View previous topic :: View next topic |
Author |
Message |
pross Apprentice
Joined: 30 Jul 2003 Posts: 156 Location: Ireland
|
Posted: Tue Feb 22, 2011 6:42 pm Post subject: |
|
|
bingyuan wrote: | Would you please make ck's BFS alternative? |
you can de-select in menuconfig _________________ My Overlay [pross] | bfs-sources |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
|
Back to top |
|
|
bingyuan n00b
Joined: 20 Feb 2011 Posts: 13
|
Posted: Wed Feb 23, 2011 12:29 am Post subject: |
|
|
kernelOfTruth wrote: | pross wrote: | bingyuan wrote: | Would you please make ck's BFS alternative? |
you can de-select in menuconfig |
it's not included (yet)
I added it but it wouldn't work
so I have to remake the whole patch(set)
it probably will make it in in the next version |
Thx for your work! |
|
Back to top |
|
|
SlashBeast Retired Dev
Joined: 23 May 2006 Posts: 2922
|
Posted: Fri Feb 25, 2011 8:27 am Post subject: |
|
|
can you provide standalone patches for SMP in dmcrypt? I was experimenting with some patches from devicemaper patchwork however with no much luck. (compiled, booted, no speedup, random freezes).
Also the fixes for ext4 and zcache would be cool. |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
Posted: Fri Feb 25, 2011 2:56 pm Post subject: |
|
|
SlashBeast wrote: | can you provide standalone patches for SMP in dmcrypt? I was experimenting with some patches from devicemaper patchwork however with no much luck. (compiled, booted, no speedup, random freezes).
Also the fixes for ext4 and zcache would be cool. |
* Patchwork dm crypt: scale to multiple CPUs
* zcache: look up the patch-names on lkml or spinics.net, marc.info, ...
zcache-linux-2.6.37-110205
[PATCH] staging: zcache: fix memory leak
after that apply: 2.6.37_plus_v16_zcache_linux-next_fixes.patch
and zcache_[PATCH] Allow sharing xvmalloc for zram and zcache
* ext4 fixes can be taken from the list: look the up on lkml, spinics, marc.info, etc.
I'm currently updating the patch(set) & fixing a bug which prevents it from booting on ext4 & BFS from working _________________ https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa
Hardcore Gentoo Linux user since 2004 |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
Posted: Fri Feb 25, 2011 5:14 pm Post subject: |
|
|
@bingyuan:
see :
Update10:
Quote: | testing:
* currently only BFS & CFS (with zcache) builds patches available, tarball with broken-out patches, several diff-patches will be available soon
* compile-tested - boot-tested (the one with compaction)
* includes zcache but not TOI
* if there are problems on 32 bit try the one without compaction
bump to 2.6.37.2_plus_v1
reverted (from 2.6.37.1): [PATCH] Change wait_for_completion_*_timeout to return a signed long to make BFS compile
apply the patches on top of 2.6.37.2
modular design
(cumulative optional patches for your convenience)
base-patch:
2.6.37.2_plus_v1-base-non_BFS-compaction-CFS-TOI-zcache-btrfs_new.patch (3.15 MB)
• 0 custom-flags
• 1 BFQ
• 2 dm-crypt multi-cpu
• 3 fadvise (v6), for rsync
• 4 ck-patchset
• 5 zram (+fixes) + zram & xvmalloc 64K page fixes and optimizations
• 6 incorrect unlikely & likely cleanups
• 7 io-less dirty throttling
• 8 vmscan: protect exectuable page from inactive list scan
• 9 mmu-preemptible (v6)
• 10 jhash 3
• 11 mm: Avoid possible deadlock caused by too_many_isolated()
• 12 fix __set_page_dirty_no_writeback() return value
• 13 some cfq, pm, reiserfs, gcc 4.5 inlining fixes & more
• 14 SFB network scheduler
• 15 squashfs + lzma support
• 16 usb mouse polling
• .
• 18 kswapd
• .
• 21 ext4 fixes
• .
• 23 memmove, cfq, smp, rtmutex fixes; usb-storage support for ums-realtek
• 24 ureadahead, radix_tree fixes
• .
• 26 coordinated flush requests
• 27 delayed tickless during load, ksoftirq fixes, uvesa/vesafb fixes; "remove 8 bytes of padding from block_device on 64bit builds"
• 28 inode data integrity patches [probably needs to be merged manually, when not applying patches in ladder order]
• 29 reiser4
• 30 slqb
• .
