lkml.org 
[lkml]   [1998]   [Jan]   [4]   [last100]   RSS Feed
Views: [more markup]  [less markup]  [headers]  [forward] 
 
Messages in this thread
/
From(Alan Cox)
SubjectBadblocks impact on overall performance
DateSat, 3 Jan 1998 23:50:31 +0000 (GMT)
> When I do a badblocks -w on a reasonable size disk
> (I just put a 3gig scsi on my system), the system can
> go out to lunch for a few seconds (becomes very non-interactive).
> 
> Does a badblocks effectively  flush out the buffer cache?

No but two things occur which are bad

1.	The request queue gets filled unfairly by the one task because
	there's a lack of request fairness (its not normally an issue)

2.	The queue fills and blocks other drivers up because Linus still
	hasnt included the patch that was developed to keep a per
	device queue (SCSI doesnt have this problem as it empties the
	request queues onto its own queues), but stuff like IDE and
	especially slower devices - eg floppies are dire for it.

	(and yes Linus can we please have the multiple queues patch in 
	2.2 pending the new io_request stuff in 2.3...)

Alan

\
 
 \ /
  Last update: 2005-03-22 12:41    [W:1.341 / U:0.500 seconds]
©2003-2008 Jasper Spaans