On Wed, 7 Jun 2000, Alexander Viro wrote:
On Wed, 7 Jun 2000, Andrea Arcangeli wrote:
On Tue, 6 Jun 2000, Alexander Viro wrote:
Erm... Andrea, could you tell why was -i_count left
non-atomic? Since iput() can be called without the big lock...
Comments?
It's not atomic because we
On Tue, 6 Jun 2000, Alexander Viro wrote:
Erm... Andrea, could you tell why was -i_count left
non-atomic? Since iput() can be called without the big lock...
Comments?
It's not atomic because we have to synchronze the i_count changes with the
list changes (see __iget), so making it
On Thu, 20 Apr 2000, Stephen C. Tweedie wrote:
much difference, and it may be the stick we need to beat Linus into
believing that this change is really quite important.
The change is really quite important and it will return to be important as
it was importnat in 2.2.x when raid5 will work
On Sun, 2 Apr 2000, Andrew Morton wrote:
Should this also be applied to 2.2.15?
Not mandatory. It's not a must to set nr_segments to zero (also in 2.3.x).
But the elevator debugging code to better check that none regular request
can sit on the queue with req-q == NULL also assumes special
On Sat, 1 Apr 2000, Andrew Morton wrote:
A week of two back I was discussing a couple of elevator hiccups with
Andrea. I have now found a way to reproduce them. /bin/sync.
mnm:/usr/src/linux alias make="make -j3"
mnm:/usr/src/linux while true ; do sync ; sleep 15 ; done
mnm:/usr/src/linux
On Sat, 1 Apr 2000, Manfred Spraul wrote:
From: "Andrea Arcangeli" [EMAIL PROTECTED]
And the previous one (as said in earlier email) made no sense since 16:06
should be the sixth gscd and you said you don't have such device even
compiled into the kernel.
Perhaps the debug c
On Sun, 2 Apr 2000, Andrew Morton wrote:
Wow. I hadn't spotted that 'counter % 100'. They're spewing out now,
even without 'sync'
I've attached here one for device 21:05 (aka 33:05, aka /dev/hde5)
Here the fix:
--- 2.3.99-pre4-pre2aa/drivers/ide/ide.c.~1~Mon Mar 27 22:44:47 2000
+++
On Mon, 10 Jan 2000, Theodore Y. Ts'o wrote:
No. It's not an ext2 problem, and it's bad form to put safety checks
at the wrong layer. Invariably, that sort of magic test never gets
Agreed.
You can put safety checks in in the kernel level of MD, and in the MD
software tools; that should
On Mon, 10 Jan 2000, Theodore Y. Ts'o wrote:
There may be an issue for some block device drivers that might use 512
byte sectors (such as IDE disks) on a 32-bit system, but that won't be
an problem if you're using an MD device to put together the disks for
the 2TB filesystem.
Wrong, because
On Mon, 10 Jan 2000, Tigran Aivazian wrote:
I am sorry to be "picking" but 2^32*512 bytes is not 2048 T, it is not
even 2.1T but a mere 2T.
Sorry, I am always been bad in doing math calcs :)
Andrea
On Mon, 10 Jan 2000, Manfred Spraul wrote:
2^32 * 512 == 2^32 * 2^9 == 2^^41
2^10 kilo
2^20 mega
2^30 giga
2^40 terra
--- 2^^41 == 2 terrabyte.
Yes, so it's an issue also for ext2. Anyway all my emails still apply,
with the difference that the probabily to hit the bug increases. I
On Mon, 10 Jan 2000, Manfred Spraul wrote:
Now the RAID tools MUST NOT mount the 4 TB disk array, or you'll destroy
your data.
[the calculation
bh[i]-b_rsector = bh[i]-b_blocknr*(bh[i]-b_size9);
in ll_rw_blk.c would overflow]
Why? rsector is a multiple of 512bytes of course. But rsector
On Fri, 7 Jan 2000, Manfred Spraul wrote:
eg it seems that lvm doesn't contain the checks for lvm's 2 TB on
32-bit platforms.
Actually lvm has real issues as well. When the real issues will gets fixed
we'll add the avoid-admin-to-do-silly-things checks too :).
Andrea
On Thu, 6 Jan 2000, Manfred Spraul wrote:
I agree that ext2 would support 8TB with 4K blocks, but the block device
drivers don't support that.
Andi infact never talked about i386.
