Skip to contentSkip to conversation

Cleanup of aiodio runtest files

PATCH In review Cyril Hrubis 3 participants 9 messages Mar 19 – Mar 26

lore
Share & tools
lore
#0 lore raw

cover Cleanup of aiodio runtest files

The aiodio runtest files were added back in 2009 and since them not much was changed there. Given that these test as far as I can tell, does not find many bugs[1], we can reduce the runtime a bit.

This is a first patch that removes duplicate tests and limits the maximal file size for the sparse tests. If we agree to do so, I will look at the ltp-aiodio.part4 and try to parametrize the testruns with different sizes and number of workers, as we do with the rest of the files.

Also since we are touching these files, should we rename/split them as well? E.g. ltp-aiodio.part1 should be problably called aio-cp. The part2 aio-sparse, etc.

[1] I did SUSE bugzilla serach and there are two filesystem bugs found in 2019, one in 2018 and one in 2016. Also as far as I can tell in these cases the bug didn't seem to be strictly related to AIO and were also found by other tests.

Cyril Hrubis (2):
runtest/ltp-aidio.part4: Remove dio_sparse
runtest: ltp-aiodio.part2: Limit the maximal filesize

 runtest/ltp-aiodio.part2 | 16 ++++++++--------
 runtest/ltp-aiodio.part4 | 32 --------------------------------
 2 files changed, 8 insertions(+), 40 deletions(-)
Signature · 5 lines
2.52.0


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
#1 lore raw

PATCH 1/2 runtest/ltp-aidio.part4: Remove dio_sparse

After a discussion we agreed that there is a sufficient coverage for dio_sparse in ltp-aiodio.part2 already. There is no point in running these tests again in part4.

---
 runtest/ltp-aiodio.part4 | 32 --------------------------------
 1 file changed, 32 deletions(-)
runtest/ltp-aiodio.part4 +0 -32
index de00b8a78..5d3028bf5 100644
@@ -1,38 +1,6 @@
 aio01 aio01
 aio02 aio02
 
-#Running dio_sparse & dirty tests
-DI000 dirty
-DS000 dio_sparse
-DI001 dirty
-DS001 dio_sparse
-DI002 dirty
-DS002 dio_sparse
-DI003 dirty
-DS003 dio_sparse
-DI004 dirty
-DS004 dio_sparse
-DI005 dirty
-DS005 dio_sparse
-DI006 dirty
-DS006 dio_sparse
-DI007 dirty
-DS007 dio_sparse
-DI008 dirty
-DS008 dio_sparse
-DI009 dirty
-DS009 dio_sparse
-#iteration on dio_sparse
-DIO00 dio_sparse
-DIO01 dio_sparse
-DIO02 dio_sparse
-DIO03 dio_sparse
-DIO04 dio_sparse
-DIO05 dio_sparse
-DIO06 dio_sparse
-DIO07 dio_sparse
-DIO08 dio_sparse
-DIO09 dio_sparse
 #Running aiodio_append
 AD000 aiodio_append
 AD001 aiodio_append
Signature · 5 lines
2.52.0


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
Reviewed-by on v1 1/2
#2 lore raw

Re: [LTP] [PATCH 1/2] runtest/ltp-aidio.part4: Remove dio_sparse

Quoted 6 lines · Cyril Hrubis

After a discussion we agreed that there is a sufficient coverage for dio_sparse in ltp-aiodio.part2 already. There is no point in running these tests again in part4.

