Backported fix for premature return from burn_disc_erase on some drives
This commit is contained in:
parent
bf4545e803
commit
3eedf0327b
@ -320,8 +320,8 @@ void burn_disc_erase_sync(struct burn_drive *d, int fast)
|
||||
/* read the initial 0 stage */
|
||||
while (!d->test_unit_ready(d) && d->get_erase_progress(d) == 0)
|
||||
sleep(1);
|
||||
while (!d->test_unit_ready(d) &&
|
||||
(d->progress.sector = d->get_erase_progress(d)) > 0)
|
||||
while ((d->progress.sector = d->get_erase_progress(d)) > 0 ||
|
||||
!d->test_unit_ready(d))
|
||||
sleep(1);
|
||||
d->progress.sector = 0x10000;
|
||||
d->busy = BURN_DRIVE_IDLE;
|
||||
|
Loading…
Reference in New Issue
Block a user