -if (is_write) { - req->aiocb =dma_blk_write (blk, &req->qsg, offset, -BDRV_SECTOR_SIZE , nvme_rw_cb, req); - }else { - req->aiocb =dma_blk_read (blk, &req->qsg, offset, -BDRV_SECTOR_SIZE , nvme_rw_cb, req); - } - }else { -if (is_write) { - req->aiocb =blk_aio_pwritev (blk, offset, &req->iov,0 , - nvme_rw_cb, req); - }else { - req->aiocb =blk_aio_preadv (blk, offset, &req->iov,0 , - nvme_rw_cb, req); - } - } - -break ; - } - -return NVME_NO_COMPLETE ; -} -static uint16_t nvme_flush (NvmeCtrl *n,NvmeRequest *req) { -NvmeNamespace *ns = req->ns; -return nvme_do_aio (ns->blkconf.blk,0 ,0 , req); +block_acct_start (blk_get_stats (n->conf.blk), &req->acct,0 , +BLOCK_ACCT_FLUSH ); + req->aiocb =blk_aio_flush (n->conf.blk, nvme_rw_cb, req); +return NVME_NO_COMPLETE ; }static uint16_t nvme_write_zeroes (NvmeCtrl *n,NvmeRequest *req)@@ -1009,7 +953,11 @@ static uint16_t nvme_write_zeroes(NvmeCtrl *n, NvmeRequest *req)return status; } -return nvme_do_aio (ns->blkconf.blk, offset, count, req); +block_acct_start (blk_get_stats (n->conf.blk), &req->acct,0 , +BLOCK_ACCT_WRITE ); + req->aiocb =blk_aio_pwrite_zeroes (n->conf.blk, offset, count, +BDRV_REQ_MAY_UNMAP , nvme_rw_cb, req); +return NVME_NO_COMPLETE ; }static uint16_t nvme_rw (NvmeCtrl *n,NvmeRequest *req)@@ -1023,6 +971,7 @@ static uint16_t nvme_rw(NvmeCtrl *n, NvmeRequest *req)uint64_t data_offset =nvme_l2b (ns, slba);enum BlockAcctType acct = req->cmd.opcode ==NVME_CMD_WRITE ?BLOCK_ACCT_WRITE :BLOCK_ACCT_READ ; +BlockBackend *blk = ns->blkconf.blk;uint16_t status; trace_pci_nvme_rw(nvme_cid(req), nvme_io_opc_str(rw->opcode),@@ -1045,7 +994,25 @@ static uint16_t nvme_rw(NvmeCtrl *n, NvmeRequest *req) goto invalid; } -return nvme_do_aio (ns->blkconf.blk, data_offset, data_size, req); +block_acct_start (blk_get_stats (blk), &req->acct, data_size, acct); +if (req->qsg.sg) { +if (acct ==BLOCK_ACCT_WRITE ) { + req->aiocb =dma_blk_write (blk, &req->qsg, data_offset, +BDRV_SECTOR_SIZE , nvme_rw_cb, req); + }else { + req->aiocb =dma_blk_read (blk, &req->qsg, data_offset, +BDRV_SECTOR_SIZE , nvme_rw_cb, req); + } + }else { +if (acct ==BLOCK_ACCT_WRITE ) { + req->aiocb =blk_aio_pwritev (blk, data_offset, &req->iov,0 , + nvme_rw_cb, req); + }else { + req->aiocb =blk_aio_preadv (blk, data_offset, &req->iov,0 , + nvme_rw_cb, req); + } + } +return NVME_NO_COMPLETE ; invalid:block_acct_invalid (blk_get_stats (ns->blkconf.blk), acct);Signature · 1 line
2.24.1
Signature · 1 line
One of us - No more doubt, silence or taboo about mental illness.
Re: [PATCH 1/9] hw/block/nvme: remove pointless rw indirection
Quoted 7 lines · Klaus Jensen
Quoted 6 lines · Keith Busch
The code switches on the opcode to invoke a function specific to that opcode. There's no point in consolidating back to a common function that just switches on that same opcode without any actual common code. Restore the opcode specific behavior without going back through another level of switches.
Quoted 6 lines
Quoted 1 line
Signed-off-by Keith Busch <[email protected]>Reviewed-by Klaus Jensen <[email protected]>Point taken. I could've sweared I had a better reason for this.
Can you also remove the nvme_do_aio trace event?
Re: [PATCH 1/9] hw/block/nvme: remove pointless rw indirection
Quoted 8 lines · Klaus Jensen
Quoted 7 lines · Klaus Jensen
Quoted 6 lines · Keith Busch
The code switches on the opcode to invoke a function specific to that opcode. There's no point in consolidating back to a common function that just switches on that same opcode without any actual common code. Restore the opcode specific behavior without going back through another level of switches.
Quoted 8 lines
Quoted 6 lines
Quoted 1 line
Signed-off-by Keith Busch <[email protected]>Reviewed-by Klaus Jensen <[email protected]>Point taken. I could've sweared I had a better reason for this.
Can you also remove the
nvme_do_aiotrace event?
Ah, thanks for pointing that out. I'll fix up the trace.
I think you may have a case where this might make sense still in the making? If so we can reevaluate when it's ready.
Re: [PATCH 1/9] hw/block/nvme: remove pointless rw indirection
In reply to Klaus JensenQuoted 80 lines · Klaus Jensen
Quoted 7 lines · Keith Busch
The code switches on the opcode to invoke a function specific to that opcode. There's no point in consolidating back to a common function that just switches on that same opcode without any actual common code. Restore the opcode specific behavior without going back through another level of switches.
Signed-off-by Keith Busch <[email protected]>Reviewed-by Klaus Jensen <[email protected]>Point taken. I could've sweared I had a better reason for this.
Quoted 67 lines
--- hw/block/nvme.c | 91 ++++++++++++++++--------------------------------- 1 file changed, 29 insertions(+), 62 deletions(-)hw/block/nvme.c +0
-55@@ -927,68 +927,12 @@ static void nvme_rw_cb(void *opaque, int ret)nvme_enqueue_req_completion (nvme_cq (req), req); } -static uint16_t nvme_do_aio (BlockBackend *blk,int64_t offset,size_t len, -NvmeRequest *req) -{ -BlockAcctCookie *acct = &req->acct; -BlockAcctStats *stats =blk_get_stats (blk); - -bool is_write = false; - -trace_pci_nvme_do_aio (nvme_cid (req), req->cmd.opcode, -nvme_io_opc_str (req->cmd.opcode),blk_name (blk), - offset, len); - -switch (req->cmd.opcode) { -case NVME_CMD_FLUSH : -block_acct_start (stats, acct,0 ,BLOCK_ACCT_FLUSH ); - req->aiocb =blk_aio_flush (blk, nvme_rw_cb, req); -break ; - -case NVME_CMD_WRITE_ZEROES : -block_acct_start (stats, acct, len,BLOCK_ACCT_WRITE ); - req->aiocb =blk_aio_pwrite_zeroes (blk, offset, len, -BDRV_REQ_MAY_UNMAP , nvme_rw_cb, - req); -break ; - -case NVME_CMD_WRITE : - is_write = true; - -/* fallthrough */ - -case NVME_CMD_READ : -block_acct_start (stats, acct, len, - is_write ?BLOCK_ACCT_WRITE :BLOCK_ACCT_READ ); - -if (req->qsg.sg) { -if (is_write) { - req->aiocb =dma_blk_write (blk, &req->qsg, offset, -BDRV_SECTOR_SIZE , nvme_rw_cb, req); - }else { - req->aiocb =dma_blk_read (blk, &req->qsg, offset, -BDRV_SECTOR_SIZE , nvme_rw_cb, req); - } - }else { -if (is_write) { - req->aiocb =blk_aio_pwritev (blk, offset, &req->iov,0 , - nvme_rw_cb, req); - }else { - req->aiocb =blk_aio_preadv (blk, offset, &req->iov,0 , - nvme_rw_cb, req); - } - } - -break ; - } -
Quoted 8 lines
hw/block/nvme.c static void nvme_rw_cb(void *opaque, int ret) · line 932
Quoted 8 lines
hw/block/nvme.c +1
-5-return NVME_NO_COMPLETE ; -} -static uint16_t nvme_flush (NvmeCtrl *n,NvmeRequest *req) { -NvmeNamespace *ns = req->ns; -return nvme_do_aio (ns->blkconf.blk,0 ,0 , req); + block_acct_start(blk_get_stats(n->conf.blk), &req->acct, 0,
Uh, ouch!
This and the rest needs to be changed to ns->blkconf.blk and not n->conf.blk.
Quoted 61 lines
Quoted 58 lines
hw/block/nvme.c +28
-2+ BLOCK_ACCT_FLUSH); + req->aiocb =blk_aio_flush (n->conf.blk, nvme_rw_cb, req); +return NVME_NO_COMPLETE ; }static uint16_t nvme_write_zeroes (NvmeCtrl *n,NvmeRequest *req)@@ -1009,7 +953,11 @@ static uint16_t nvme_write_zeroes(NvmeCtrl *n, NvmeRequest *req)return status; } -return nvme_do_aio (ns->blkconf.blk, offset, count, req); +block_acct_start (blk_get_stats (n->conf.blk), &req->acct,0 , +BLOCK_ACCT_WRITE ); + req->aiocb =blk_aio_pwrite_zeroes (n->conf.blk, offset, count, +BDRV_REQ_MAY_UNMAP , nvme_rw_cb, req); +return NVME_NO_COMPLETE ; }static uint16_t nvme_rw (NvmeCtrl *n,NvmeRequest *req)@@ -1023,6 +971,7 @@ static uint16_t nvme_rw(NvmeCtrl *n, NvmeRequest *req)uint64_t data_offset =nvme_l2b (ns, slba);enum BlockAcctType acct = req->cmd.opcode ==NVME_CMD_WRITE ?BLOCK_ACCT_WRITE :BLOCK_ACCT_READ ; +BlockBackend *blk = ns->blkconf.blk;uint16_t status; trace_pci_nvme_rw(nvme_cid(req), nvme_io_opc_str(rw->opcode),@@ -1045,7 +994,25 @@ static uint16_t nvme_rw(NvmeCtrl *n, NvmeRequest *req) goto invalid; } -return nvme_do_aio (ns->blkconf.blk, data_offset, data_size, req); +block_acct_start (blk_get_stats (blk), &req->acct, data_size, acct); +if (req->qsg.sg) { +if (acct ==BLOCK_ACCT_WRITE ) { + req->aiocb =dma_blk_write (blk, &req->qsg, data_offset, +BDRV_SECTOR_SIZE , nvme_rw_cb, req); + }else { + req->aiocb =dma_blk_read (blk, &req->qsg, data_offset, +BDRV_SECTOR_SIZE , nvme_rw_cb, req); + } + }else { +if (acct ==BLOCK_ACCT_WRITE ) { + req->aiocb =blk_aio_pwritev (blk, data_offset, &req->iov,0 , + nvme_rw_cb, req); + }else { + req->aiocb =blk_aio_preadv (blk, data_offset, &req->iov,0 , + nvme_rw_cb, req); + } + } +return NVME_NO_COMPLETE ; invalid:block_acct_invalid (blk_get_stats (ns->blkconf.blk), acct);Signature · 1 line
2.24.1Signature · 1 line
One of us - No more doubt, silence or taboo about mental illness.