Looks good to me. Feel free to add:

Honza

Quoted 50 lines
---
 runtest/ltp-aiodio.part4 | 32 --------------------------------
 1 file changed, 32 deletions(-)
runtest/ltp-aiodio.part4 +0 -32
index de00b8a78..5d3028bf5 100644
@@ -1,38 +1,6 @@
 aio01 aio01
 aio02 aio02
 
-#Running dio_sparse & dirty tests
-DI000 dirty
-DS000 dio_sparse
-DI001 dirty
-DS001 dio_sparse
-DI002 dirty
-DS002 dio_sparse
-DI003 dirty
-DS003 dio_sparse
-DI004 dirty
-DS004 dio_sparse
-DI005 dirty
-DS005 dio_sparse
-DI006 dirty
-DS006 dio_sparse
-DI007 dirty
-DS007 dio_sparse
-DI008 dirty
-DS008 dio_sparse
-DI009 dirty
-DS009 dio_sparse
-#iteration on dio_sparse
-DIO00 dio_sparse
-DIO01 dio_sparse
-DIO02 dio_sparse
-DIO03 dio_sparse
-DIO04 dio_sparse
-DIO05 dio_sparse
-DIO06 dio_sparse
-DIO07 dio_sparse
-DIO08 dio_sparse
-DIO09 dio_sparse
 #Running aiodio_append
 AD000 aiodio_append
 AD001 aiodio_append
Signature · 1 line
2.52.0
Signature · 5 lines
Jan Kara <[email protected]>
SUSE Labs, CR

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
#4 lore raw

PATCH 2/2 runtest: ltp-aiodio.part2: Limit the maximal filesize

In reply to Cyril Hrubis

The test was operating with filesize up to ~500MB which, in certain cases, was making the runtime too slow. Some test variants, especially with many processes, were running up to 15 mintues on modern hardware. After a discussion we agreed that we can reduce the maximal file size to one quater without sacrificing test coverage.

I've also assumed that there was a typo in ADSP072 since the buffer size was set to be order of magnitude bigger than the file size.

---
 runtest/ltp-aiodio.part2 | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)
runtest/ltp-aiodio.part2 +8 -8
index 599c9fd2f..c82f6fd02 100644
@@ -25,8 +25,8 @@ ADSP023 aiodio_sparse -o 4 -w 1024k -s 4096k -n 6
 ADSP024 aiodio_sparse -o 4 -w 2048k -s 8192k -n 6
 ADSP025 aiodio_sparse -o 4 -w 4096k -s 16384k -n 6
 ADSP026 aiodio_sparse -o 4 -w 18192k -s 72768k -n 6
-ADSP027 aiodio_sparse -o 4 -w 18192k -s 518192k -n 6
-ADSP028 aiodio_sparse -o 4 -w 65536k -s 262144k -n 6
+ADSP027 aiodio_sparse -o 4 -w 18192k -s 129548k -n 6
+ADSP028 aiodio_sparse -o 4 -w 65536k -s 129548k -n 6
 ADSP029 aiodio_sparse -o 6 -w 65536k -n 6
 ADSP030 aiodio_sparse -o 8 -w 128k -s 1024k -n 6
 ADSP031 aiodio_sparse -o 16 -w 256k -s 4096k -n 6
@@ -55,7 +55,7 @@ ADSP053 dio_sparse  -w 2048k -s 2048k -n 2
 ADSP054 dio_sparse  -w 4096k -s 4096k -n 2
 ADSP055 dio_sparse  -w 8192k -s 8192k -n 2
 ADSP056 dio_sparse  -w 18192k -s 18192k -n 2
-ADSP057 dio_sparse  -w 518192k -s 518192k -n 2
+ADSP057 dio_sparse  -w 129548k -s 129548k -n 2
 ADSP058 dio_sparse  -w 58192k -s 58192k -n 4
 ADSP059 dio_sparse  -w 58192k -s 58192k -n 6
 ADSP060 dio_sparse  -w 256k -s 256k -n 6
@@ -65,14 +65,14 @@ ADSP063 dio_sparse  -w 2048k -s 2048k -n 6
 ADSP064 dio_sparse  -w 2048k -s 4096k -n 6
 ADSP065 dio_sparse  -w 8192k -s 8192k -n 6
 ADSP066 dio_sparse  -w 18192k -s 18192k -n 6
-ADSP067 dio_sparse  -w 58192k -s 518192k -n 6
-ADSP068 dio_sparse  -w 518192k -s 518192k -n 6
+ADSP067 dio_sparse  -w 58192k -s 129548k -n 6
+ADSP068 dio_sparse  -w 129548k -s 129548k -n 6
 ADSP069 dio_sparse  -w 1024k -s 2048k -n 6
 ADSP070 dio_sparse  -w 4096k -s 4096k -n 32
 ADSP071 dio_sparse  -w 8192k -s 8192k -n 64
-ADSP072 dio_sparse  -w 518192k -s 18192k -n 128
-ADSP073 dio_sparse  -w 518192k -s 518192k -n 512
-ADSP074 dio_sparse  -w 518192k -s 518192k -n 1000
+ADSP072 dio_sparse  -w 129548k -s 129548k -n 128
+ADSP073 dio_sparse  -w 129548k -s 129548k -n 512
+ADSP074 dio_sparse  -w 129548k -s 129548k -n 1000
 ADSP075 dio_sparse  -w 4k -s 2k -o 2k -n 2
 ADSP076 dio_sparse  -w 2k -s 1k -o 1k -n 2
 ADSP077 dio_sparse  -w 1k -s 512 -o 512 -n 2
Signature · 5 lines
2.52.0


