Skip to contentSkip to conversation

block/nvme: Minor tracing improvements

PATCH Applied Philippe Mathieu-Daudé 3 participants 6 messages Jan 27, 2021 – Feb 2, 2021 Block layer

lore
Share & tools
lore
#0 lore raw

cover block/nvme: Minor tracing improvements

Fix an incorrect trace event (sometimes it is better to have no information rather than having incorrect info...) and add another event to display the spec implemented by the controller.

Philippe Mathieu-Daudé (2):
block/nvme: Properly display doorbell stride length in trace event block/nvme: Trace NVMe spec version supported by the controller

 block/nvme.c       | 8 +++++++-
 block/trace-events | 1 +
 2 files changed, 8 insertions(+), 1 deletion(-)
Signature · 1 line
2.26.2
#1 lore raw

PATCH 1/2 block/nvme: Properly display doorbell stride length in trace event

Commit 15b2260bef3 ("block/nvme: Trace controller capabilities") misunderstood the doorbell stride value from the datasheet, use the correct one. The 'doorbell_scale' variable used few lines later is correct.

---
 block/nvme.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
block/nvme.c +1 -1
index 5a6fbacf4a5..80c4318d8fc 100644
@@ -745,7 +745,7 @@ static int nvme_init(BlockDriverState *bs, const char *device, int namespace,
     trace_nvme_controller_capability("Contiguous Queues Required",
                                      NVME_CAP_CQR(cap));
     trace_nvme_controller_capability("Doorbell Stride",
-                                     2 << (2 + NVME_CAP_DSTRD(cap)));
+                                     1 << (2 + NVME_CAP_DSTRD(cap)));
     trace_nvme_controller_capability("Subsystem Reset Supported",
                                      NVME_CAP_NSSRS(cap));
     trace_nvme_controller_capability("Memory Page Size Minimum",
Signature · 1 line
2.26.2
Reviewed-by on v1 1/2
#2 lore raw

Re: [PATCH 1/2] block/nvme: Properly display doorbell stride length in trace event

Quoted 6 lines · Philippe Mathieu-Daudé

Commit 15b2260bef3 ("block/nvme: Trace controller capabilities") misunderstood the doorbell stride value from the datasheet, use the correct one. The 'doorbell_scale' variable used few lines later is correct.

Quoted 21 lines
---
 block/nvme.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
block/nvme.c +1 -1
index 5a6fbacf4a5..80c4318d8fc 100644
@@ -745,7 +745,7 @@ static int nvme_init(BlockDriverState *bs, const char *device, int namespace,
     trace_nvme_controller_capability("Contiguous Queues Required",
                                      NVME_CAP_CQR(cap));
     trace_nvme_controller_capability("Doorbell Stride",
-                                     2 << (2 + NVME_CAP_DSTRD(cap)));
+                                     1 << (2 + NVME_CAP_DSTRD(cap)));
     trace_nvme_controller_capability("Subsystem Reset Supported",
                                      NVME_CAP_NSSRS(cap));
     trace_nvme_controller_capability("Memory Page Size Minimum",
Signature · 1 line
2.26.2
Signature · 1 line
One of us - No more doubt, silence or taboo about mental illness.
#3 lore raw

PATCH 2/2 block/nvme: Trace NVMe spec version supported by the controller

In reply to Philippe Mathieu-Daudé

NVMe controllers implement different versions of the spec, and different features of it. It is useful to gather this information when debugging.

---
 block/nvme.c       | 6 ++++++
 block/trace-events | 1 +
 2 files changed, 7 insertions(+)
block/nvme.c +6 -0
index 80c4318d8fc..2b5421e7aa6 100644
@@ -708,6 +708,7 @@ static int nvme_init(BlockDriverState *bs, const char *device, int namespace,
     AioContext *aio_context = bdrv_get_aio_context(bs);
     int ret;
     uint64_t cap;
+    uint32_t ver;
     uint64_t timeout_ms;
     uint64_t deadline, now;
     volatile NvmeBar *regs = NULL;
@@ -764,6 +765,11 @@ static int nvme_init(BlockDriverState *bs, const char *device, int namespace,
     bs->bl.request_alignment = s->page_size;
     timeout_ms = MIN(500 * NVME_CAP_TO(cap), 30000);
 
+    ver = le32_to_cpu(regs->vs);
+    trace_nvme_controller_spec_version(extract32(ver, 16, 16),
+                                       extract32(ver, 8, 8),
+                                       extract32(ver, 0, 8));
+
     /* Reset device to get a clean state. */
     regs->cc = cpu_to_le32(le32_to_cpu(regs->cc) & 0xFE);
     /* Wait for CSTS.RDY = 0. */
block/trace-events +1 -0
index 8368f4acb0b..ecbc32a80a9 100644
@@ -136,6 +136,7 @@ qed_aio_write_main(void *s, void *acb, int ret, uint64_t offset, size_t len) "s
 # nvme.c
 nvme_controller_capability_raw(uint64_t value) "0x%08"PRIx64
 nvme_controller_capability(const char *desc, uint64_t value) "%s: %"PRIu64
+nvme_controller_spec_version(uint32_t mjr, uint32_t mnr, uint32_t ter) "Specification supported: %u.%u.%u"
 nvme_kick(void *s, unsigned q_index) "s %p q #%u"
 nvme_dma_flush_queue_wait(void *s) "s %p"
 nvme_error(int cmd_specific, int sq_head, int sqid, int cid, int status) "cmd_specific %d sq_head %d sqid %d cid %d status 0x%x"
Signature · 1 line
2.26.2
Reviewed-by on v1 2/2
#4 lore raw

Re: [PATCH 2/2] block/nvme: Trace NVMe spec version supported by the controller

Quoted 5 lines · Philippe Mathieu-Daudé

NVMe controllers implement different versions of the spec, and different features of it. It is useful to gather this information when debugging.

Quoted 45 lines
---
 block/nvme.c       | 6 ++++++
 block/trace-events | 1 +
 2 files changed, 7 insertions(+)
block/nvme.c +6 -0
index 80c4318d8fc..2b5421e7aa6 100644
@@ -708,6 +708,7 @@ static int nvme_init(BlockDriverState *bs, const char *device, int namespace,
     AioContext *aio_context = bdrv_get_aio_context(bs);
     int ret;
     uint64_t cap;
+    uint32_t ver;
     uint64_t timeout_ms;
     uint64_t deadline, now;
     volatile NvmeBar *regs = NULL;
@@ -764,6 +765,11 @@ static int nvme_init(BlockDriverState *bs, const char *device, int namespace,
     bs->bl.request_alignment = s->page_size;
     timeout_ms = MIN(500 * NVME_CAP_TO(cap), 30000);
 
+    ver = le32_to_cpu(regs->vs);
+    trace_nvme_controller_spec_version(extract32(ver, 16, 16),
+                                       extract32(ver, 8, 8),
+                                       extract32(ver, 0, 8));
+
     /* Reset device to get a clean state. */
     regs->cc = cpu_to_le32(le32_to_cpu(regs->cc) & 0xFE);
     /* Wait for CSTS.RDY = 0. */
block/trace-events +1 -0
index 8368f4acb0b..ecbc32a80a9 100644
@@ -136,6 +136,7 @@ qed_aio_write_main(void *s, void *acb, int ret, uint64_t offset, size_t len) "s
 # nvme.c
 nvme_controller_capability_raw(uint64_t value) "0x%08"PRIx64
 nvme_controller_capability(const char *desc, uint64_t value) "%s: %"PRIu64
+nvme_controller_spec_version(uint32_t mjr, uint32_t mnr, uint32_t ter) "Specification supported: %u.%u.%u"
 nvme_kick(void *s, unsigned q_index) "s %p q #%u"
 nvme_dma_flush_queue_wait(void *s) "s %p"
 nvme_error(int cmd_specific, int sq_head, int sqid, int cid, int status) "cmd_specific %d sq_head %d sqid %d cid %d status 0x%x"
Signature · 1 line
2.26.2
Signature · 1 line
One of us - No more doubt, silence or taboo about mental illness.
Applied
#5 lore raw
In reply to Philippe Mathieu-Daudé

Am 27.01.2021 um 22:21 hat Philippe Mathieu-Daudé geschrieben:

Quoted 4 lines

Fix an incorrect trace event (sometimes it is better to have no information rather than having incorrect info...) and add another event to display the spec implemented by the controller.

Thanks, applied to the block branch.

Kevin