Andrea
On Tue, 21 Dec 1999, Benjamin C.R. LaHaise wrote:
The buffer dirty lists are the wrong place to be dealing with this. We
The only reason for not using buffer.c is to make sure to not insert bugs
in such file.
need a lightweight, fast way of monitoring the system's dirty buffer/page
The
On Tue, 21 Dec 1999, Stephen C. Tweedie wrote:
refile_buffer() checks in buffer.c. Ideally there should be a
system-wide upper bound on dirty data: if each different filesystem
starts to throttle writes at 50% of physical memory then you only
need two different filesystems to
On Tue, 21 Dec 1999, Stephen C. Tweedie wrote:
We cannot use the buffer.c dirty list anyway because bdflush can write
those buffers to disk at any time. Transactions have to control the
So you are talking about replacing this line:
dirty = size_buffers_type[BUF_DIRTY] PAGE_SHIFT;
On Sun, 5 Dec 1999, Alexander Viro wrote:
The whole point being that _nobody_ has it opened. It was attached to a
packet when sender called sendmsg(). That is, set of descriptors passed
by caller of sendmsg() had been converted into the set of struct file *
When recvmsg() is called by the
On Fri, 3 Dec 1999, Richard Gooch wrote:
What? Having group write on the directory? No thanks.
You can't hardlink a directory.
I'll tell another way that will let you understand correctly for sure.
I want that the i_link of an inode can be changed only by an user that has
write permissions on
On Fri, 3 Dec 1999, Alexander Viro wrote:
Andrea, you _do_ realize that CODA is not a Linux-only thing? So it's
Have I ever talked about coda in first place?
Andrea
On Fri, 3 Dec 1999, Alexander Viro wrote:
... and F- on UNIX SA 101 - if you don't know the reasons to keep /tmp on
a separate filesystem.
Would you call this a solution? This is a very ugly workaround. The fact
this works is only a side effect of the limitations of the hardlink.
So another
On Thu, 2 Dec 1999, Alexander Viro wrote:
such games link() is the least of your problems - it's effect can be
completely reproduced with plain open(). exec 42/bar/foo and several
hours after that sh -c /dev/fd/42 will do the trick - fork() preserves
open descriptors.
If there was really a
On Sun, 28 Nov 1999, Manfred Spraul wrote:
for(;;){
read(fd,data,sizeof(data));
if(EOF)
break;
process_data(data);
}
Do you know if we have to ensure that no record is returned twice, no
record is missed? I just checked the POSIX standard, but I found
On Mon, 18 Oct 1999, Alexander Viro wrote:
? The same way you are doing it with pagecache.
I think if I would have just understood I wouldn't be asking here ;).
WTF would we _need_ to know? Think about it as about memory caching. You
can cache by virtual address and you can cache by physical
On Mon, 18 Oct 1999, Stephen C. Tweedie wrote:
Data and metadata are completely different. On, say, a large and busy
web or ftp server, you really don't care about a 1G metadata limit, but
a 1G page cache limit is much more painful.
Sure I completly agree. It looked you was talking about
On Tue, 12 Oct 1999, Stephen C. Tweedie wrote:
changes. The ext2 truncate code is really, really careful to provide
I was _not_ talking about ext2 at all. I was talking about the bforget and
brelse semantics. As bforget fallback to brelse you shouldn't expect
bforget to really destroy the
On 11 Oct 1999, Eric W. Biederman wrote:
What about adding to the end of ext2_alloc_block:
It's _equally_ slow. Do you seen my patch? I prefer to do the query at the
higher lever to save cpu cache.
Andrea
On Tue, 12 Oct 1999, Stephen C. Tweedie wrote:
Umm, your last proposal was to do a hash lookup on each new page cache
buffer mapping. That is a significant performance cost, which IMHO is
not exactly the right direction either. :)
It's not obvious that the only thing to consider are
On Fri, 8 Oct 1999, Mikulas Patocka wrote:
Here goes quick'n'dirty patch. It does bforget(). It should prevent file
corruption.
wrong patch. bforget give you no guarantee at all. bfoget always fallback
to brelse if necessary.
What I said about bforget in my old email is still true. The _only_
On Sat, 21 Aug 1999, Bradley D. LaRonde wrote:
but booting with initrd gives the following error (then goes on, apparently
working fine):
Hmm... probably I missed something specific to initrd (I have not tried it
myself btw). I'll have a look at it now.
does not get the error. dd'ing an image
-unlinking stuff on a page that you are asking to go in the
freelist.
Andrea Arcangeli
;
if (next) {
bh-b_next = next;
next-b_pprev = bh-b_next;
}
*bhp = bh;
bh-b_pprev = bhp;
nr_hashed_buffers++;
}
Andrea Arcangeli
(bh);
bh-b_dev = B_FREE;
insert_into_queues(bh);
I just said exactly that some days ago in my _first_ email with the
subject `buffer.c glitches'. Both invalidate_buffers() and set_blocksize()
should put all invalidated buffers in the freelist.
Andrea Arcangeli
On Fri, 5 Nov 1999, Mikulas Patocka wrote:
What if user does cat /dev/hda1/dev/null on mounted device while
filesystem driver does getblk? ... the buffer gets corrupted.
I was considering the case where the buffer returned by getblk is _just_
uptodate. If the block is just uptodate then all
34 matches
Mail list logo