• 32 linux-phc
• 33 Improve drain pages performance on large systems; fbcondecor
optional:
• memory compaction
• BFS, CFS [both are mutually exclusive]
• some 2.6.38 fixes & changes
• TOI [zcache and TOI are mutually exclusive]
• latest btrfs fixes & changes
|
_________________ https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa
Hardcore Gentoo Linux user since 2004 |
|
Back to top |
|
|
raj7095 n00b
Joined: 08 Feb 2011 Posts: 59
|
Posted: Fri Feb 25, 2011 9:47 pm Post subject: |
|
|
sweet. that is amazing. |
|
Back to top |
|
|
raj7095 n00b
Joined: 08 Feb 2011 Posts: 59
|
Posted: Fri Feb 25, 2011 10:51 pm Post subject: |
|
|
odd, i got an error while compiling the cfs version. it says: kernel/sched.c:3280: error: implicit declaration of function ‘TIMER_DEFERRED_INITIALIZER’ |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
|
Back to top |
|
|
tranquilcool Veteran
Joined: 25 Mar 2005 Posts: 1179
|
Posted: Fri Feb 25, 2011 11:58 pm Post subject: |
|
|
sorry but page 1 is real confusion. who knows which patch is which.
i'll wait until things are cleared. _________________ this is a strange strange world. |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
|
Back to top |
|
|
raj7095 n00b
Joined: 08 Feb 2011 Posts: 59
|
Posted: Sat Feb 26, 2011 12:49 am Post subject: |
|
|
i am using the one WITH compaction and zcache. the version that isnt cfs from 2.6.38 |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
|
Back to top |
|
|
raj7095 n00b
Joined: 08 Feb 2011 Posts: 59
|
Posted: Sat Feb 26, 2011 1:19 am Post subject: |
|
|
it compiles cleanly now. don't get that error anymore. |
|
Back to top |
|
|
raj7095 n00b
Joined: 08 Feb 2011 Posts: 59
|
Posted: Sat Feb 26, 2011 1:37 am Post subject: |
|
|
cfs from 2.6.38 makes my cursor feel a little smoother than 2.6.37. weird thing is cfs with additional stuff compiles fine without doing anything. anyways, on a side note, i now figured out why bfs in zen kernel felt slower. it was just zen tune settings. it turns out lowering rr_interval than the default setting leads to a much lower throughput. so, once we selected desktop profile, it slowed down the entire kernel because rr_interval is set way too low. thats why your kernel with bfs felt much faster compared to zen kernel, but when i lowered the rr_interval in your kernel, it slowed down as well. |
|
Back to top |
|
|
tranquilcool Veteran
Joined: 25 Mar 2005 Posts: 1179
|
Posted: Sat Feb 26, 2011 2:54 am Post subject: |
|
|
with either
* 2.6.37.2_plus_v1_BFS_compaction_zcache_new.patch (3.53 MB)
OR
* 2.6.37.2_plus_v1_BFS_no-compaction_zcache.patch (3.5 MB)
no zcache reported in dmesg.
