Skip to contentSkip to conversation

Attendance request

ATTEND Dave Kleikamp 4 participants 8 messages Feb 2, 2012 – Feb 3, 2012 VFS

lore
Share & tools
lore
#0 lore raw

Committee members,

Please consider inviting me to the Storage, Filesystem, & MM Summit. I am working in Chris Mason's team at Oracle focusing on storage I/O.

My latest focus is on completing the work started on by Zach Brown to add a kernel interface to allow kernel code to submit aio, and having the loop device submit direct IO requests to the underlying file system.

Zach's prototype:
http://kerneltrap.org/mailarchive/linux-fsdevel/2009/10/22/6516163

I should have patches ready for review within the next couple weeks.

Having gotten back into the filesystem/storage area after briefly working the embedded world at IBM, I look forward to participating in the summit and seeing everyone face-to-face again.

Thanks,
Dave "Shaggy" Kleikamp

#1 lore raw
Quoted 4 lines

My latest focus is on completing the work started on by Zach Brown to add a kernel interface to allow kernel code to submit aio, and having the loop device submit direct IO requests to the underlying file system.

You poor dear!

If you need any input, just holler. I'm still around.

- z

#2 lore raw
Quoted 8 lines · Zach Brown
Quoted 4 lines

My latest focus is on completing the work started on by Zach Brown to add a kernel interface to allow kernel code to submit aio, and having the loop device submit direct IO requests to the underlying file system.

You poor dear!

If you need any input, just holler. I'm still around.

I'll certainly cc you when I have patches ready for review. Your patches didn't suffer very much bit rot at all.

Thanks,
Shaggy

#3 lore raw
In reply to Dave Kleikamp
Quoted 8 lines · Dave Kleikamp

Committee members,

Please consider inviting me to the Storage, Filesystem, & MM Summit. I am working in Chris Mason's team at Oracle focusing on storage I/O.

My latest focus is on completing the work started on by Zach Brown to add a kernel interface to allow kernel code to submit aio, and having the loop device submit direct IO requests to the underlying file system.

We should talk; I'm currently working on an async bmap function for ext4, which I then intended to use for a modified high-level AIO support framework that allowed for async metadata access. That is, I'm interested in solving the "AIO isn't really 'A'" problem.

That's a different problem that what you're working on, but it's obviously going to touch the same code. (Actually, my original plan was to not touch the generic code in __blockdev_direct_IO and below and create some ext4-specific code for now, but if you're going to be messing around in that code path anyway, we should coordinate....)

- Ted

#4 lore raw
Quoted 5 lines

We should talk; I'm currently working on an async bmap function for ext4, which I then intended to use for a modified high-level AIO support framework that allowed for async metadata access. That is, I'm interested in solving the "AIO isn't really 'A'" problem.

Well, for that one blocking case in the submission path.

I certainly understand the desire to just make things less painful in this case, but I do wonder if the long-term win is to sink resources into trying to get threads executings async op so that the code paths don't have to be butchered into state machines around each potential blocking op.

It's a lot of work to get it transparent and efficient, I know. But then we'd be out of the business of obfuscating our code paths.

- z

#5 lore raw
Quoted 8 lines · Zach Brown

Well, for that one blocking case in the submission path.

I certainly understand the desire to just make things less painful in this case, but I do wonder if the long-term win is to sink resources into trying to get threads executings async op so that the code paths don't have to be butchered into state machines around each potential blocking op.

I only intend to do this for reads and non-allocating writes; in those cases, it's actually pretty easy. And if everything is already in cache, it's a lot easier to make things fast and not require diving into workqueue or some other kernel thread.

My concern with just stuffing every single io_submit into a kernel thread is the overhead involved, especially if there is a high rate of AIO submission (think high-speed PCIe-attached flash). Either you will end up blocking waiting for a kernel thread to be available, or you will burn a huge amount of kernel memory for the thread stacks if nothing else.

- Ted

#6 lore raw
Quoted 3 lines

I only intend to do this for reads and non-allocating writes; in those cases, it's actually pretty easy.

*nod*

Quoted 3 lines

My concern with just stuffing every single io_submit into a kernel thread is the overhead involved, especially if there is a high rate of AIO submission (think high-speed PCIe-attached flash).

Oh, I certainly agree. It'd be real work to find out if its possible to make it perform reasonably. I'd still like to see someone support that work so I'll keep mentioning it :).

- z

#7 lore raw
In reply to Ted Ts'o
Quoted 6 lines · Ted Ts'o
Quoted 5 lines · Dave Kleikamp

Committee members,

Please consider inviting me to the Storage, Filesystem, & MM Summit. I am working in Chris Mason's team at Oracle focusing on storage I/O.

Quoted 8 lines
Quoted 3 lines

My latest focus is on completing the work started on by Zach Brown to add a kernel interface to allow kernel code to submit aio, and having the loop device submit direct IO requests to the underlying file system.

We should talk; I'm currently working on an async bmap function for ext4, which I then intended to use for a modified high-level AIO support framework that allowed for async metadata access. That is, I'm interested in solving the "AIO isn't really 'A'" problem.

Hi Ted,

Async metadata accessing is on our TODO list too. Would you like to share more information to us? Maybe we can discuss this topic in ext4 developer's workshop. Thank you.

Regards,
Zheng

Quoted 12 lines

That's a different problem that what you're working on, but it's obviously going to touch the same code. (Actually, my original plan was to not touch the generic code in __blockdev_direct_IO and below and create some ext4-specific code for now, but if you're going to be messing around in that code path anyway, we should coordinate....)

- Ted
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html