Re: Block layer projects that I haven't had time for

[Ming Lin]

On Mon, Dec 8, 2014 at 3:48 AM, Dongsu Park
<dongsu.park@xxxxxxxxxxxxxxxx> wrote:
> Thanks for the reply.
>
> On 05.12.2014 19:02, Kent Overstreet wrote:
>> On Thu, Dec 04, 2014 at 12:00:27PM +0100, Dongsu Park wrote:
>> > Playing a little with your block_stuff tree based on 3.15, however,
>> > I think there still seems to be a couple of issues.
>> > First of all, it doesn't work with virtio-blk. A testing Qemu VM panics
>> > at the very early stage of booting. This issue should be addressed as
>> > the first step, so that other parts can be tested.
>>
>> Really? I was testing with virtio-blk, that's odd..
>
> The culprit seems to be the plugging commit.
> Before that change, it works well also with virtio-blk.
> Though that's not the only issue...
>
>> > Moreover, I've already tried to rebase these patches on top of current
>> > mainline, 3.18-rc7. It's now compilable, but it seems to introduce
>> > more bugs about direct-IO. I didn't manage to find out the reason.
>> > I'd need to also look at the previous review comments in [1], [2].
>> >
>> > Don't you have other trees based on top of 3.17 or higher?
>> > If not, can I create my own tree based on 3.18-rc7 to publish?
>>
>> Yeah, I'd post what you have now and I'll try and take a look.
>
> I've created a git tree to include what I have right now.
> Please see <https://github.com/dongsupark/linux>.
>
> To be able to handle different issues one by one,
> I got the entire tree separated out into 4 branches based on 3.18.
>
> * block-generic-req-for-next : the most stable branch you can test with.
>   With this branch, you can test most of block drivers as well as file
>   systems with less critical bugs. Though it's not 100% perfect yet,
>   e.g. btrfs doesn't seem to work quite well. Thus more tests are needed.
>
> * block-mpage-bvecs-for-next : block-generic-req-for-next + multipage bvecs.
>   This branch shows a critical issue that writing blocks to ext4 rootfs
>   causes the whole system to crash. Need-to-investigate.

I tried block-mpage-bvecs-for-next branch on qemu-kvm with ext4 rootfs.
Run "sync" will stuck in kernel.

[  480.751901] INFO: task sync:4424 blocked for more than 120 seconds.
[  480.753064]       Not tainted 3.18.0-00025-g46c8231 #39
[  480.753720] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs"
disables this message.
[  480.754737] sync            D ffff88001fc11180     0  4424   4338 0x00000000
[  480.755719]  ffff88001cdfbc98 0000000000000086 ffff88001cdfbba8
ffff880014adefc0
[  480.756810]  0000000000011180 0000000000004000 ffffffff81813460
ffff880014adefc0
[  480.758102]  ffff88001cdfbbc8 ffffffff812f08be ffff88001cdfbc18
ffff880014adf028
[  480.759454] Call Trace:
[  480.759852]  [<ffffffff812f08be>] ? debug_smp_processor_id+0x17/0x19
[  480.760609]  [<ffffffff8106093e>] ? __enqueue_entity+0x69/0x6b
[  480.761318]  [<ffffffff8106017e>] ? __dequeue_entity+0x33/0x38
[  480.762026]  [<ffffffff810601ab>] ? set_next_entity+0x28/0x7d
[  480.762739]  [<ffffffff8105a4fb>] ? get_parent_ip+0xf/0x3f
[  480.763425]  [<ffffffff8108562b>] ? ktime_get+0x50/0x8f
[  480.763848]  [<ffffffff8148abdb>] ? bit_wait_timeout+0x60/0x60
[  480.764555]  [<ffffffff8148a6be>] schedule+0x6a/0x6c
[  480.765186]  [<ffffffff8148a74f>] io_schedule+0x8f/0xcd
[  480.765841]  [<ffffffff8148ac19>] bit_wait_io+0x3e/0x42
[  480.766493]  [<ffffffff8148ae80>] __wait_on_bit+0x4d/0x86
[  480.767183]  [<ffffffff810d4302>] ? find_get_pages_tag+0x106/0x133
[  480.767847]  [<ffffffff810d4a63>] wait_on_page_bit+0x76/0x78
[  480.768532]  [<ffffffff8106ab59>] ? wake_atomic_t_function+0x2d/0x2d
[  480.769262]  [<ffffffff810d511f>] filemap_fdatawait_range+0x7e/0x11d
[  480.769992]  [<ffffffff8148a639>] ? preempt_schedule+0x36/0x51
[  480.770677]  [<ffffffff8105a4fb>] ? get_parent_ip+0xf/0x3f
[  480.771848]  [<ffffffff810d51df>] filemap_fdatawait+0x21/0x23
[  480.772530]  [<ffffffff811458ce>] sync_inodes_sb+0x158/0x1aa
[  480.773201]  [<ffffffff81480303>] ? br_mdb_dump+0x225/0x495
[  480.773885]  [<ffffffff81149ad8>] ? fdatawrite_one_bdev+0x18/0x18
[  480.774592]  [<ffffffff81149aec>] sync_inodes_one_sb+0x14/0x16
[  480.775278]  [<ffffffff81125937>] iterate_supers+0x6f/0xc4
[  480.775847]  [<ffffffff81149bf4>] sys_sync+0x35/0x83
[  480.776460]  [<ffffffff8148da52>] system_call_fastpath+0x12/0x17


Here is a quick hack.

diff --git a/block/bio.c b/block/bio.c
index 4020ccc..fbc7108 100644
--- a/block/bio.c
+++ b/block/bio.c
@@ -829,6 +829,11 @@ int bio_add_page(struct bio *bio, struct page *page,
                if (bvec_to_phys(bv) + bv->bv_len ==
                    page_to_phys(page) + offset) {
                        bv->bv_len += len;
+                       /*
+                        * Page is not added to bio vec.
+                        * Clear PG_writeback so
filemap_fdatawait_range() won't wait for it.
+                        */
+                       TestClearPageWriteback(page);
                        goto done;
                }
        }

Thanks,
Ming
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/