lkml.org 
[lkml]   [2013]   [Jan]   [13]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectRe: [block] allow blk_flush_policy to return REQ_FSEQ_DATA independent of *FLUSH
From
Hello,
Re-reading the code, I see that I was wrong with the previous comment of mine. So, in summary

i) If block driver does not support FLUSH and FUA, then __generic_make_request() checks will clear BIO FLUSH and FUA flags and hence blk_insert_flush() will not be invoked.

ii) If block driver clears FLUSH & FUA flags while IO is in flight, then there is possibility of IO missing __generic_make_request() checks and hitting issue with blk_insert_flush() being discussed here.

iii) If block driver sets only REQ_FUA without REQ_FLUSH then __generic_make_request() checks will not clear BIO flags and hence blk_insert_flush() will be invoked which will hit the blk_insert_flush() issue being discussed here. However, setting REQ_FUA without REQ_FLUSH is not as per documentation and it is invalid for block driver to do so.

Thanks,
Ajith

On Wednesday, 9 January 2013 09:44:55 UTC+5:30, Ajith Kumar wrote:
> Hello,
>
> Thanks for the response.
>
> A block device driver during initialization would decide if it is capable of supporting FLUSH/FUA or not. Suppose driver claims FLUSH/FUA support then any bio targeted at this driver with FLUSH bit set(which is commonly set by file system like XFS for doing internal logging) has a data corruption vulnerability in case of an abrupt shutdown. So, IMO the vulnerability is not open to rare window where driver changes q->flush_flags while IO is in flight, but for a much larger window from time driver comes up and throughout it's life.
>
>
>
> Thanks,
>
> Ajith
>
>
>
> On Wednesday, 9 January 2013 00:15:31 UTC+5:30, Tejun Heo wrote:
>
> > Hello,
>
> >
>
> >
>
> >
>
> > On Tue, Jan 08, 2013 at 10:04:23AM -0800, ajithb.kumar@gmail.com wrote:
>
> >
>
> > > Hi,
>
> >
>
> > > Could you please provide clarity on the following.
>
> >
>
> > > "> Hmmm... yes, this can become a correctness issue if (and only if)
>
> >
>
> > > > blk_queue_flush() is called to change q->flush_flags while requests
>
> >
>
> > > > are in-flight;"
>
> >
>
> > >
>
> >
>
> > > Could you please clarify as to why is it a correctness issue only if
>
> >
>
> > > blk_queue_flush() is used to change flush_flags when requests are in
>
> >
>
> > > flight ? As I understand, XFS does set WRITE_FLUSH_FUA flag in
>
> >
>
> > > _xfs_buf_ioapply() function irrespective of whether the underlying
>
> >
>
> > > device supports flush capabilities or not which will flow into
>
> >
>
> > > blk_insert_flush(). Is my reading of the code correct and is there
>
> >
>
> > > a general correctness issue here which potentially results in XFS
>
> >
>
> > > file system corruption in case of an abrupt shutdown independent of
>
> >
>
> > > q->flush_flags getting changed while request is in flight.
>
> >
>
> >
>
> >
>
> > My memory is kinda fuzzy at this point but if a queue doesn't support
>
> >
>
> > flush, its flush_flags should be zero and
>
> >
>
> > generic_make_request_checks() will clear REQ_FLUSH|REQ_FUA from
>
> >
>
> > bio->bi_rw so we never hit blk_insert_flush() and the request will be
>
> >
>
> > processed as a normal IO one; however, if REQ_FLUSH goes off after a
>
> >
>
> > request passed generic_make_request_checks() but before
>
> >
>
> > blk_flush_policy(), it'll become null op and its data payload won't
>
> >
>
> > get written out to the underlying device, which is data corruption.
>
> >
>
> >
>
> >
>
> > Thanks.
>
> >
>
> >
>
> >
>
> > --
>
> >
>
> > tejun
>
> >
>
> > --
>
> >
>
> > 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: 2013-01-13 09:21    [W:0.028 / U:1.252 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site