Signature · 1 line
One of us - No more doubt, silence or taboo about mental illness.
RE: [PATCH 1/9] hw/block/nvme: remove pointless rw indirection
In reply to Keith BuschQuoted 82 lines
-----Original Message-----
From: Keith Busch <[email protected]>
Sent: Wednesday, September 30, 2020 6:04 PM
To: [email protected]; [email protected]; Klaus Jensen <[email protected]>
Cc: Niklas Cassel <[email protected]>; Dmitry Fomichev <[email protected]>; Kevin Wolf <[email protected]>; Philippe Mathieu-Daudé <[email protected]>; Keith Busch <[email protected]>
Subject: [PATCH 1/9] hw/block/nvme: remove pointless rw indirectionThe code switches on the opcode to invoke a function specific to that opcode. There's no point in consolidating back to a common function that just switches on that same opcode without any actual common code. Restore the opcode specific behavior without going back through another level of switches.
Signed-off-by Keith Busch <[email protected]>--- hw/block/nvme.c | 91 ++++++++++++++++--------------------------------- 1 file changed, 29 insertions(+), 62 deletions(-)hw/block/nvme.c +0
-53@@ -927,68 +927,12 @@ static void nvme_rw_cb(void *opaque, int ret)nvme_enqueue_req_completion (nvme_cq (req), req); } -static uint16_t nvme_do_aio (BlockBackend *blk,int64_t offset,size_t len, -NvmeRequest *req) -{ -BlockAcctCookie *acct = &req->acct; -BlockAcctStats *stats =blk_get_stats (blk); - -bool is_write = false; - -trace_pci_nvme_do_aio (nvme_cid (req), req->cmd.opcode, -nvme_io_opc_str (req->cmd.opcode),blk_name (blk), - offset, len); - -switch (req->cmd.opcode) { -case NVME_CMD_FLUSH : -block_acct_start (stats, acct,0 ,BLOCK_ACCT_FLUSH ); - req->aiocb =blk_aio_flush (blk, nvme_rw_cb, req); -break ; - -case NVME_CMD_WRITE_ZEROES : -block_acct_start (stats, acct, len,BLOCK_ACCT_WRITE ); - req->aiocb =blk_aio_pwrite_zeroes (blk, offset, len, -BDRV_REQ_MAY_UNMAP , nvme_rw_cb, - req); -break ; - -case NVME_CMD_WRITE : - is_write = true; - -/* fallthrough */ - -case NVME_CMD_READ : -block_acct_start (stats, acct, len, - is_write ?BLOCK_ACCT_WRITE :BLOCK_ACCT_READ ); - -if (req->qsg.sg) { -if (is_write) { - req->aiocb =dma_blk_write (blk, &req->qsg, offset, -BDRV_SECTOR_SIZE , nvme_rw_cb, req); - }else { - req->aiocb =dma_blk_read (blk, &req->qsg, offset, -BDRV_SECTOR_SIZE , nvme_rw_cb, req); - } - }else { -if (is_write) { - req->aiocb =blk_aio_pwritev (blk, offset, &req->iov,0 , - nvme_rw_cb, req); - }else { - req->aiocb =blk_aio_preadv (blk, offset, &req->iov,0 , - nvme_rw_cb, req); - } - } - -break ;
hw/block/nvme.c static void nvme_rw_cb(void *opaque, int ret) · line 932
Quoted 10 lines
hw/block/nvme.c +1
-7- } - -return NVME_NO_COMPLETE ; -} -static uint16_t nvme_flush (NvmeCtrl *n,NvmeRequest *req) { -NvmeNamespace *ns = req->ns; -return nvme_do_aio (ns->blkconf.blk,0 ,0 , req); + block_acct_start(blk_get_stats(n->conf.blk), &req->acct, 0,
This should be block_acct_start(blk_get_stats(req->ns->blkconf.blk), &req->acct, 0,
hw/block/nvme.c static void nvme_rw_cb(void *opaque, int ret) · line 934
Quoted 2 lines
hw/block/nvme.c +2
-0+ BLOCK_ACCT_FLUSH); + req->aiocb =blk_aio_flush (n->conf.blk, nvme_rw_cb, req);
and herereq->aiocb = blk_aio_flush(req->ns->blkconf.blk, nvme_rw_cb, req);
hw/block/nvme.c static uint16_t nvme_write_zeroes(NvmeCtrl *n, · line 957
Quoted 11 lines
hw/block/nvme.c +2
-1+return NVME_NO_COMPLETE ; }static uint16_t nvme_write_zeroes (NvmeCtrl *n,NvmeRequest *req)@@ -1009,7 +953,11 @@ static uint16_t nvme_write_zeroes(NvmeCtrl *n, NvmeRequest *req)return status; } -return nvme_do_aio (ns->blkconf.blk, offset, count, req); + block_acct_start(blk_get_stats(n->conf.blk), &req->acct, 0,
block_acct_start(blk_get_stats(ns->blkconf.blk), &req->acct, 0,
hw/block/nvme.c static uint16_t nvme_write_zeroes(NvmeCtrl *n, NvmeRequest *req) · line 958
Quoted 2 lines
hw/block/nvme.c +2
-0+ BLOCK_ACCT_WRITE); + req->aiocb = blk_aio_pwrite_zeroes(n->conf.blk, offset, count,
Here as well,
req->aiocb = blk_aio_pwrite_zeroes(ns->blkconf.blk, offset, count,
Quoted 44 lines
hw/block/nvme.c +2
-0+ BDRV_REQ_MAY_UNMAP, nvme_rw_cb, req); +return NVME_NO_COMPLETE ; }static
uint16_tnvme_rw(NvmeCtrl *n, NvmeRequest *req)hw/block/nvme.c +1
-0@@ -1023,6 +971,7 @@ static uint16_t nvme_rw(NvmeCtrl *n, NvmeRequest *req)uint64_t data_offset =nvme_l2b (ns, slba);enum BlockAcctType acct = req->cmd.opcode ==NVME_CMD_WRITE ?BLOCK_ACCT_WRITE :BLOCK_ACCT_READ ; +BlockBackend *blk = ns->blkconf.blk;uint16_t status;trace_pci_nvme_rw(nvme_cid(req), nvme_io_opc_str(rw->opcode),hw/block/nvme.c +19
-1@@ -1045,7 +994,25 @@ static uint16_t nvme_rw(NvmeCtrl *n, NvmeRequest *req) goto invalid; } -return nvme_do_aio (ns->blkconf.blk, data_offset, data_size, req); +block_acct_start (blk_get_stats (blk), &req->acct, data_size, acct); +if (req->qsg.sg) { +if (acct ==BLOCK_ACCT_WRITE ) { + req->aiocb =dma_blk_write (blk, &req->qsg, data_offset, +BDRV_SECTOR_SIZE , nvme_rw_cb, req); + }else { + req->aiocb =dma_blk_read (blk, &req->qsg, data_offset, +BDRV_SECTOR_SIZE , nvme_rw_cb, req); + } + }else { +if (acct ==BLOCK_ACCT_WRITE ) { + req->aiocb =blk_aio_pwritev (blk, data_offset, &req->iov,0 , + nvme_rw_cb, req); + }else { + req->aiocb =blk_aio_preadv (blk, data_offset, &req->iov,0 , + nvme_rw_cb, req); + } + } +return NVME_NO_COMPLETE ; invalid:block_acct_invalid(blk_get_stats(ns->blkconf.blk), acct);--
2.24.1
PATCH 2/9 hw/block/nvme: fix log page offset check
In reply to Keith BuschReturn error if the requested offset starts after the size of the log being returned. Also, move the check for earlier in the function so we're not doing unnecessary calculations.
--- hw/block/nvme.c | 22 ++++++++++------------ 1 file changed, 10 insertions(+), 12 deletions(-)
hw/block/nvme.c +10 -12
@@ -1179,6 +1179,10 @@ static uint16_t nvme_smart_info(NvmeCtrl *n, uint8_t rae, uint32_t buf_len,return NVME_INVALID_FIELD |NVME_DNR ; } +if (off >=sizeof (smart)) { +return NVME_INVALID_FIELD |NVME_DNR ; + } +for (int i =1 ; i <= n->num_namespaces; i++) {NvmeNamespace *ns =nvme_ns (n, i);if (!ns) {
@@ -1193,10 +1197,6 @@ static uint16_t nvme_smart_info(NvmeCtrl *n, uint8_t rae, uint32_t buf_len, write_commands += s->nr_ops[BLOCK_ACCT_WRITE ]; } -if (off >sizeof (smart)) { -return NVME_INVALID_FIELD |NVME_DNR ; - } - trans_len =MIN (sizeof (smart) - off, buf_len);memset (&smart,0x0 ,sizeof (smart));
@@ -1234,12 +1234,11 @@ static uint16_t nvme_fw_log_info(NvmeCtrl *n, uint32_t buf_len, uint64_t off, .afi = 0x1, }; -strpadcpy ((char *)&fw_log.frs1,sizeof (fw_log.frs1),"1.0" ,' ' ); - -if (off >sizeof (fw_log)) { +if (off >=sizeof (fw_log)) {return NVME_INVALID_FIELD |NVME_DNR ; } +strpadcpy ((char *)&fw_log.frs1,sizeof (fw_log.frs1),"1.0" ,' ' ); trans_len =MIN (sizeof (fw_log) - off, buf_len); return nvme_dma(n, (uint8_t *) &fw_log + off, trans_len,
@@ -1252,16 +1251,15 @@ static uint16_t nvme_error_info(NvmeCtrl *n, uint8_t rae, uint32_t buf_len,uint32_t trans_len;NvmeErrorLog errlog; -if (!rae) { -nvme_clear_events (n,NVME_AER_TYPE_ERROR ); +if (off >=sizeof (errlog)) { +return NVME_INVALID_FIELD |NVME_DNR ; } -if (off >sizeof (errlog)) { -return NVME_INVALID_FIELD |NVME_DNR ; +if (!rae) { +nvme_clear_events (n,NVME_AER_TYPE_ERROR ); }memset (&errlog,0x0 ,sizeof (errlog)); - trans_len =MIN (sizeof (errlog) - off, buf_len); return nvme_dma(n, (uint8_t *)&errlog, trans_len,
Signature · 1 line
2.24.1
RE: [PATCH 2/9] hw/block/nvme: fix log page offset check
Quoted 10 lines
-----Original Message-----
From: Keith Busch <[email protected]>
Sent: Wednesday, September 30, 2020 6:04 PM
To: [email protected]; [email protected]; Klaus Jensen <[email protected]>
Cc: Niklas Cassel <[email protected]>; Dmitry Fomichev <[email protected]>; Kevin Wolf <[email protected]>; Philippe Mathieu-Daudé <[email protected]>; Keith Busch <[email protected]>
Subject: [PATCH 2/9] hw/block/nvme: fix log page offset check
Quoted 5 lines
Return error if the requested offset starts after the size of the log being returned. Also, move the check for earlier in the function so we're not doing unnecessary calculations.