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
#5 lore raw

Re: [LTP] [PATCH 2/2] runtest: ltp-aiodio.part2: Limit the maximal filesize

Quoted 20 lines · Cyril Hrubis

The test was operating with filesize up to ~500MB which, in certain cases, was making the runtime too slow. Some test variants, especially with many processes, were running up to 15 mintues on modern hardware. After a discussion we agreed that we can reduce the maximal file size to one quater without sacrificing test coverage.

I've also assumed that there was a typo in ADSP072 since the buffer size was set to be order of magnitude bigger than the file size.

---
 runtest/ltp-aiodio.part2 | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)
runtest/ltp-aiodio.part2 +0 -0
index 599c9fd2f..c82f6fd02 100644

runtest/ltp-aiodio.part2 ADSP023 aiodio_sparse -o 4 -w 1024k -s 4096k -n 6 · line 29

Quoted 7 lines
runtest/ltp-aiodio.part2 +2 -2
@@ -25,8 +25,8 @@ ADSP023 aiodio_sparse -o 4 -w 1024k -s 4096k -n 6
 ADSP024 aiodio_sparse -o 4 -w 2048k -s 8192k -n 6
 ADSP025 aiodio_sparse -o 4 -w 4096k -s 16384k -n 6
 ADSP026 aiodio_sparse -o 4 -w 18192k -s 72768k -n 6
-ADSP027 aiodio_sparse -o 4 -w 18192k -s 518192k -n 6
-ADSP028 aiodio_sparse -o 4 -w 65536k -s 262144k -n 6
+ADSP027 aiodio_sparse -o 4 -w 18192k -s 129548k -n 6
+ADSP028 aiodio_sparse -o 4 -w 65536k -s 129548k -n 6

I think the write size of 64MB is too large for 128MB file with 4 AIO contexts - I think the size will be autotrimmed to 32MB. So you might as well just specify -w 32768k.

Otherwise this looks good to me.

Honza

Signature · 5 lines
Jan Kara <[email protected]>
SUSE Labs, CR

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
#6 lore raw

Re: [LTP] [PATCH 2/2] runtest: ltp-aiodio.part2: Limit the maximal filesize

Hi!

Quoted 15 lines
Quoted 15 lines

The test was operating with filesize up to ~500MB which, in certain cases, was making the runtime too slow. Some test variants, especially with many processes, were running up to 15 mintues on modern hardware. After a discussion we agreed that we can reduce the maximal file size to one quater without sacrificing test coverage.

I've also assumed that there was a typo in ADSP072 since the buffer size was set to be order of magnitude bigger than the file size.

