Re: [PATCH 05/12] linux: Reduce worst case wait to 10 seconds for disks

[Thomas Schmitt]

Hi,

looking at
  
https://git.savannah.gnu.org/cgit/hurd/gnumach.git/tree/linux/src/drivers/block/ide.c
i get the impression that WAIT_WORSTCASE is for a single ATAPI
transaction. So a long delay loop would not necessarily be a case for
WAIT_WORSTCASE.

But my Pioneer BDR-S09 drive shows long delays with its reply to the
START/STOP UNIT which loads the tray. All my other drives reply quickly
and then indicate Not Ready to TEST UNIT READY for the same time.

With a DVD+RW in its open tray it needs 9.7 seconds to reply to
START/STOP UNIT.

  xorriso -scsi_log on -outdev /dev/sr1
reports on stderr (among many other lines):

  START/STOP UNIT
  1b 00 00 00 03 00
    9702684 us     [ 9854648 ]

The third line tells that the ioctl(SG_IO) lasted 9702684 microseconds.
(This time span was determined by two calls to gettimeofday(2), not by
sg_io_hdr_t.duration.)
At the end of ioctl(SG_IO), libburn was running since 9854648 microseconds.

With a DVD-RW it needed 8.8 seconds, with a BD-RE it needed 8.3 seconds.


I wrote:
> Worst time measured by libburn's SCSI logging was about 25 seconds.
> That was in autumn 2020 with a freshly bought Pioneer BD drive

Re-reading my unsuccessful patch for Linux automatic tray loading, where
i proposed 40 seconds for a waiting loop around TEST UNIT READY, it
must have been an ASUS DRW-24D5MT DVD burner of 2018.


Have a nice day :)

Thomas