lkml.org 
[lkml]   [2004]   [Mar]   [19]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH] barrier patch set
On Fri, Mar 19 2004, Jeff Garzik wrote:
> Jens Axboe wrote:
> >Cosmetic stuff that will get ironed out. You can find the patches here:
> >
> >ftp://ftp.kernel.org/pub/linux/kernel/people/axboe/patches/v2.6/2.6.5-rc1-mm2/
> >
> >ide-barrier-2.6.5-rc1-mm2-1
> > ide/core part
>
>
> WRT ATA and flush-cache... before the IDE pieces of this patch are
> merged, IMO it is a requirement that the entire flush-cache stuff gets a
> review. ide_get_error_location() is one of the important pieces

Well what are you waiting for :-). Honestly, the bits have been "out
there" for 2 years or so, so it's not like it's anything brand new.

ide_get_error_location() is new for 2.6, I consider the 2.6 code to have
basically 100% coverage where the 2.4 one never really did...

> (great!). Another important piece is to make sure that a drive's
> flush-cache capability is correctly deduced and set up from the
> identify-device. The steps look like
>
> - check identify-device word 83, bits 12 (flush cache) and 13 (flush
> cache ext)
> - issue set-features command to get flush-cache into proper state
> (enabled or disabled, as the user desires), if identify-device word 86
> indicates it is not already in the state you seek.
> - re-read identify [packet] device page from device, make sure
> flush-cache[-ext] is enabled. A slacker could just make sure the
> set-features command completed successfully, but to be 100% correct you
> need to re-read the identify-device page. :/
>
> NOTE 1: these steps are specific to the flush-cache command, and are
> only vaguely related to write caching (which must be tested-for and
> enabled separately). It is important to go through these steps
> separately for write-caching and flush-cache[-ext].
>
> Write-caching and flush-cache-command state should always be considered
> separately, even though the two are often used together. I want to
> avoid the LG cdrom debacle, where not properly checking for, and setting
> up, flush-cache resulted in turning cdroms into bricks.
>
> NOTE 2: Don't forget to check for 0x0000 or 0xffff in word 86, of
> __only__ identify-packet-device. These special case meanings do not
> appear for ATA devices, only ATAPI devices.
>
> NOTE 3: flush-cache-ext should always be used on lba48 devices, even if
> the request was inside the lba28 limit.
>
> NOTE 4: flush-cache-ext does not exist on ATAPI devices.

Only ide-disk is supported. But that's some great info, I'll go over the
code this weekend and make sure we're doing ok and correct any
deviations.

--
Jens Axboe

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 14:01    [W:0.131 / U:0.032 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site