---
 runtest/ltp-aiodio.part2 | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)
Quoted 16 lines

runtest/ltp-aiodio.part2 ADSP023 aiodio_sparse -o 4 -w 1024k -s 4096k -n 6 · line 29

Quoted 12 lines
runtest/ltp-aiodio.part2 +2 -2
index 599c9fd2f..c82f6fd02 100644
@@ -25,8 +25,8 @@ ADSP023 aiodio_sparse -o 4 -w 1024k -s 4096k -n 6
 ADSP024 aiodio_sparse -o 4 -w 2048k -s 8192k -n 6
 ADSP025 aiodio_sparse -o 4 -w 4096k -s 16384k -n 6
 ADSP026 aiodio_sparse -o 4 -w 18192k -s 72768k -n 6
-ADSP027 aiodio_sparse -o 4 -w 18192k -s 518192k -n 6
-ADSP028 aiodio_sparse -o 4 -w 65536k -s 262144k -n 6
+ADSP027 aiodio_sparse -o 4 -w 18192k -s 129548k -n 6
+ADSP028 aiodio_sparse -o 4 -w 65536k -s 129548k -n 6

I think the write size of 64MB is too large for 128MB file with 4 AIO contexts - I think the size will be autotrimmed to 32MB. So you might as well just specify -w 32768k.

Right, will fix that.

Quoted 1 line

Otherwise this looks good to me.

Do I need to send v2 or can I add your Reviewed-by: if I fix the write size before applying the patches?

Signature · 5 lines
Cyril Hrubis
[email protected]

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
Reviewed-by
#7 lore raw

Re: [LTP] [PATCH 2/2] runtest: ltp-aiodio.part2: Limit the maximal filesize

Quoted 33 lines · Cyril Hrubis

Hi!

Quoted 15 lines
Quoted 15 lines

The test was operating with filesize up to ~500MB which, in certain cases, was making the runtime too slow. Some test variants, especially with many processes, were running up to 15 mintues on modern hardware. After a discussion we agreed that we can reduce the maximal file size to one quater without sacrificing test coverage.

I've also assumed that there was a typo in ADSP072 since the buffer size was set to be order of magnitude bigger than the file size.

---
 runtest/ltp-aiodio.part2 | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)
Quoted 16 lines

runtest/ltp-aiodio.part2 ADSP023 aiodio_sparse -o 4 -w 1024k -s 4096k -n 6 · line 29

Quoted 12 lines
runtest/ltp-aiodio.part2 +2 -2
index 599c9fd2f..c82f6fd02 100644
@@ -25,8 +25,8 @@ ADSP023 aiodio_sparse -o 4 -w 1024k -s 4096k -n 6
 ADSP024 aiodio_sparse -o 4 -w 2048k -s 8192k -n 6
 ADSP025 aiodio_sparse -o 4 -w 4096k -s 16384k -n 6
 ADSP026 aiodio_sparse -o 4 -w 18192k -s 72768k -n 6
-ADSP027 aiodio_sparse -o 4 -w 18192k -s 518192k -n 6
-ADSP028 aiodio_sparse -o 4 -w 65536k -s 262144k -n 6
+ADSP027 aiodio_sparse -o 4 -w 18192k -s 129548k -n 6
+ADSP028 aiodio_sparse -o 4 -w 65536k -s 129548k -n 6

I think the write size of 64MB is too large for 128MB file with 4 AIO contexts - I think the size will be autotrimmed to 32MB. So you might as well just specify -w 32768k.

Quoted 6 lines

Right, will fix that.

Quoted 1 line

Otherwise this looks good to me.

Do I need to send v2 or can I add your Reviewed-by: if I fix the write size before applying the patches?

Feel free to add:

one that one is fixed.

Honza

Signature · 5 lines
Jan Kara <[email protected]>
SUSE Labs, CR

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp