search

linux-nvme / nvme-trace

Decode nvme trace events
GNU General Public License v2.0
17 stars 4 forks source link

Add comments on events #1

Open TimHe95 opened 2 years ago

TimHe95 commented 2 years ago

What does the events means in /sys/kernel/debug/tracing/events/nvme? There are four events in this directory:

  1. nvme_async_event
  2. nvme_complete_rq
  3. nvme_setup_cmd
  4. nvme_sq

My kernel version: 5.4.0 I want to monitor the depth of each queue in real time. I suppose when I see nvme_setup_cmd, this means I can +1 the depth and when I see nvme_sq, I can -1 the depth. Is that right?

Thanks!

TimHe95 commented 2 years ago

And what is the difference between nvme_complete_rq and nvme_sq:

          <idle>-0       [001] d.h. 86191.129911: nvme_complete_rq: nvme0: disk=nvme0n1, qid=2, cmdid=83, res=0x0, retries=0, flags=0x0, status=0x0
          <idle>-0       [001] d.h. 86191.129912: nvme_sq: nvme0: disk=nvme0n1, qid=2, head=118, tail=118
          <idle>-0       [001] d.h. 86191.129913: nvme_complete_rq: nvme0: disk=nvme0n1, qid=2, cmdid=84, res=0x0, retries=0, flags=0x0, status=0x0
          <idle>-0       [001] d.h. 86191.129914: nvme_sq: nvme0: disk=nvme0n1, qid=2, head=118, tail=118
          <idle>-0       [001] d.h. 86191.129914: nvme_complete_rq: nvme0: disk=nvme0n1, qid=2, cmdid=85, res=0x0, retries=0, flags=0x0, status=0x0
          <idle>-0       [001] d.h. 86191.129915: nvme_sq: nvme0: disk=nvme0n1, qid=2, head=118, tail=118
          <idle>-0       [001] d.h. 86191.129915: nvme_complete_rq: nvme0: disk=nvme0n1, qid=2, cmdid=86, res=0x0, retries=0, flags=0x0, status=0x0
          <idle>-0       [001] d.h. 86191.129916: nvme_sq: nvme0: disk=nvme0n1, qid=2, head=118, tail=118
          <idle>-0       [001] d.h. 86191.129917: nvme_complete_rq: nvme0: disk=nvme0n1, qid=2, cmdid=87, res=0x0, retries=0, flags=0x0, status=0x0
          <idle>-0       [001] d.h. 86191.129918: nvme_sq: nvme0: disk=nvme0n1, qid=2, head=118, tail=118
          <idle>-0       [001] d.h. 86191.129918: nvme_complete_rq: nvme0: disk=nvme0n1, qid=2, cmdid=88, res=0x0, retries=0, flags=0x0, status=0x0
          <idle>-0       [001] d.h. 86191.129919: nvme_sq: nvme0: disk=nvme0n1, qid=2, head=118, tail=118
          <idle>-0       [001] d.h. 86191.129919: nvme_complete_rq: nvme0: disk=nvme0n1, qid=2, cmdid=89, res=0x0, retries=0, flags=0x0, status=0x0
          <idle>-0       [001] d.h. 86191.129921: nvme_sq: nvme0: disk=nvme0n1, qid=2, head=118, tail=118
          <idle>-0       [001] d.h. 86191.129921: nvme_complete_rq: nvme0: disk=nvme0n1, qid=2, cmdid=90, res=0x0, retries=0, flags=0x0, status=0x0
          <idle>-0       [001] d.h. 86191.129922: nvme_sq: nvme0: disk=nvme0n1, qid=2, head=118, tail=118
          <idle>-0       [001] d.h. 86191.129922: nvme_complete_rq: nvme0: disk=nvme0n1, qid=2, cmdid=91, res=0x0, retries=0, flags=0x0, status=0x0
          <idle>-0       [001] d.h. 86191.129923: nvme_sq: nvme0: disk=nvme0n1, qid=2, head=118, tail=118

It seems that they occur together and has very small time gap (1 us).