On Oct 7, 2021, at 3:31 PM, David Wysochanski [off-list ref] wrote:

On Thu, Oct 7, 2021 at 3:25 PM Chuck Lever III [off-list ref] wrote:

quoted

quoted

On Oct 7, 2021, at 3:16 PM, David Wysochanski [off-list ref] wrote:

On Thu, Oct 7, 2021 at 12:22 PM Chuck Lever III [off-list ref] wrote:

quoted

quoted

On Oct 7, 2021, at 12:17 PM, Chuck Lever [off-list ref] wrote:

There are two new events that report slightly different information
for readpage and readpages.

For readpage:
          fsx-1387  [006]   380.761896: nfs_aops_readpage:    fileid=00:28:2 fhandle=0x36fbbe51 version=1752899355910932437 page_index=32

The index of a synchronous single-page read is reported.

For readpages:

          fsx-1387  [006]   380.760847: nfs_aops_readpages:   fileid=00:28:2 fhandle=0x36fbbe51 version=1752899355909932456 nr_pages=3

The count of pages requested is reported. readpages does not wait
for the READ requests to complete.

Signed-off-by: Chuck Lever <redacted>

Well obviously I forgot to update the patch description.
I can send a v3 later to do that.

Why not just call the tracepoints nfs_readpage and nfs_readpages?

Because there is already an nfs_readpage_done() tracepoint.

Ah ok.  FWIW, you could use nfs_readpage_enter() and
nfs_readpage_exit() similar to nfs_rmdir() for example.

Sure. I was hoping to tag all NFS address space ops methods so
they could turned on with a single command, like:

  # trace-cmd record -e nfs:nfs_aop_\*

But maybe that's not valuable.


--
Chuck Lever