Date: Thu, 23 Jan 2003 15:49:21 -0800 From: Juli Mallett <jmallett@FreeBSD.org> To: Nate Lawson <nate@root.org> Cc: current@FreeBSD.org Subject: Re: Test this! Patch to make newfs(8) use libufs. Message-ID: <20030123154921.A66976@FreeBSD.org> In-Reply-To: <20030123153811.A66302@FreeBSD.org>; from jmallett@FreeBSD.org on Thu, Jan 23, 2003 at 03:38:12PM -0800 References: <20030123142924.A61481@FreeBSD.org> <Pine.BSF.4.21.0301231507540.73010-100000@root.org> <20030123153811.A66302@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
* De: Juli Mallett <jmallett@FreeBSD.org> [ Data: 2003-01-23 ]
[ Subjecte: Re: Test this! Patch to make newfs(8) use libufs. ]
> * De: Nate Lawson <nate@root.org> [ Data: 2003-01-23 ]
> [ Subjecte: Re: Test this! Patch to make newfs(8) use libufs. ]
> > On Thu, 23 Jan 2003, Juli Mallett wrote:
> > > * De: Nate Lawson <nate@root.org> [ Data: 2003-01-23 ]
> > > [ Subjecte: Re: Test this! Patch to make newfs(8) use libufs. ]
> > > > On Thu, 23 Jan 2003, Juli Mallett wrote:
> > > > > Thanks to Ruslan's reminder that tunefs now uses libufs and tunefs is
> > > > > one of the crunched programs, I realised that I really needed to make
> > > > > newfs(8) use libufs. To show off that it *can* help us reduce space,
> > > > > a good bit in some cases.
> > > >
> > > > Good to see this. Does libufs do an fsync() in bwrite() or do you need to
> > > > do that manually in place of the previous calls to wtfsflush()?
> > >
> > > It doesn't. Do you think it should? I'd rather do that than litter the
> > > newfs code.
> >
> > Not needed. On second look, I see you removed the wc[] caching and fall
> > back to just write(). (BTW, what's the performance difference with your
> > patch?) If newfs doesn't complete successfully due to a system crash, it
> > doesn't matter if data written was flushed to disk, just run it again.
>
> *nod*. Not to mention that I don't think you need it for writing to a raw
> device?
>
> As for performance, I don't know. Test it and find out? :)
I just went to do this, and found that pwrite is failing, saying EBADF.
Could this be because of the failed ioctl? I'm not sure why this is
happening. Any thoughts would be appreciated.
12228 newfs RET open 4
12228 newfs CALL fstat(0x4,0xbfbff798)
12228 newfs RET fstat 0
12228 newfs CALL ioctl(0x4,DIOCGSECTORSIZE,0x80509c0)
12228 newfs RET ioctl 0
12228 newfs CALL ioctl(0x4,DIOCGMEDIASIZE,0xbfbff790)
12228 newfs RET ioctl 0
12228 newfs CALL ioctl(0x4,DIOCGDINFO,0x8050480)
12228 newfs RET ioctl -1 errno 25 Inappropriate ioctl for device
12228 newfs CALL gettimeofday(0xbfbff5c8,0)
12228 newfs RET gettimeofday 0
12228 newfs CALL open(0x28111300,0,0)
12228 newfs NAMI "/dev/random"
12228 newfs RET open 5
12228 newfs CALL read(0x5,0x2811d0a4,0x7c)
12228 newfs GIO fd 5 read 124 bytes
"\^A9\M^I\^S\M-`b\M^Y\M-T~\M-I\b{\M-mN\M-b\M-{(\^Q\M-k|n\M-( :\M-pA\
S\^Y\M-|\M-m\M^P\M-M*\M-%\^_e\M^G>\M-v\M-H\M-;2*\M-e\M-E\M-d\M^Y\M^U=\
\^WrB\^S\M-*\^UuL\0\M-=\M-M7@i\M-t`\M^Y\M-+\M-U\M^X\M-jA{\M-L cq_TvW\
\^_\M-5\M-Af\M-)\^TP8
\^U\M-.
*\^V\^^\M-=\^R\M-e\M-'I\M-Tm\M^BY\M^M\M-W\M^JZ\M-y\240gX\M-=\M^Z\M-)\0\
\^P\^R\M^]\^Xq\M-F0\M-3"
12228 newfs RET read 124/0x7c
12228 newfs CALL close(0x5)
12228 newfs RET close 0
12228 newfs CALL pwrite(0x4,0x8050a24,0x200,0,0x1fff,0)
12228 newfs RET pwrite -1 errno 9 Bad file descriptor
I didn't notice this before as the test ran fine :/ Last time I trust a test
I didn't write to be honest. Also, forget to check the return value of code.
Any help would be appreciated.
Thanx,
juli.
--
Juli Mallett <jmallett@FreeBSD.org>
AIM: BSDFlata -- IRC: juli on EFnet.
OpenDarwin, Mono, FreeBSD Developer.
ircd-hybrid Developer, EFnet addict.
FreeBSD on MIPS-Anything on FreeBSD.
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20030123154921.A66976>