Signed-off-by Keith Busch <[email protected]>
Reviewed- by: Dmitry Fomichev <[email protected]>
Quoted 72 lines
--- hw/block/nvme.c | 22 ++++++++++------------ 1 file changed, 10 insertions(+), 12 deletions(-)hw/block/nvme.c +4
-0@@ -1179,6 +1179,10 @@ static uint16_t nvme_smart_info(NvmeCtrl *n, uint8_t rae, uint32_t buf_len, return NVME_INVALID_FIELD | NVME_DNR; } +if (off >=sizeof (smart)) { +return NVME_INVALID_FIELD |NVME_DNR ; + } +for (int i =1 ; i <= n->num_namespaces; i++) {NvmeNamespace *ns =nvme_ns (n, i);if (!ns) {hw/block/nvme.c +0
-4@@ -1193,10 +1197,6 @@ static uint16_t nvme_smart_info(NvmeCtrl *n, uint8_t rae, uint32_t buf_len, write_commands += s->nr_ops[BLOCK_ACCT_WRITE]; } -if (off >sizeof (smart)) { -return NVME_INVALID_FIELD |NVME_DNR ; - } - trans_len =MIN (sizeof (smart) - off, buf_len);memset(&smart, 0x0, sizeof(smart));hw/block/nvme.c +2
-3@@ -1234,12 +1234,11 @@ static uint16_t nvme_fw_log_info(NvmeCtrl *n, uint32_t buf_len, uint64_t off, .afi = 0x1, }; -strpadcpy ((char *)&fw_log.frs1,sizeof (fw_log.frs1),"1.0" ,' ' ); - -if (off >sizeof (fw_log)) { +if (off >=sizeof (fw_log)) {return NVME_INVALID_FIELD |NVME_DNR ; } +strpadcpy ((char *)&fw_log.frs1,sizeof (fw_log.frs1),"1.0" ,' ' ); trans_len =MIN (sizeof (fw_log) - off, buf_len);return nvme_dma(n, (
uint8_t*)&fw_log+ off,trans_len,hw/block/nvme.c +4
-5@@ -1252,16 +1251,15 @@ static uint16_t nvme_error_info(NvmeCtrl *n, uint8_t rae, uint32_t buf_len, uint32_t trans_len;NvmeErrorLog errlog; -if (!rae) { -nvme_clear_events (n,NVME_AER_TYPE_ERROR ); +if (off >=sizeof (errlog)) { +return NVME_INVALID_FIELD |NVME_DNR ; } -if (off >sizeof (errlog)) { -return NVME_INVALID_FIELD |NVME_DNR ; +if (!rae) { +nvme_clear_events (n,NVME_AER_TYPE_ERROR ); }memset (&errlog,0x0 ,sizeof (errlog)); - trans_len =MIN (sizeof (errlog) - off, buf_len);return nvme_dma(n, (
uint8_t*)&errlog,trans_len,
--
2.24.1
Re: [PATCH 2/9] hw/block/nvme: fix log page offset check
In reply to Keith BuschQuoted 5 lines · Keith Busch
Return error if the requested offset starts after the size of the log being returned. Also, move the check for earlier in the function so we're not doing unnecessary calculations.
Signed-off-by Keith Busch <[email protected]>
Quoted 70 lines
--- hw/block/nvme.c | 22 ++++++++++------------ 1 file changed, 10 insertions(+), 12 deletions(-)hw/block/nvme.c +10
-12@@ -1179,6 +1179,10 @@ static uint16_t nvme_smart_info(NvmeCtrl *n, uint8_t rae, uint32_t buf_len,return NVME_INVALID_FIELD |NVME_DNR ; } +if (off >=sizeof (smart)) { +return NVME_INVALID_FIELD |NVME_DNR ; + } +for (int i =1 ; i <= n->num_namespaces; i++) {NvmeNamespace *ns =nvme_ns (n, i);if (!ns) {@@ -1193,10 +1197,6 @@ static uint16_t nvme_smart_info(NvmeCtrl *n, uint8_t rae, uint32_t buf_len, write_commands += s->nr_ops[BLOCK_ACCT_WRITE ]; } -if (off >sizeof (smart)) { -return NVME_INVALID_FIELD |NVME_DNR ; - } - trans_len =MIN (sizeof (smart) - off, buf_len);memset (&smart,0x0 ,sizeof (smart));@@ -1234,12 +1234,11 @@ static uint16_t nvme_fw_log_info(NvmeCtrl *n, uint32_t buf_len, uint64_t off, .afi = 0x1, }; -strpadcpy ((char *)&fw_log.frs1,sizeof (fw_log.frs1),"1.0" ,' ' ); - -if (off >sizeof (fw_log)) { +if (off >=sizeof (fw_log)) {return NVME_INVALID_FIELD |NVME_DNR ; } +strpadcpy ((char *)&fw_log.frs1,sizeof (fw_log.frs1),"1.0" ,' ' ); trans_len =MIN (sizeof (fw_log) - off, buf_len); return nvme_dma(n, (uint8_t *) &fw_log + off, trans_len,@@ -1252,16 +1251,15 @@ static uint16_t nvme_error_info(NvmeCtrl *n, uint8_t rae, uint32_t buf_len,uint32_t trans_len;NvmeErrorLog errlog; -if (!rae) { -nvme_clear_events (n,NVME_AER_TYPE_ERROR ); +if (off >=sizeof (errlog)) { +return NVME_INVALID_FIELD |NVME_DNR ; } -if (off >sizeof (errlog)) { -return NVME_INVALID_FIELD |NVME_DNR ; +if (!rae) { +nvme_clear_events (n,NVME_AER_TYPE_ERROR ); }memset (&errlog,0x0 ,sizeof (errlog)); - trans_len =MIN (sizeof (errlog) - off, buf_len); return nvme_dma(n, (uint8_t *)&errlog, trans_len,Signature · 1 line
2.24.1
Signature · 1 line
One of us - No more doubt, silence or taboo about mental illness.
Re: [PATCH 2/9] hw/block/nvme: fix log page offset check
In reply to Keith BuschQuoted 8 lines · Keith Busch
Return error if the requested offset starts after the size of the log being returned. Also, move the check for earlier in the function so we're not doing unnecessary calculations.
Signed-off-by Keith Busch <[email protected]>--- hw/block/nvme.c | 22 ++++++++++------------ 1 file changed, 10 insertions(+), 12 deletions(-)
PATCH 3/9 hw/block/nvme: support per-namespace smart log
In reply to Keith BuschLet the user specify a specific namespace if they want to get access stats for a specific namespace.
--- hw/block/nvme.c | 66 +++++++++++++++++++++++++++----------------- include/block/nvme.h | 1 + 2 files changed, 41 insertions(+), 26 deletions(-)
hw/block/nvme.c +40 -26
@@ -1164,48 +1164,62 @@ static uint16_t nvme_create_sq(NvmeCtrl *n, NvmeRequest *req)return NVME_SUCCESS ; } +struct nvme_stats { +uint64_t units_read; +uint64_t units_written; +uint64_t read_commands; +uint64_t write_commands; +}; + +static void nvme_set_blk_stats (NvmeNamespace *ns,struct nvme_stats *stats) +{ +BlockAcctStats *s =blk_get_stats (ns->blkconf.blk); + + stats->units_read += s->nr_bytes[BLOCK_ACCT_READ ] >>BDRV_SECTOR_BITS ; + stats->units_written += s->nr_bytes[BLOCK_ACCT_WRITE ] >>BDRV_SECTOR_BITS ; + stats->read_commands += s->nr_ops[BLOCK_ACCT_READ ]; + stats->write_commands += s->nr_ops[BLOCK_ACCT_WRITE ]; +} +static uint16_t nvme_smart_info (NvmeCtrl *n,uint8_t rae,uint32_t buf_len,uint64_t off,NvmeRequest *req) {uint32_t nsid =le32_to_cpu (req->cmd.nsid); - +struct nvme_stats stats = {0 }; +NvmeSmartLog smart = {0 };uint32_t trans_len; +NvmeNamespace *ns;time_t current_ms; -uint64_t units_read =0 , units_written =0 ; -uint64_t read_commands =0 , write_commands =0 ; -NvmeSmartLog smart; - -if (nsid && nsid !=0xffffffff ) { -return NVME_INVALID_FIELD |NVME_DNR ; - }if (off >=sizeof (smart)) {return NVME_INVALID_FIELD |NVME_DNR ; } -for (int i =1 ; i <= n->num_namespaces; i++) { -NvmeNamespace *ns =nvme_ns (n, i); -if (!ns) { -continue ; - } - -BlockAcctStats *s =blk_get_stats (ns->blkconf.blk); +if (nsid !=0xffffffff ) { + ns =nvme_ns (n, nsid); +if (!ns) +return NVME_INVALID_NSID |NVME_DNR ; +nvme_set_blk_stats (ns, &stats); + }else { +int i; - units_read += s->nr_bytes[BLOCK_ACCT_READ ] >>BDRV_SECTOR_BITS ; - units_written += s->nr_bytes[BLOCK_ACCT_WRITE ] >>BDRV_SECTOR_BITS ; - read_commands += s->nr_ops[BLOCK_ACCT_READ ]; - write_commands += s->nr_ops[BLOCK_ACCT_WRITE ]; +for (i =1 ; i <= n->num_namespaces; i++) { + ns =nvme_ns (n, i); +if (!ns) { +continue ; + } +nvme_set_blk_stats (ns, &stats); + } } trans_len =MIN (sizeof (smart) - off, buf_len); -memset (&smart,0x0 ,sizeof (smart)); - - smart.data_units_read[0 ] =cpu_to_le64 (DIV_ROUND_UP (units_read,1000 )); - smart.data_units_written[0 ] =cpu_to_le64 (DIV_ROUND_UP (units_written, + smart.data_units_read[0 ] =cpu_to_le64 (DIV_ROUND_UP (stats.units_read, +1000 )); + smart.data_units_written[0 ] =cpu_to_le64 (DIV_ROUND_UP (stats.units_written,1000 )); - smart.host_read_commands[0 ] =cpu_to_le64 (read_commands); - smart.host_write_commands[0 ] =cpu_to_le64 (write_commands); + smart.host_read_commands[0 ] =cpu_to_le64 (stats.read_commands); + smart.host_write_commands[0 ] =cpu_to_le64 (stats.write_commands); smart.temperature =cpu_to_le16 (n->temperature);
@@ -2708,7 +2722,7 @@ static void nvme_init_ctrl(NvmeCtrl *n, PCIDevice *pci_dev) id->acl =3 ; id->aerl = n->params.aerl; id->frmw = (NVME_NUM_FW_SLOTS <<1 ) |NVME_FRMW_SLOT1_RO ; - id->lpa =NVME_LPA_EXTENDED ; + id->lpa =NVME_LPA_NS_SMART |NVME_LPA_EXTENDED ;/* recommended default value (~70 C) */ id->wctemp =cpu_to_le16 (NVME_TEMPERATURE_WARNING );
include/block/nvme.h +1 -0
@@ -849,6 +849,7 @@ enum NvmeIdCtrlFrmw { };enum NvmeIdCtrlLpa { +NVME_LPA_NS_SMART =1 <<0 ,NVME_LPA_EXTENDED =1 <<2 , };
Signature · 1 line
2.24.1
Re: [PATCH 3/9] hw/block/nvme: support per-namespace smart log
Quoted 3 lines · Keith Busch
Let the user specify a specific namespace if they want to get access stats for a specific namespace.
