On Po 06-02-06 22:13:20, Nigel Cunningham wrote:
> On Monday 06 February 2006 20:59, Pavel Machek wrote:
> > I do not see why I'd have to modify freezer for supporting swap
> > files. It is really unrelated part. If I want to support swap files,
> > I'll just bmap() them from userspace, then write to blockdevice
> > directly, like lilo or grub does.
>
> It would be needed because they're on filesystems. Freezing the
> filesystems, then trying to eat memory to the degree that the system tries
> to swap would lead to a deadlock if that swap was on a frozen filesystem.
> Conversely, trying to freeze the memory before freezing processes would be
> unreliable because it relies on not racing against other processes trying
> to work.
I need to think about that one... it should be possible to allocate
space in swap file *before* freeze, and get disk address via
bmap... are you sure swapfile accesses go through filesystems? I
thought they use something bmap-like internally.
> > > d. Storage of the image.
> > >
> > > As it currently stands, the interface between userspace and the kernel
> > > for uswsusp looks clean and simple. This is mainly, however, because
> > > it only supports writing to swap, and strictly synchronously.
> >
> > ...and it is going to stay that way.
>
> No plans for writing to anywhere else but swap?
Yes, but with bmap, it is possible to do that with current
interface. Early version even supported swap partitions.
> > > userspace for getting the sector numbers of storage. Finally, you'd
> > > want to use bio functions to submit the I/O, and a kernel routine to
> > > handle the completion. Then you'd need some mechanism to wait for or
> > > check for completion of I/O on a particular page or all pages. Of
> > > course you might decide not to do async I/O because it's too complex,
> > > but then you'd take the performance hit you currently have, and we
> > > wouldn't have an apples with apples comparison.
> >
> > Submit bios from userspace? Eeek? We have perfectly working async io
>
> Yes. But then how do you submit I/O to files on filesystems you can't
> mount? You're stuck with swap partitions only forever?
read/write directly block device is perfectly possible from
userspace. Small problem is finding suspend header with the map, but
you somehow solved that already, right?
> > interface for userland, and BTW going async will not give you too much
> > of performance advantage here...
>
> How do you know that? Suspend2 has async I/O, and can write the image as
> fast as the drive can take it. Some testing I did a while ago showed a max
> throughput of 16MB/s for swsusp vs 35 (what the drive is capable of) for
> Suspend2. Add LZF compression and it's 70MB/s vs 16MB/s.
Userspace is perfectly capable of saturating I/O subsystem. No magical
"async io" is needed for that. time dd if=/dev/zero of=/dev/hda
bs=... if you don't believe me.
> > Current uswsusp is 3K lines of code in kernel, 1K lines of code in
> > userspace.
> >
> > When we are done, we'll have perhaps 2.5K lines of code in kernel
> > (in-kernel swap writing support goes away), and maybe 20K lines in
> > userspace.
>
> Without adding which out of async I/O, compression, encryption, swap file
> support, and ordinary file support?
I'll get same bandwidth as you, without need for async I/O. Async I/O
is not really a feature, suspend speed is. (There are existing
interfaces for doing AIO from userspace, anyway, but I'm pretty sure
they will not be needed
Compression, encryption, swapfile/normal file is planned, and should
not affect kernel code.
Pavel
--
Web maintainer for suspend.sf.net (www.sf.net/projects/suspend) wanted...
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
[Index of Archives]
[Kernel Newbies]
[Netfilter]
[Bugtraq]
[Photo]
[Stuff]
[Gimp]
[Yosemite News]
[MIPS Linux]
[ARM Linux]
[Linux Security]
[Linux RAID]
[Video 4 Linux]
[Linux for the blind]
[Linux Resources]
