Skip to contentSkip to conversation

readahead: partial sendfile fix

PATCH Ram Pai 2 participants 2 messages Feb 12, 2007 – Mar 8, 2007 1 message not in the archive

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

Ram Pai

author
#1 lore raw
Quoted 33 lines · Fengguang Wu

Enable readahead to handle partially done read requests, e.g.

        sendfile(188, 1921, [1478592], 19553028) = 37440
        sendfile(188, 1921, [1516032], 19515588) = 28800
        sendfile(188, 1921, [1544832], 19486788) = 37440
        sendfile(188, 1921, [1582272], 19449348) = 14400
        sendfile(188, 1921, [1596672], 19434948) = 37440
        sendfile(188, 1921, [1634112], 19397508) = 37440

In the above strace log,
- some lighttpd is doing _sequential_ reading
- every sendfile() returns with only _partial_ work done

page_cache_readahead() expects that if it returns @next_index, it will be
called exactly at @next_index next time. That's not true here. So the pattern
will be falsely recognized as a random read trace.

Also documented in "Linux AIO Performance and Robustness for Enterprise
Workloads" section 3.5:

sendfile(fd, 0, 2GB, fd2) = 8192,
tells readahead about up to 128KB of the read
sendfile(fd, 8192, 2GB - 8192, fd2) = 8192,
tells readahead about 8KB - 132KB of the read
sendfile(fd, 16384, 2GB - 16384, fd2) = 8192,
tells readahead about 16KB-140KB of the read
...
This confuses the readahead logic about the I/O pattern which
appears

        to be 0-128K, 8K-132K, 16K-140K instead of clear sequentiality
Quoted 8 lines

from
0-2GB that is really appropriate.

Retry based AIO shares the same read pattern and readahead problem. In this case, simply disabling readahead on restarted aio is not a good option:
we still need to call into readahead in the rare case of (req_size > ra_max).

The solution you proposed seems kludgy to me. If you determine that the its a restarted aio, then start reading from where readahead had left reading from earlier. To me a simple fix is:

- if (unlikely(aio_restarted()))
- next_index = last_index; /* Avoid repeat readahead */

+ if (unlikely(aio_restarted()))
+ next_index = min(prev_index+1, last_index);

No?
RP

Quoted 45 lines
---
 mm/filemap.c   |    3 ---
 mm/readahead.c |    9 +++++++++
 2 files changed, 9 insertions(+), 3 deletions(-)
mm/readahead.c +2 -0
@@ -581,6 +581,15 @@ page_cache_readahead(struct address_spac
        int sequential;
 
        /*
+        * A previous read request is partially completed,
+        * causing the retried/continued read calls into us

prematurely.

mm/readahead.c +4 -0
+        */
+       if (ra->start < offset &&
+                       offset < ra->prev_page &&
+                                ra->prev_page < ra->ahead_start +

ra->ahead_size)

mm/readahead.c +3 -0
+               goto out;
+
+       /*
         * We avoid doing extra work and bogusly perturbing the

readahead
* window expansion logic.
*/

mm/filemap.c +0 -3
@@ -915,9 +915,6 @@ void do_generic_mapping_read(struct addr
        if (!isize)
                goto out;
 
-       if (unlikely(aio_restarted()))
-               next_index = last_index; /* Avoid repeat readahead */
 
 
 
 
-
        end_index = (isize - 1) >> PAGE_CACHE_SHIFT;
        for (;;) {
                struct page *page;
#2 lore raw

Ram Pai,

Sorry for the long delay, I was just back from the winter vacation.

Quoted 1 line · Ram Pai

The solution you proposed seems kludgy to me. If you determine that the

I dislike it, either.

Quoted 11 lines

its a restarted aio, then start reading from where readahead had left reading from earlier. To me a simple fix is:

(quoted diff) +2 -2
-       if (unlikely(aio_restarted()))
-               next_index = last_index; /* Avoid repeat readahead */
 
+       if (unlikely(aio_restarted()))
+  		next_index = min(prev_index+1, last_index);

No?

Can be even simpler, if we _only_ want to fix the aio case:

+ if (unlikely(aio_restarted()))
+ next_index = prev_index + 1;

Regards,
Wu