I don't think this makes sense for v1.3+.
NVM Express v1.3d, Section 5.14.1.2: "There is no namespace specific information defined in the SMART / Health log page in this revision of the specification. therefore the controller log page and namespace specific log page contain identical information".
I have no idea why the TWG decided this, but that's the way it is ;)
Quoted 123 lines
Signed-off-by Keith Busch <[email protected]>--- hw/block/nvme.c | 66 +++++++++++++++++++++++++++----------------- include/block/nvme.h | 1 + 2 files changed, 41 insertions(+), 26 deletions(-)hw/block/nvme.c +40
-26@@ -1164,48 +1164,62 @@ static uint16_t nvme_create_sq(NvmeCtrl *n, NvmeRequest *req)return NVME_SUCCESS ; } +struct nvme_stats { +uint64_t units_read; +uint64_t units_written; +uint64_t read_commands; +uint64_t write_commands; +}; + +static void nvme_set_blk_stats (NvmeNamespace *ns,struct nvme_stats *stats) +{ +BlockAcctStats *s =blk_get_stats (ns->blkconf.blk); + + stats->units_read += s->nr_bytes[BLOCK_ACCT_READ ] >>BDRV_SECTOR_BITS ; + stats->units_written += s->nr_bytes[BLOCK_ACCT_WRITE ] >>BDRV_SECTOR_BITS ; + stats->read_commands += s->nr_ops[BLOCK_ACCT_READ ]; + stats->write_commands += s->nr_ops[BLOCK_ACCT_WRITE ]; +} +static uint16_t nvme_smart_info (NvmeCtrl *n,uint8_t rae,uint32_t buf_len,uint64_t off,NvmeRequest *req) {uint32_t nsid =le32_to_cpu (req->cmd.nsid); - +struct nvme_stats stats = {0 }; +NvmeSmartLog smart = {0 };uint32_t trans_len; +NvmeNamespace *ns;time_t current_ms; -uint64_t units_read =0 , units_written =0 ; -uint64_t read_commands =0 , write_commands =0 ; -NvmeSmartLog smart; - -if (nsid && nsid !=0xffffffff ) { -return NVME_INVALID_FIELD |NVME_DNR ; - }if (off >=sizeof (smart)) {return NVME_INVALID_FIELD |NVME_DNR ; } -for (int i =1 ; i <= n->num_namespaces; i++) { -NvmeNamespace *ns =nvme_ns (n, i); -if (!ns) { -continue ; - } - -BlockAcctStats *s =blk_get_stats (ns->blkconf.blk); +if (nsid !=0xffffffff ) { + ns =nvme_ns (n, nsid); +if (!ns) +return NVME_INVALID_NSID |NVME_DNR ; +nvme_set_blk_stats (ns, &stats); + }else { +int i; - units_read += s->nr_bytes[BLOCK_ACCT_READ ] >>BDRV_SECTOR_BITS ; - units_written += s->nr_bytes[BLOCK_ACCT_WRITE ] >>BDRV_SECTOR_BITS ; - read_commands += s->nr_ops[BLOCK_ACCT_READ ]; - write_commands += s->nr_ops[BLOCK_ACCT_WRITE ]; +for (i =1 ; i <= n->num_namespaces; i++) { + ns =nvme_ns (n, i); +if (!ns) { +continue ; + } +nvme_set_blk_stats (ns, &stats); + } } trans_len =MIN (sizeof (smart) - off, buf_len); -memset (&smart,0x0 ,sizeof (smart)); - - smart.data_units_read[0 ] =cpu_to_le64 (DIV_ROUND_UP (units_read,1000 )); - smart.data_units_written[0 ] =cpu_to_le64 (DIV_ROUND_UP (units_written, + smart.data_units_read[0 ] =cpu_to_le64 (DIV_ROUND_UP (stats.units_read, +1000 )); + smart.data_units_written[0 ] =cpu_to_le64 (DIV_ROUND_UP (stats.units_written,1000 )); - smart.host_read_commands[0 ] =cpu_to_le64 (read_commands); - smart.host_write_commands[0 ] =cpu_to_le64 (write_commands); + smart.host_read_commands[0 ] =cpu_to_le64 (stats.read_commands); + smart.host_write_commands[0 ] =cpu_to_le64 (stats.write_commands); smart.temperature =cpu_to_le16 (n->temperature);@@ -2708,7 +2722,7 @@ static void nvme_init_ctrl(NvmeCtrl *n, PCIDevice *pci_dev) id->acl =3 ; id->aerl = n->params.aerl; id->frmw = (NVME_NUM_FW_SLOTS <<1 ) |NVME_FRMW_SLOT1_RO ; - id->lpa =NVME_LPA_EXTENDED ; + id->lpa =NVME_LPA_NS_SMART |NVME_LPA_EXTENDED ;/* recommended default value (~70 C) */ id->wctemp =cpu_to_le16 (NVME_TEMPERATURE_WARNING );include/block/nvme.h +1
-0@@ -849,6 +849,7 @@ enum NvmeIdCtrlFrmw { };enum NvmeIdCtrlLpa { +NVME_LPA_NS_SMART =1 <<0 ,NVME_LPA_EXTENDED =1 <<2 , };Signature · 1 line
2.24.1
Signature · 1 line
One of us - No more doubt, silence or taboo about mental illness.
Re: [PATCH 3/9] hw/block/nvme: support per-namespace smart log
Quoted 5 lines · Klaus Jensen
Quoted 3 lines · Keith Busch
Let the user specify a specific namespace if they want to get access stats for a specific namespace.
Quoted 8 lines
I don't think this makes sense for v1.3+.
NVM Express v1.3d, Section 5.14.1.2: "There is no namespace specific information defined in the SMART / Health log page in this revision of the specification. therefore the controller log page and namespace specific log page contain identical information".
I have no idea why the TWG decided this, but that's the way it is ;)
I don't think they did that. The behavior you're referring to is specific to controllers that operate a particular way: "If the log page is not supported on a per namespace basis ...". They were trying to provide a spec compliant way for controllers to return a success status if you supplied a valid NSID when the controller doesn't support per-namespace smart data..
The previous paragraph is more clear on this: "The controller may also support requesting the log page on a per namespace basis, as indicated by bit 0 of the LPA field in the Identify Controller data structure".
Re: [PATCH 3/9] hw/block/nvme: support per-namespace smart log
Quoted 21 lines · Keith Busch
Quoted 5 lines · Klaus Jensen
Quoted 3 lines · Keith Busch
Let the user specify a specific namespace if they want to get access stats for a specific namespace.
Quoted 8 lines
I don't think this makes sense for v1.3+.
NVM Express v1.3d, Section 5.14.1.2: "There is no namespace specific information defined in the SMART / Health log page in this revision of the specification. therefore the controller log page and namespace specific log page contain identical information".
I have no idea why the TWG decided this, but that's the way it is ;)
I don't think they did that. The behavior you're referring to is specific to controllers that operate a particular way: "If the log page is not supported on a per namespace basis ...". They were trying to provide a spec compliant way for controllers to return a success status if you supplied a valid NSID when the controller doesn't support per-namespace smart data..
Quoted 3 lines
The previous paragraph is more clear on this: "The controller may also support requesting the log page on a per namespace basis, as indicated by bit 0 of the LPA field in the Identify Controller data structure".
OK, so I agree that it makes sense for it to be supported on a per namespace basis, but I think the spec is just keeping the door open for future namespace specific stuff in the log page - currently there is none.
Figure 94 (the actual SMART log page) says that the Data Units Read/Written are controller wide, so there really is no namespace specific information. Maybe this could be in the context of shared namespaces? How would a controller know how much data has been read/written from it without asking the other controllers? What if a controller is detached from the namespace - you'd lose those numbers.
Re: [PATCH 3/9] hw/block/nvme: support per-namespace smart log
Quoted 11 lines · Klaus Jensen
OK, so I agree that it makes sense for it to be supported on a per namespace basis, but I think the spec is just keeping the door open for future namespace specific stuff in the log page - currently there is none.
Figure 94 (the actual SMART log page) says that the Data Units Read/Written are controller wide, so there really is no namespace specific information. Maybe this could be in the context of shared namespaces? How would a controller know how much data has been read/written from it without asking the other controllers? What if a controller is detached from the namespace - you'd lose those numbers.
That text is wrong. There is no "controller" scope to the smart log. Figure 191 says the smart scope is to the subsystem or the namespace. It doesn't matter which controller performed an IO to a particular namespace; the log needs to report the same information regardless of which controller you query. How that is coordinated within the subsystem is a detail not defined by spec.
Not that that particular detail matters here, as we don't support multi-controller subsystems (yet!). But the smart log text has missed an update to reflect this, so it looks like trivial ECN material to me.
