Date: Tue, 12 Feb 2008 14:33:20 +0200 From: Andriy Gapon <avg@icyb.net.ua> To: Poul-Henning Kamp <phk@phk.freebsd.dk> Cc: freebsd-fs@FreeBSD.org, freebsd-hackers@FreeBSD.org, Pav Lucistnik <pav@FreeBSD.org>, Bruce Evans <bde@zeta.org.au> Subject: Re: fs/udf: vm pages "overlap" while reading large dir Message-ID: <47B19210.3010508@icyb.net.ua> In-Reply-To: <4299.1202816854@critter.freebsd.dk> References: <4299.1202816854@critter.freebsd.dk>
next in thread | previous in thread | raw e-mail | index | archive | help
on 12/02/2008 13:47 Poul-Henning Kamp said the following: > In message <47B181F2.2070808@icyb.net.ua>, Andriy Gapon writes: > >> 2.3. this code passes to bread blkno that is calculated as 4*sector, >> where sector is a number of a physical 2048-byte sector. [**] >> [**] - I think that this is a requirement of buffcache system, because >> internally it performs many calculations that seem to assume that block >> size is always 512. > > Yes, the buf-cache works in 512 bytes units throughout. > >> 3.1. for a fresh buf getlbk would assign the following: >> bsize = bo->bo_bsize; >> offset = blkno * bsize; > > That's clearly wrong. > > We need to assert that the blkno is aligned to the start of a sector > and use the 512 byte units, so I guess it would be: > > offset = dbtob(blkno); > KASSERT(!(offset & (bsize - 1)), ("suitable diagnostic")); > > Thank you for this very insightful and neat suggestion! I think that it must work but I will try test it tonight on whatever media and FS-s I have available. Thank you again! P.S. hope to not get '"suitable diagnostic"' from something like msdosfs :-) -- Andriy Gapon
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?47B19210.3010508>