9fans - fans of the OS Plan 9 from Bell Labs
 help / color / mirror / Atom feed
* [9fans] plan9ports venti with arenas on fat filesystem
@ 2007-10-12  1:02 Adrian Tritschler
  2007-10-12  2:28 ` Russ Cox
  0 siblings, 1 reply; 3+ messages in thread
From: Adrian Tritschler @ 2007-10-12  1:02 UTC (permalink / raw)
  To: Fans of the OS Plan 9 from Bell Labs

I've been playing/experimenting with a p9p venti with arenas in 512M
files on linux ext3 file systems with no problems, today I decided to
set up a second venti on an external USB drive so I could have a backup
of it.  It seems that I can't create arenas on files on the USB fat32 disk.

% cd /mnt/venti
% venti/fmtarenas arenas6 ./arenas6
fmtarenas ./arenas6: 1 arenas, 536,076,288 bytes storage, 524,288 bytes
for index map
2007/1012 10:25:17 err 4: write ./arenas6 offset 0xc0000 count 49152 buf
8074000 returned 16384: Success
2007/1012 10:25:17 err 4: can't write arena set: write ./arenas6 offset
0xc0000 count 49152 buf 8074000 returned 16384: Success
can't write back arena partition header for ./arenas6: can't write arena
set: write ./arenas6 offset 0xc0000 count 49152 buf 8074000 returned
16384: Success

Copying the file to the ext3 system, performing the venti/fmtarenas then
copying it back works.

% rsync -P arenas[67] $HOME/tmp
arenas6
   536870912 100%    4.96MB/s    0:01:43 (xfer#1, to-check=1/2)
arenas7
   536870912 100%    4.61MB/s    0:01:51 (xfer#2, to-check=0/2)

sent 1073873034 bytes  received 64 bytes  5006401.39 bytes/sec
total size is 1073741824  speedup is 1.00

% cd $HOME/tmp
% venti/fmtarenas arenas6 ./arenas6
fmtarenas ./arenas6: 1 arenas, 536,076,288 bytes storage, 524,288 bytes
for index map
% venti/fmtarenas arenas7 ./arenas7
fmtarenas ./arenas7: 1 arenas, 536,076,288 bytes storage, 524,288 bytes
for index map

Then back on the USB fat32 system all *seems* to work:

% cd /mnt/venti
% rsync -P  $HOME/tmp/arenas6 .
 :
% venti/fmtindex venti.conf
fmtindex: 8 arenas, 23,800 index buckets, 4,288,479,232 bytes storage
% venti/venti -d -h tcp!localhost!1090 -a tcp!localhost!17036
2007/1012 10:40:42 venti: conf...httpd
tcp!localhost!1090...init...icache 0 bytes = 1,000 entries; 4 scache
sync...announce tcp!localhost!17036...serving.

As soon as venti/wrarena starts to write to it, many errors occur, all
seem to be of the form:

2007/1012 10:43:15 err 4: write ./isect1 offset 0x73c000 count 65536 buf
810c000 returned 16384: Success
venti/venti: part ./isect1 addr 0x10c000: icachewritesect writepart:
write ./isect1 offset 0x73c000 count 65536 buf 810c000 returned 16384:
Success

Is the correct answer, "just don't do that", am I doing something wrong,
or this a bug that anyone is interested in fixing?

	Adrian



^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [9fans] plan9ports venti with arenas on fat filesystem
  2007-10-12  1:02 [9fans] plan9ports venti with arenas on fat filesystem Adrian Tritschler
@ 2007-10-12  2:28 ` Russ Cox
  2007-10-12  8:19   ` Robert William Fuller
  0 siblings, 1 reply; 3+ messages in thread
From: Russ Cox @ 2007-10-12  2:28 UTC (permalink / raw)
  To: 9fans

The code is in /usr/local/plan9/src/cmd/venti/srv/part.c.
And it is very complicated.  Far too complicated, but hey,
that's modern Unix.

It looks like pwrite is being asked to write 64k and only
writing 16k.  That alone shouldn't be a problem unless
the code believes the fs block size is > 16k.

I don't see anything obviously wrong (it's that complicated)
but I would certainly accept bug fixes.  ;-)

Russ



^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [9fans] plan9ports venti with arenas on fat filesystem
  2007-10-12  2:28 ` Russ Cox
@ 2007-10-12  8:19   ` Robert William Fuller
  0 siblings, 0 replies; 3+ messages in thread
From: Robert William Fuller @ 2007-10-12  8:19 UTC (permalink / raw)
  To: Fans of the OS Plan 9 from Bell Labs

Russ Cox wrote:
> The code is in /usr/local/plan9/src/cmd/venti/srv/part.c.
> And it is very complicated.  Far too complicated, but hey,
> that's modern Unix.
> 
> It looks like pwrite is being asked to write 64k and only
> writing 16k.  That alone shouldn't be a problem unless
> the code believes the fs block size is > 16k.

A FAT filesystem of that size should have a block size of 64K, no?


^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2007-10-12  8:19 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2007-10-12  1:02 [9fans] plan9ports venti with arenas on fat filesystem Adrian Tritschler
2007-10-12  2:28 ` Russ Cox
2007-10-12  8:19   ` Robert William Fuller

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).