Klaus Jensen
Reviewed-byRe: [PATCH 3/9] hw/block/nvme: support per-namespace smart log
Quoted 13 lines · Keith Busch
Quoted 11 lines · Klaus Jensen
OK, so I agree that it makes sense for it to be supported on a per namespace basis, but I think the spec is just keeping the door open for future namespace specific stuff in the log page - currently there is none.
Figure 94 (the actual SMART log page) says that the Data Units Read/Written are controller wide, so there really is no namespace specific information. Maybe this could be in the context of shared namespaces? How would a controller know how much data has been read/written from it without asking the other controllers? What if a controller is detached from the namespace - you'd lose those numbers.
Quoted 7 lines
That text is wrong. There is no "controller" scope to the smart log. Figure 191 says the smart scope is to the subsystem or the namespace. It doesn't matter which controller performed an IO to a particular namespace; the log needs to report the same information regardless of which controller you query. How that is coordinated within the subsystem is a detail not defined by spec.
Oh! Thats new in v1.4. So they updated that, but forgot figure 194. In v1.3 it is controller and namespace scope.
Anyway, fair enough. In that case,
Re: [PATCH 3/9] hw/block/nvme: support per-namespace smart log
In reply to Keith BuschQuoted 36 lines · Keith Busch
Let the user specify a specific namespace if they want to get access stats for a specific namespace.
Signed-off-by Keith Busch <[email protected]>--- hw/block/nvme.c | 66 +++++++++++++++++++++++++++----------------- include/block/nvme.h | 1 + 2 files changed, 41 insertions(+), 26 deletions(-)hw/block/nvme.c +3
-9static uint16_t nvme_smart_info (NvmeCtrl *n,uint8_t rae,uint32_t buf_len,uint64_t off,NvmeRequest *req) {uint32_t nsid =le32_to_cpu (req->cmd.nsid); - +struct nvme_stats stats = {0 }; +NvmeSmartLog smart = {0 };uint32_t trans_len; +NvmeNamespace *ns;time_t current_ms; -uint64_t units_read =0 , units_written =0 ; -uint64_t read_commands =0 , write_commands =0 ; -NvmeSmartLog smart; - -if (nsid && nsid !=0xffffffff ) { -return NVME_INVALID_FIELD |NVME_DNR ; - }if (off >=sizeof (smart)) {return NVME_INVALID_FIELD |NVME_DNR ; } -for (int i =1 ; i <= n->num_namespaces; i++) {
hw/block/nvme.c static uint16_t nvme_create_sq(NvmeCtrl *n, NvmeRequest *req) · line 1201
Quoted 10 lines
hw/block/nvme.c +4
-6-NvmeNamespace *ns =nvme_ns (n, i); -if (!ns) { -continue ; - } - -BlockAcctStats *s =blk_get_stats (ns->blkconf.blk); +if (nsid !=0xffffffff ) { + ns =nvme_ns (n, nsid); +if (!ns) +return NVME_INVALID_NSID |NVME_DNR ;
Btw, this is failing style check (missing braces).
RE: [PATCH 3/9] hw/block/nvme: support per-namespace smart log
In reply to Keith BuschQuoted 72 lines
-----Original Message-----
From: Keith Busch <[email protected]>
Sent: Wednesday, September 30, 2020 6:04 PM
To: [email protected]; [email protected]; Klaus Jensen <[email protected]>
Cc: Niklas Cassel <[email protected]>; Dmitry Fomichev <[email protected]>; Kevin Wolf <[email protected]>; Philippe Mathieu-Daudé <[email protected]>; Keith Busch <[email protected]>
Subject: [PATCH 3/9] hw/block/nvme: support per-namespace smart logLet the user specify a specific namespace if they want to get access stats for a specific namespace.
Signed-off-by Keith Busch <[email protected]>--- hw/block/nvme.c | 66 +++++++++++++++++++++++++++----------------- include/block/nvme.h | 1 + 2 files changed, 41 insertions(+), 26 deletions(-)hw/block/nvme.c +20
-9@@ -1164,48 +1164,62 @@ static uint16_t nvme_create_sq(NvmeCtrl *n, NvmeRequest *req)return NVME_SUCCESS ; } +struct nvme_stats { +uint64_t units_read; +uint64_t units_written; +uint64_t read_commands; +uint64_t write_commands; +}; + +static void nvme_set_blk_stats (NvmeNamespace *ns,struct nvme_stats *stats) +{ +BlockAcctStats *s =blk_get_stats (ns->blkconf.blk); + + stats->units_read += s->nr_bytes[BLOCK_ACCT_READ ] >>BDRV_SECTOR_BITS ; + stats->units_written += s->nr_bytes[BLOCK_ACCT_WRITE ] >>BDRV_SECTOR_BITS ; + stats->read_commands += s->nr_ops[BLOCK_ACCT_READ ]; + stats->write_commands += s->nr_ops[BLOCK_ACCT_WRITE ]; +} +static uint16_t nvme_smart_info (NvmeCtrl *n,uint8_t rae,uint32_t buf_len,uint64_t off,NvmeRequest *req) {uint32_t nsid =le32_to_cpu (req->cmd.nsid); - +struct nvme_stats stats = {0 }; +NvmeSmartLog smart = {0 };uint32_t trans_len; +NvmeNamespace *ns;time_t current_ms; -uint64_t units_read =0 , units_written =0 ; -uint64_t read_commands =0 , write_commands =0 ; -NvmeSmartLog smart; - -if (nsid && nsid !=0xffffffff ) { -return NVME_INVALID_FIELD |NVME_DNR ; - }if (off >=sizeof (smart)) {return NVME_INVALID_FIELD |NVME_DNR ; } -for (int i =1 ; i <= n->num_namespaces; i++) {
hw/block/nvme.c static uint16_t nvme_create_sq(NvmeCtrl *n, · line 1206
Quoted 10 lines
hw/block/nvme.c +4
-6-NvmeNamespace *ns =nvme_ns (n, i); -if (!ns) { -continue ; - } - -BlockAcctStats *s =blk_get_stats (ns->blkconf.blk); +if (nsid !=0xffffffff ) { + ns =nvme_ns (n, nsid); +if (!ns) +return NVME_INVALID_NSID |NVME_DNR ;
checkpatch will complain about missing curly braces here
Quoted 64 lines
hw/block/nvme.c +10
-7+nvme_set_blk_stats (ns, &stats); + }else { +int i; - units_read += s->nr_bytes[BLOCK_ACCT_READ ] >>BDRV_SECTOR_BITS ; - units_written += s->nr_bytes[BLOCK_ACCT_WRITE ] >>BDRV_SECTOR_BITS ; - read_commands += s->nr_ops[BLOCK_ACCT_READ ]; - write_commands += s->nr_ops[BLOCK_ACCT_WRITE ]; +for (i =1 ; i <= n->num_namespaces; i++) { + ns =nvme_ns (n, i); +if (!ns) { +continue ; + } +nvme_set_blk_stats (ns, &stats); + } } trans_len =MIN (sizeof (smart) - off, buf_len); -memset (&smart,0x0 ,sizeof (smart)); - - smart.data_units_read[0] = cpu_to_le64(DIV_ROUND_UP(units_read,1000));
- smart.data_units_written[0] =
cpu_to_le64(DIV_ROUND_UP(units_written,
+ smart.data_units_read[0] =
cpu_to_le64(DIV_ROUND_UP(stats.units_read,hw/block/nvme.c +2
-0+ 1000)); + smart.data_units_written[0] =cpu_to_le64(DIV_ROUND_UP(stats.units_written,
1000));hw/block/nvme.c +3
-3- smart.host_read_commands[0 ] =cpu_to_le64 (read_commands); - smart.host_write_commands[0 ] =cpu_to_le64 (write_commands); + smart.host_read_commands[0 ] =cpu_to_le64 (stats.read_commands); + smart.host_write_commands[0 ] =cpu_to_le64 (stats.write_commands); smart.temperature =cpu_to_le16 (n->temperature);@@ -2708,7 +2722,7 @@ static void nvme_init_ctrl(NvmeCtrl *n, PCIDevice *pci_dev) id->acl =3 ; id->aerl = n->params.aerl; id->frmw = (NVME_NUM_FW_SLOTS <<1 ) |NVME_FRMW_SLOT1_RO ; - id->lpa =NVME_LPA_EXTENDED ; + id->lpa =NVME_LPA_NS_SMART |NVME_LPA_EXTENDED ;/* recommended default value (~70 C) */ id->wctemp = cpu_to_le16(NVME_TEMPERATURE_WARNING);include/block/nvme.h +1
-1@@ -849,6 +849,7 @@ enum NvmeIdCtrlFrmw { };enum NvmeIdCtrlLpa { +NVME_LPA_NS_SMART =1 <<0 ,NVME_LPA_EXTENDED =1 <<2 , }; --2.24.1
PATCH 4/9 hw/block/nvme: validate command set selected
In reply to Keith BuschFail to start the controller if the user requests a command set that the controller does not support.
--- hw/block/nvme.c | 6 +++++- hw/block/trace-events | 1 + include/block/nvme.h | 4 ++++ 3 files changed, 10 insertions(+), 1 deletion(-)
hw/block/nvme.c +5 -1
@@ -2049,6 +2049,10 @@ static int nvme_start_ctrl(NvmeCtrl *n)trace_pci_nvme_err_startfail_acq_misaligned (n->bar.acq);return -1 ; } +if (unlikely (!(NVME_CAP_CSS (n->bar.cap) & (1 <<NVME_CC_CSS (n->bar.cc))))) { +trace_pci_nvme_err_startfail_css (NVME_CC_CSS (n->bar.cc)); +return -1 ; + }if (unlikely (NVME_CC_MPS (n->bar.cc) <NVME_CAP_MPSMIN (n->bar.cap))) { trace_pci_nvme_err_startfail_page_too_small(
@@ -2750,7 +2754,7 @@ static void nvme_init_ctrl(NvmeCtrl *n, PCIDevice *pci_dev)NVME_CAP_SET_MQES (n->bar.cap,0x7ff );NVME_CAP_SET_CQR (n->bar.cap,1 );NVME_CAP_SET_TO (n->bar.cap,0xf ); -NVME_CAP_SET_CSS (n->bar.cap,1 ); +NVME_CAP_SET_CSS (n->bar.cap,NVME_CAP_CSS_NVM );NVME_CAP_SET_MPSMAX (n->bar.cap,4 ); n->bar.vs =NVME_SPEC_VER ;
hw/block/trace-events +1 -0
@@ -133,6 +133,7 @@ pci_nvme_err_startfail_cqent_too_small(uint8_t log2ps, uint8_t maxlog2ps) "nvme_ pci_nvme_err_startfail_cqent_too_large(uint8_t log2ps, uint8_t maxlog2ps) "nvme_start_ctrl failed because the completion queue entry size is too large: log2size=%u, max=%u" pci_nvme_err_startfail_sqent_too_small(uint8_t log2ps, uint8_t maxlog2ps) "nvme_start_ctrl failed because the submission queue entry size is too small: log2size=%u, min=%u" pci_nvme_err_startfail_sqent_too_large(uint8_t log2ps, uint8_t maxlog2ps) "nvme_start_ctrl failed because the submission queue entry size is too large: log2size=%u, max=%u" +pci_nvme_err_startfail_css(uint8_t css) "nvme_start_ctrl failed because invalid command set selected:%u" pci_nvme_err_startfail_asqent_sz_zero(void) "nvme_start_ctrl failed because the admin submission queue size is zero" pci_nvme_err_startfail_acqent_sz_zero(void) "nvme_start_ctrl failed because the admin completion queue size is zero" pci_nvme_err_startfail(void) "setting controller enable bit failed"
include/block/nvme.h +4 -0
@@ -82,6 +82,10 @@ enum NvmeCapMask {#define NVME_CAP_SET_PMRS (cap, val) (cap |= (uint64_t)(val & CAP_PMR_MASK)\ << CAP_PMR_SHIFT) +enum NvmeCapCss { +NVME_CAP_CSS_NVM =1 <<0 , +}; +enum NvmeCcShift {CC_EN_SHIFT =0 ,CC_CSS_SHIFT =4 ,
Signature · 1 line
2.24.1
Re: [PATCH 4/9] hw/block/nvme: validate command set selected
Quoted 4 lines · Keith Busch
Fail to start the controller if the user requests a command set that the controller does not support.
Signed-off-by Keith Busch <[email protected]>
Quoted 61 lines
--- hw/block/nvme.c | 6 +++++- hw/block/trace-events | 1 + include/block/nvme.h | 4 ++++ 3 files changed, 10 insertions(+), 1 deletion(-)hw/block/nvme.c +5
-1@@ -2049,6 +2049,10 @@ static int nvme_start_ctrl(NvmeCtrl *n)trace_pci_nvme_err_startfail_acq_misaligned (n->bar.acq);return -1 ; } +if (unlikely (!(NVME_CAP_CSS (n->bar.cap) & (1 <<NVME_CC_CSS (n->bar.cc))))) { +trace_pci_nvme_err_startfail_css (NVME_CC_CSS (n->bar.cc)); +return -1 ; + }if (unlikely (NVME_CC_MPS (n->bar.cc) <NVME_CAP_MPSMIN (n->bar.cap))) { trace_pci_nvme_err_startfail_page_too_small(@@ -2750,7 +2754,7 @@ static void nvme_init_ctrl(NvmeCtrl *n, PCIDevice *pci_dev)NVME_CAP_SET_MQES (n->bar.cap,0x7ff );NVME_CAP_SET_CQR (n->bar.cap,1 );NVME_CAP_SET_TO (n->bar.cap,0xf ); -NVME_CAP_SET_CSS (n->bar.cap,1 ); +NVME_CAP_SET_CSS (n->bar.cap,NVME_CAP_CSS_NVM );NVME_CAP_SET_MPSMAX (n->bar.cap,4 ); n->bar.vs =NVME_SPEC_VER ;hw/block/trace-events +1
-0@@ -133,6 +133,7 @@ pci_nvme_err_startfail_cqent_too_small(uint8_t log2ps, uint8_t maxlog2ps) "nvme_ pci_nvme_err_startfail_cqent_too_large(uint8_t log2ps, uint8_t maxlog2ps) "nvme_start_ctrl failed because the completion queue entry size is too large: log2size=%u, max=%u" pci_nvme_err_startfail_sqent_too_small(uint8_t log2ps, uint8_t maxlog2ps) "nvme_start_ctrl failed because the submission queue entry size is too small: log2size=%u, min=%u" pci_nvme_err_startfail_sqent_too_large(uint8_t log2ps, uint8_t maxlog2ps) "nvme_start_ctrl failed because the submission queue entry size is too large: log2size=%u, max=%u" +pci_nvme_err_startfail_css(uint8_t css) "nvme_start_ctrl failed because invalid command set selected:%u" pci_nvme_err_startfail_asqent_sz_zero(void) "nvme_start_ctrl failed because the admin submission queue size is zero" pci_nvme_err_startfail_acqent_sz_zero(void) "nvme_start_ctrl failed because the admin completion queue size is zero" pci_nvme_err_startfail(void) "setting controller enable bit failed"include/block/nvme.h +4
-0@@ -82,6 +82,10 @@ enum NvmeCapMask {#define NVME_CAP_SET_PMRS (cap, val) (cap |= (uint64_t)(val & CAP_PMR_MASK)\ << CAP_PMR_SHIFT) +enum NvmeCapCss { +NVME_CAP_CSS_NVM =1 <<0 , +}; +enum NvmeCcShift {CC_EN_SHIFT =0 ,CC_CSS_SHIFT =4 ,Signature · 1 line
2.24.1
Signature · 1 line
One of us - No more doubt, silence or taboo about mental illness.
PATCH 5/9 hw/block/nvme: support for admin-only command set
In reply to Keith Busch--- hw/block/nvme.c | 1 + include/block/nvme.h | 3 ++- 2 files changed, 3 insertions(+), 1 deletion(-)
hw/block/nvme.c +1 -0
@@ -2755,6 +2755,7 @@ static void nvme_init_ctrl(NvmeCtrl *n, PCIDevice *pci_dev)NVME_CAP_SET_CQR (n->bar.cap,1 );NVME_CAP_SET_TO (n->bar.cap,0xf );NVME_CAP_SET_CSS (n->bar.cap,NVME_CAP_CSS_NVM ); +NVME_CAP_SET_CSS (n->bar.cap,NVME_CAP_CSS_ADMIN_ONLY );NVME_CAP_SET_MPSMAX (n->bar.cap,4 ); n->bar.vs =NVME_SPEC_VER ;
include/block/nvme.h +2 -1
@@ -83,7 +83,8 @@ enum NvmeCapMask { << CAP_PMR_SHIFT)enum NvmeCapCss { -NVME_CAP_CSS_NVM =1 <<0 , +NVME_CAP_CSS_NVM =1 <<0 , +NVME_CAP_CSS_ADMIN_ONLY =1 <<7 , };enum NvmeCcShift {
Signature · 1 line
2.24.1
RE: [PATCH 5/9] hw/block/nvme: support for admin-only command set
Quoted 21 lines
-----Original Message-----
From: Keith Busch <[email protected]>
Sent: Wednesday, September 30, 2020 6:04 PM
To: [email protected]; [email protected]; Klaus Jensen <[email protected]>
Cc: Niklas Cassel <[email protected]>; Dmitry Fomichev <[email protected]>; Kevin Wolf <[email protected]>; Philippe Mathieu-Daudé <[email protected]>; Keith Busch <[email protected]>
Subject: [PATCH 5/9] hw/block/nvme: support for admin-only command setSigned-off-by Keith Busch <[email protected]>--- hw/block/nvme.c | 1 + include/block/nvme.h | 3 ++- 2 files changed, 3 insertions(+), 1 deletion(-)hw/block/nvme.c +0
-0
hw/block/nvme.c static void nvme_init_ctrl(NvmeCtrl *n, PCIDevice · line 2759
Quoted 5 lines
hw/block/nvme.c +1
-0@@ -2755,6 +2755,7 @@ static void nvme_init_ctrl(NvmeCtrl *n, PCIDevice *pci_dev)NVME_CAP_SET_CQR (n->bar.cap,1 );NVME_CAP_SET_TO (n->bar.cap,0xf );NVME_CAP_SET_CSS (n->bar.cap,NVME_CAP_CSS_NVM ); +NVME_CAP_SET_CSS (n->bar.cap,NVME_CAP_CSS_ADMIN_ONLY );
This could be
- NVME_CAP_SET_CSS(n->bar.cap, NVME_CAP_CSS_NVM);
+ NVME_CAP_SET_CSS(n->bar.cap, (NVME_CAP_CSS_NVM | NVME_CAP_CSS_ADMIN_ONLY));
Unfortunately, parentheses are needed above because NVME_CAP_SET_CSS macro and other similar macros use "val" instead of (val). A possible cleanup topic...
Quoted 19 lines
hw/block/nvme.c +0
-0NVME_CAP_SET_MPSMAX (n->bar.cap,4 );n->bar.vs = NVME_SPEC_VER;include/block/nvme.h +2
-2@@ -83,7 +83,8 @@ enum NvmeCapMask { << CAP_PMR_SHIFT)enum NvmeCapCss { -NVME_CAP_CSS_NVM =1 <<0 , +NVME_CAP_CSS_NVM =1 <<0 , +NVME_CAP_CSS_ADMIN_ONLY =1 <<7 , };enum NvmeCcShift { --2.24.1
Re: [PATCH 5/9] hw/block/nvme: support for admin-only command set
In reply to Keith BuschQuoted 2 lines · Keith Busch
Signed-off-by Keith Busch <[email protected]>---
Quoted 34 lines
hw/block/nvme.c | 1 + include/block/nvme.h | 3 ++- 2 files changed, 3 insertions(+), 1 deletion(-)hw/block/nvme.c +1
-0@@ -2755,6 +2755,7 @@ static void nvme_init_ctrl(NvmeCtrl *n, PCIDevice *pci_dev)NVME_CAP_SET_CQR (n->bar.cap,1 );NVME_CAP_SET_TO (n->bar.cap,0xf );NVME_CAP_SET_CSS (n->bar.cap,NVME_CAP_CSS_NVM ); +NVME_CAP_SET_CSS (n->bar.cap,NVME_CAP_CSS_ADMIN_ONLY );NVME_CAP_SET_MPSMAX (n->bar.cap,4 ); n->bar.vs =NVME_SPEC_VER ;include/block/nvme.h +2
-1@@ -83,7 +83,8 @@ enum NvmeCapMask { << CAP_PMR_SHIFT)enum NvmeCapCss { -NVME_CAP_CSS_NVM =1 <<0 , +NVME_CAP_CSS_NVM =1 <<0 , +NVME_CAP_CSS_ADMIN_ONLY =1 <<7 , };enum NvmeCcShift {Signature · 1 line
2.24.1
Signature · 1 line
One of us - No more doubt, silence or taboo about mental illness.
PATCH 6/9 hw/block/nvme: reject io commands if only admin command set selected
In reply to Keith BuschFrom: Klaus Jensen <[email protected]>
If the host sets CC.CSS to 111b, all commands submitted to I/O queues should be completed with status Invalid Command Opcode.
