> Welcome to the Twilight Zone! Here is the trace from an iozone run
> against WfW with debugging turned on in smbfs/file.c:
>
> You will note the glaring absence of any log entries pertaining to
> smb_read operations <g>. Nonetheless, reads are occurring (albeit at
> slower rates than would be expected with cacheing), so where is the data
> coming from? BTW, I'm confident that read tracing is operational, as it
> does pickup activity from cat-ing existing files on the server.
Hi Steve,
The only explanation I can think of is that the flood of messages
overloaded your logging. I tried running
iozone 1 8192
and got a more modest set of log messages showing the transition from
write to read. You should see an fsync, a close, then the file being
opened, and the count of the number of pages cached. On the NT server,
it shows 256 pages cached after finishing the write.
Regards,
Bill