Note that this is technically a v1.4 feature, but it does not hurt to implement before we finally bump the reported version implemented.
--- hw/block/nvme.c | 4 ++++ include/block/nvme.h | 5 +++++ 2 files changed, 9 insertions(+)
hw/block/nvme.c +4 -0
@@ -1026,6 +1026,10 @@ static uint16_t nvme_io_cmd(NvmeCtrl *n, NvmeRequest *req)trace_pci_nvme_io_cmd (nvme_cid (req), nsid,nvme_sqid (req), req->cmd.opcode,nvme_io_opc_str (req->cmd.opcode)); +if (NVME_CC_CSS (n->bar.cc) ==NVME_CC_CSS_ADMIN_ONLY ) { +return NVME_INVALID_OPCODE |NVME_DNR ; + } +if (!nvme_nsid_valid (n, nsid)) {return NVME_INVALID_NSID |NVME_DNR ; }
include/block/nvme.h +5 -0
@@ -115,6 +115,11 @@ enum NvmeCcMask {#define NVME_CC_IOSQES (cc) ((cc >> CC_IOSQES_SHIFT) & CC_IOSQES_MASK)#define NVME_CC_IOCQES (cc) ((cc >> CC_IOCQES_SHIFT) & CC_IOCQES_MASK) +enum NvmeCcCss { +NVME_CC_CSS_NVM =0x0 , +NVME_CC_CSS_ADMIN_ONLY =0x7 , +}; +enum NvmeCstsShift {CSTS_RDY_SHIFT =0 ,CSTS_CFS_SHIFT =1 ,
Signature · 1 line
2.24.1
RE: [PATCH 6/9] hw/block/nvme: reject io commands if only admin command set selected
Quoted 13 lines
-----Original Message-----
From: Keith Busch <[email protected]>
Sent: Wednesday, September 30, 2020 6:04 PM
To: [email protected]; [email protected]; Klaus Jensen <[email protected]>
Cc: Niklas Cassel <[email protected]>; Dmitry Fomichev <[email protected]>; Kevin Wolf <[email protected]>; Philippe Mathieu-Daudé <[email protected]>; Keith Busch <[email protected]>
Subject: [PATCH 6/9] hw/block/nvme: reject io commands if only admin command set selectedFrom: Klaus Jensen <[email protected]>
Quoted 8 lines
If the host sets CC.CSS to 111b, all commands submitted to I/O queues should be completed with status Invalid Command Opcode.
Note that this is technically a v1.4 feature, but it does not hurt to implement before we finally bump the reported version implemented.
Signed-off-by Klaus Jensen <[email protected]>Signed-off-by Keith Busch <[email protected]>
Quoted 41 lines
--- hw/block/nvme.c | 4 ++++ include/block/nvme.h | 5 +++++ 2 files changed, 9 insertions(+)hw/block/nvme.c +4
-0@@ -1026,6 +1026,10 @@ static uint16_t nvme_io_cmd(NvmeCtrl *n, NvmeRequest *req)trace_pci_nvme_io_cmd (nvme_cid (req ),nsid ,nvme_sqid (req ), req->cmd.opcode, nvme_io_opc_str(req->cmd.opcode)); +if (NVME_CC_CSS (n->bar.cc) ==NVME_CC_CSS_ADMIN_ONLY ) { +return NVME_INVALID_OPCODE |NVME_DNR ; + } +if (!nvme_nsid_valid (n, nsid)) {return NVME_INVALID_NSID |NVME_DNR ;}include/block/nvme.h +5
-0@@ -115,6 +115,11 @@ enum NvmeCcMask {#define NVME_CC_IOSQES (cc) ((cc >> CC_IOSQES_SHIFT) & CC_IOSQES_MASK)#define NVME_CC_IOCQES (cc) ((cc >> CC_IOCQES_SHIFT) & CC_IOCQES_MASK) +enum NvmeCcCss { +NVME_CC_CSS_NVM =0x0 , +NVME_CC_CSS_ADMIN_ONLY =0x7 , +}; +enum NvmeCstsShift {
CSTS_RDY_SHIFT= 0,
CSTS_CFS_SHIFT= 1,
--
2.24.1
PATCH 7/9 hw/block/nvme: add nsid to get/setfeat trace events
In reply to Keith BuschFrom: Klaus Jensen <[email protected]>
Include the namespace id in the pci_nvme_{get,set}feat trace events.
--- hw/block/nvme.c | 4 ++-- hw/block/trace-events | 4 ++-- 2 files changed, 4 insertions(+), 4 deletions(-)
hw/block/nvme.c +2 -2
@@ -1643,7 +1643,7 @@ static uint16_t nvme_get_feature(NvmeCtrl *n, NvmeRequest *req) [NVME_ARBITRATION ] =NVME_ARB_AB_NOLIMIT , }; -trace_pci_nvme_getfeat (nvme_cid (req), fid, sel, dw11); +trace_pci_nvme_getfeat (nvme_cid (req), nsid, fid, sel, dw11);if (!nvme_feature_support[fid]) {return NVME_INVALID_FIELD |NVME_DNR ;
@@ -1781,7 +1781,7 @@ static uint16_t nvme_set_feature(NvmeCtrl *n, NvmeRequest *req)uint8_t fid =NVME_GETSETFEAT_FID (dw10);uint8_t save =NVME_SETFEAT_SAVE (dw10); -trace_pci_nvme_setfeat (nvme_cid (req), fid, save, dw11); +trace_pci_nvme_setfeat (nvme_cid (req), nsid, fid, save, dw11);if (save) {return NVME_FID_NOT_SAVEABLE |NVME_DNR ;
hw/block/trace-events +2 -2
@@ -53,8 +53,8 @@ pci_nvme_identify_ns(uint32_t ns) "nsid %"PRIu32"" pci_nvme_identify_nslist(uint32_t ns) "nsid %"PRIu32"" pci_nvme_identify_ns_descr_list(uint32_t ns) "nsid %"PRIu32"" pci_nvme_get_log(uint16_t cid, uint8_t lid, uint8_t lsp, uint8_t rae, uint32_t len, uint64_t off) "cid %"PRIu16" lid 0x%"PRIx8" lsp 0x%"PRIx8" rae 0x%"PRIx8" len %"PRIu32" off %"PRIu64"" -pci_nvme_getfeat(uint16_t cid, uint8_t fid, uint8_t sel, uint32_t cdw11) "cid %"PRIu16" fid 0x%"PRIx8" sel 0x%"PRIx8" cdw11 0x%"PRIx32"" -pci_nvme_setfeat(uint16_t cid, uint8_t fid, uint8_t save, uint32_t cdw11) "cid %"PRIu16" fid 0x%"PRIx8" save 0x%"PRIx8" cdw11 0x%"PRIx32"" +pci_nvme_getfeat(uint16_t cid, uint32_t nsid, uint8_t fid, uint8_t sel, uint32_t cdw11) "cid %"PRIu16" nsid 0x%"PRIx32" fid 0x%"PRIx8" sel 0x%"PRIx8" cdw11 0x%"PRIx32"" +pci_nvme_setfeat(uint16_t cid, uint32_t nsid, uint8_t fid, uint8_t save, uint32_t cdw11) "cid %"PRIu16" nsid 0x%"PRIx32" fid 0x%"PRIx8" save 0x%"PRIx8" cdw11 0x%"PRIx32"" pci_nvme_getfeat_vwcache(const char* result) "get feature volatile write cache, result=%s" pci_nvme_getfeat_numq(int result) "get feature number of queues, result=%d" pci_nvme_setfeat_numq(int reqcq, int reqsq, int gotcq, int gotsq) "requested cq_count=%d sq_count=%d, responding with cq_count=%d sq_count=%d"
Signature · 1 line
2.24.1
PATCH 8/9 hw/block/nvme: add trace event for requests with non-zero status code
In reply to Keith BuschFrom: Klaus Jensen <[email protected]>
If a command results in a non-zero status code, trace it.
--- hw/block/nvme.c | 6 ++++++ hw/block/trace-events | 1 + 2 files changed, 7 insertions(+)
hw/block/nvme.c +6 -0
@@ -777,6 +777,12 @@ static void nvme_enqueue_req_completion(NvmeCQueue *cq, NvmeRequest *req)assert (cq->cqid == req->sq->cqid);trace_pci_nvme_enqueue_req_completion (nvme_cid (req), cq->cqid, req->status); + +if (req->status) { +trace_pci_nvme_err_req_status (nvme_cid (req),nvme_nsid (req->ns), + req->status, req->cmd.opcode); + } +QTAILQ_REMOVE (&req->sq->out_req_list, req, entry);QTAILQ_INSERT_TAIL (&cq->req_list, req, entry);timer_mod (cq->timer,qemu_clock_get_ns (QEMU_CLOCK_VIRTUAL ) +500 );
hw/block/trace-events +1 -0
@@ -89,6 +89,7 @@ pci_nvme_mmio_shutdown_cleared(void) "shutdown bit cleared" # nvme traces for error conditions pci_nvme_err_mdts(uint16_t cid, size_t len) "cid %"PRIu16" len %zu" +pci_nvme_err_req_status(uint16_t cid, uint32_t nsid, uint16_t status, uint8_t opc) "cid %"PRIu16" nsid %"PRIu32" status 0x%"PRIx16" opc 0x%"PRIx8"" pci_nvme_err_addr_read(uint64_t addr) "addr 0x%"PRIx64"" pci_nvme_err_addr_write(uint64_t addr) "addr 0x%"PRIx64"" pci_nvme_err_cfs(void) "controller fatal status"
Signature · 1 line
2.24.1
RE: [PATCH 8/9] hw/block/nvme: add trace event for requests with non-zero status code
Quoted 27 lines
-----Original Message-----
From: Keith Busch <[email protected]>
Sent: Wednesday, September 30, 2020 6:04 PM
To: [email protected]; [email protected]; Klaus Jensen <[email protected]>
Cc: Niklas Cassel <[email protected]>; Dmitry Fomichev <[email protected]>; Kevin Wolf <[email protected]>; Philippe Mathieu-Daudé <[email protected]>; Keith Busch <[email protected]>
Subject: [PATCH 8/9] hw/block/nvme: add trace event for requests with non-zero status codeFrom: Klaus Jensen <[email protected]>
If a command results in a non-zero status code, trace it.
Signed-off-by Klaus Jensen <[email protected]>Signed-off-by Keith Busch <[email protected]>--- hw/block/nvme.c | 6 ++++++ hw/block/trace-events | 1 + 2 files changed, 7 insertions(+)hw/block/nvme.c +0
-0
hw/block/nvme.c static void · line 786
Quoted 10 lines
hw/block/nvme.c +6
-0@@ -777,6 +777,12 @@ static void nvme_enqueue_req_completion(NvmeCQueue *cq, NvmeRequest *req)assert (cq->cqid == req->sq->cqid);trace_pci_nvme_enqueue_req_completion (nvme_cid (req), cq->cqid, req->status); + +if (req->status) { +trace_pci_nvme_err_req_status (nvme_cid (req),nvme_nsid (req->ns), + req->status, req->cmd.opcode); + } +
Very useful.
Quoted 21 lines
hw/block/nvme.c +0
-0QTAILQ_REMOVE (&req->sq->out_req_list, req, entry);QTAILQ_INSERT_TAIL (&cq->req_list, req, entry);timer_mod(cq->timer, qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) +500);
hw/block/trace-events +0
-0@@ -89,6 +89,7 @@ pci_nvme_mmio_shutdown_cleared(void) "shutdownbit cleared"
# nvme traces for error conditions pci_nvme_err_mdts(uint16_t cid, size_t len) "cid %"PRIu16" len %zu" +pci_nvme_err_req_status(uint16_t cid, uint32_t nsid, uint16_t status, uint8_t opc) "cid %"PRIu16" nsid %"PRIu32" status 0x%"PRIx16" opc 0x%"PRIx8"" pci_nvme_err_addr_read(uint64_t addr) "addr 0x%"PRIx64"" pci_nvme_err_addr_write(uint64_t addr) "addr 0x%"PRIx64"" pci_nvme_err_cfs(void) "controller fatal status" -- 2.24.1
Re: [PATCH 8/9] hw/block/nvme: add trace event for requests with non-zero status code
In reply to Keith BuschQuoted 10 lines · Keith Busch
From: Klaus Jensen <[email protected]>
If a command results in a non-zero status code, trace it.
Signed-off-by Klaus Jensen <[email protected]>Signed-off-by Keith Busch <[email protected]>--- hw/block/nvme.c | 6 ++++++ hw/block/trace-events | 1 + 2 files changed, 7 insertions(+)
PATCH 9/9 hw/block/nvme: report actual LBA data shift in LBAF
In reply to Keith BuschFrom: Dmitry Fomichev <[email protected]>
Calculate the data shift value to report based on the set value of logical_block_size device property.
In the process, use a local variable to calculate the LBA format index instead of the hardcoded value 0. This makes the code more readable and it will make it easier to add support for multiple LBA formats in the future.
--- hw/block/nvme-ns.c | 5 +++++ 1 file changed, 5 insertions(+)
hw/block/nvme-ns.c +5 -0
@@ -47,6 +47,8 @@ static void nvme_ns_init(NvmeNamespace *ns)static int nvme_ns_init_blk (NvmeCtrl *n,NvmeNamespace *ns,Error **errp) { +int lba_index; +if (!blkconf_blocksizes (&ns->blkconf, errp)) {return -1 ; }
@@ -67,6 +69,9 @@ static int nvme_ns_init_blk(NvmeCtrl *n, NvmeNamespace *ns, Error **errp) n->features.vwc =0x1 ; } + lba_index =NVME_ID_NS_FLBAS_INDEX (ns->id_ns.flbas); + ns->id_ns.lbaf[lba_index].ds =31 -clz32 (ns->blkconf.logical_block_size); +return 0 ; }
Signature · 1 line
2.24.1
Re: [PATCH 9/9] hw/block/nvme: report actual LBA data shift in LBAF
Quoted 31 lines · Keith Busch
From: Dmitry Fomichev <[email protected]>
Calculate the data shift value to report based on the set value of
logical_block_sizedevice property.In the process, use a local variable to calculate the LBA format index instead of the hardcoded value 0. This makes the code more readable and it will make it easier to add support for multiple LBA formats in the future.
Signed-off-by Dmitry Fomichev <[email protected]>Reviewed-by Klaus Jensen <[email protected]>Signed-off-by Keith Busch <[email protected]>--- hw/block/nvme-ns.c | 5 +++++ 1 file changed, 5 insertions(+)hw/block/nvme-ns.c +2
-0@@ -47,6 +47,8 @@ static void nvme_ns_init(NvmeNamespace *ns)static int nvme_ns_init_blk (NvmeCtrl *n,NvmeNamespace *ns,Error **errp) { +int lba_index; +if (!blkconf_blocksizes (&ns->blkconf, errp)) {return -1 ; }
hw/block/nvme-ns.c static int nvme_ns_init_blk(NvmeCtrl *n, NvmeNamespace *ns, Error **errp) · line 77
Quoted 9 lines
hw/block/nvme-ns.c +3
-0@@ -67,6 +69,9 @@ static int nvme_ns_init_blk(NvmeCtrl *n, NvmeNamespace *ns, Error **errp) n->features.vwc =0x1 ; } + lba_index =NVME_ID_NS_FLBAS_INDEX (ns->id_ns.flbas); + ns->id_ns.lbaf[lba_index].ds =31 -clz32 (ns->blkconf.logical_block_size); +return 0 ; }
I think that back when I reviewed this, it was before the multi-namespace patch.
I believe this setup should move to the nvme_ns_init() function instead.
Quoted 10 lines · Keith Busch
After going through the zns enabling, I notice the controller enabling is not correct. Then I just continued maked more stuff. The series, I think, contains some of the less controversial patches from the two conflicting zns series, preceeded by some cleanups and fixes from me.
If this is all fine, I took the liberty of porting the zns enabling to it and made a public branch for consideration here:
http://git.infradead.org/qemu-nvme.git/shortlog/refs/heads/kb-zns
I rebased my patches on top of Keith's fixups as well. And in compliance with the concensus, I removed persistence.
Quoted 25 lines · Keith Busch
After going through the zns enabling, I notice the controller enabling is not correct. Then I just continued maked more stuff. The series, I think, contains some of the less controversial patches from the two conflicting zns series, preceeded by some cleanups and fixes from me.
If this is all fine, I took the liberty of porting the zns enabling to it and made a public branch for consideration here:
http://git.infradead.org/qemu-nvme.git/shortlog/refs/heads/kb-zns
Dmitry Fomichev (1):
hw/block/nvme: report actual LBA data shift in LBAFKeith Busch (5):
hw/block/nvme: remove pointless rw indirection hw/block/nvme: fix log page offset check hw/block/nvme: support per-namespace smart log hw/block/nvme: validate command set selected hw/block/nvme: support for admin-only command setKlaus Jensen (3):
hw/block/nvme: reject io commands if only admin command set selected hw/block/nvme: add nsid to get/setfeat trace events hw/block/nvme: add trace event for requests with non-zero status code
Quoted 10 lines
hw/block/nvme-ns.c | 5 ++ hw/block/nvme.c | 194 ++++++++++++++++++++---------------------- hw/block/trace-events | 6 +- include/block/nvme.h | 11 +++ 4 files changed, 114 insertions(+), 102 deletions(-)Signature · 1 line
2.24.1
These fixes all look good to me apart from the odd fixes that has been mentioned in the reviews. Since soft freeze is only two weeks away (Oct 27th), it would be nice to get this staged on nvme-next so we can get a pull sent off to Peter.
Quoted 37 lines · Klaus Jensen
Quoted 25 lines · Keith Busch
After going through the zns enabling, I notice the controller enabling is not correct. Then I just continued maked more stuff. The series, I think, contains some of the less controversial patches from the two conflicting zns series, preceeded by some cleanups and fixes from me.
If this is all fine, I took the liberty of porting the zns enabling to it and made a public branch for consideration here:
http://git.infradead.org/qemu-nvme.git/shortlog/refs/heads/kb-zns
Dmitry Fomichev (1):
hw/block/nvme: report actual LBA data shift in LBAFKeith Busch (5):
hw/block/nvme: remove pointless rw indirection hw/block/nvme: fix log page offset check hw/block/nvme: support per-namespace smart log hw/block/nvme: validate command set selected hw/block/nvme: support for admin-only command setKlaus Jensen (3):
hw/block/nvme: reject io commands if only admin command set selected hw/block/nvme: add nsid to get/setfeat trace events hw/block/nvme: add trace event for requests with non-zero status codeQuoted 10 lines
hw/block/nvme-ns.c | 5 ++ hw/block/nvme.c | 194 ++++++++++++++++++++---------------------- hw/block/trace-events | 6 +- include/block/nvme.h | 11 +++ 4 files changed, 114 insertions(+), 102 deletions(-)Signature · 1 line
2.24.1
Quoted 4 lines
These fixes all look good to me apart from the odd fixes that has been mentioned in the reviews. Since soft freeze is only two weeks away (Oct 27th), it would be nice to get this staged on nvme-next so we can get a pull sent off to Peter.
I've fixed up the comments mentioned and added the received reviews. Since it was pretty trivial fixups and passes my basic santify tests, I went ahead and pushed to nvme-next.
Quoted 38 lines · Keith Busch
Quoted 37 lines · Klaus Jensen
Quoted 35 lines · Keith Busch
After going through the zns enabling, I notice the controller enabling is not correct. Then I just continued maked more stuff. The series, I think, contains some of the less controversial patches from the two conflicting zns series, preceeded by some cleanups and fixes from me.
If this is all fine, I took the liberty of porting the zns enabling to it and made a public branch for consideration here:
http://git.infradead.org/qemu-nvme.git/shortlog/refs/heads/kb-zns
Dmitry Fomichev (1):
hw/block/nvme: report actual LBA data shift in LBAFKeith Busch (5):
hw/block/nvme: remove pointless rw indirection hw/block/nvme: fix log page offset check hw/block/nvme: support per-namespace smart log hw/block/nvme: validate command set selected hw/block/nvme: support for admin-only command setKlaus Jensen (3):
hw/block/nvme: reject io commands if only admin command set selected hw/block/nvme: add nsid to get/setfeat trace events hw/block/nvme: add trace event for requests with non-zero status codehw/block/nvme-ns.c | 5 ++ hw/block/nvme.c | 194 ++++++++++++++++++++---------------------- hw/block/trace-events | 6 +- include/block/nvme.h | 11 +++ 4 files changed, 114 insertions(+), 102 deletions(-)Signature · 1 line
2.24.1
Quoted 9 lines
Quoted 4 lines
These fixes all look good to me apart from the odd fixes that has been mentioned in the reviews. Since soft freeze is only two weeks away (Oct 27th), it would be nice to get this staged on nvme-next so we can get a pull sent off to Peter.
I've fixed up the comments mentioned and added the received reviews. Since it was pretty trivial fixups and passes my basic santify tests, I went ahead and pushed to nvme-next.
Looks good, thanks!