./zcache.stats gives the following errors;
/sys/kernel/mm/cleancache/cleancache_puts : 0
/sys/kernel/mm/cleancache/cleancache_succ_gets : 0
/sys/kernel/mm/cleancache/cleancache_failed_gets : 0
/sys/kernel/mm/frontswap/curr_pages : 0
/sys/kernel/mm/frontswap/succ_puts : 0
/sys/kernel/mm/frontswap/gets : 0
/sys/kernel/mm/zcache/curr_obj_count : cat: /sys/kernel/mm/zcache/curr_obj_count: No such file or directory
/sys/kernel/mm/zcache/curr_obj_count_max : cat: /sys/kernel/mm/zcache/curr_obj_count_max: No such file or directory
/sys/kernel/mm/zcache/curr_objnode_count : cat: /sys/kernel/mm/zcache/curr_objnode_count: No such file or directory
/sys/kernel/mm/zcache/curr_objnode_count_max : cat: /sys/kernel/mm/zcache/curr_objnode_count_max: No such file or directory
/sys/kernel/mm/zcache/flush_total : cat: /sys/kernel/mm/zcache/flush_total: No such file or directory
/sys/kernel/mm/zcache/flush_found : cat: /sys/kernel/mm/zcache/flush_found: No such file or directory
/sys/kernel/mm/zcache/flobj_total : cat: /sys/kernel/mm/zcache/flobj_total: No such file or directory
/sys/kernel/mm/zcache/flobj_found : cat: /sys/kernel/mm/zcache/flobj_found: No such file or directory
/sys/kernel/mm/zcache/failed_eph_puts : cat: /sys/kernel/mm/zcache/failed_eph_puts: No such file or directory
/sys/kernel/mm/zcache/failed_pers_puts : cat: /sys/kernel/mm/zcache/failed_pers_puts: No such file or directory
/sys/kernel/mm/zcache/zbud_curr_raw_pages : cat: /sys/kernel/mm/zcache/zbud_curr_raw_pages: No such file or directory
/sys/kernel/mm/zcache/zbud_curr_zpages : cat: /sys/kernel/mm/zcache/zbud_curr_zpages: No such file or directory
/sys/kernel/mm/zcache/zbud_curr_zbytes : cat: /sys/kernel/mm/zcache/zbud_curr_zbytes: No such file or directory
/sys/kernel/mm/zcache/zbud_cumul_zpages : cat: /sys/kernel/mm/zcache/zbud_cumul_zpages: No such file or directory
/sys/kernel/mm/zcache/zbud_cumul_zbytes : cat: /sys/kernel/mm/zcache/zbud_cumul_zbytes: No such file or directory
/sys/kernel/mm/zcache/evicted_raw_pages : cat: /sys/kernel/mm/zcache/evicted_raw_pages: No such file or directory
/sys/kernel/mm/zcache/evicted_unbuddied_pages : cat: /sys/kernel/mm/zcache/evicted_unbuddied_pages: No such file or directory
/sys/kernel/mm/zcache/evicted_buddied_pages : cat: /sys/kernel/mm/zcache/evicted_buddied_pages: No such file or directory
/sys/kernel/mm/zcache/failed_alloc : cat: /sys/kernel/mm/zcache/failed_alloc: No such file or directory
/sys/kernel/mm/zcache/failed_get_free_pages : cat: /sys/kernel/mm/zcache/failed_get_free_pages: No such file or directory
/sys/kernel/mm/zcache/compress_poor : cat: /sys/kernel/mm/zcache/compress_poor: No such file or directory
/sys/kernel/mm/zcache/put_to_flush : cat: /sys/kernel/mm/zcache/put_to_flush: No such file or directory
/sys/kernel/mm/zcache/aborted_preload : cat: /sys/kernel/mm/zcache/aborted_preload: No such file or directory
/sys/kernel/mm/zcache/aborted_shrink : cat: /sys/kernel/mm/zcache/aborted_shrink: No such file or directory
/sys/kernel/mm/zcache/zbud_buddied_count : cat: /sys/kernel/mm/zcache/zbud_buddied_count: No such file or directory
/sys/kernel/mm/zcache/zbpg_unused_list_count : cat: /sys/kernel/mm/zcache/zbpg_unused_list_count: No such file or directory
/sys/kernel/mm/zcache/zbud_unbuddied_list_counts : cat: /sys/kernel/mm/zcache/zbud_unbuddied_list_counts: No such file or directory
/sys/kernel/mm/zcache/zbud_cumul_chunk_counts : cat: /sys/kernel/mm/zcache/zbud_cumul_chunk_counts: No such file or directory
MemTotal: 3373800 kB
MemFree: 2091672 kB
Active(file): 173556 kB
Inactive(file): 473136 kB
pgpgin 617655
pgpgout 21144
pswpin 0
pswpout 0 _________________ this is a strange strange world. |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
|
Back to top |
|
|
tranquilcool Veteran
Joined: 25 Mar 2005 Posts: 1179
|
Posted: Sat Feb 26, 2011 8:03 am Post subject: |
|
|
kernelOfTruth wrote: | tranquilcool wrote: | with either
* 2.6.37.2_plus_v1_BFS_compaction_zcache_new.patch (3.53 MB)
OR
* 2.6.37.2_plus_v1_BFS_no-compaction_zcache.patch (3.5 MB)
no zcache reported in dmesg.
./zcache.stats gives the following errors;
[snip] |
yeah,
thanks for reporting !
the whole patchset is suffering from this problem
seems like a patch didn't play well
I also had rubbed my eyes in astonishment why there wasn't any mentioning of zcache in the logs:
|
problem solved. zcache now reports in logs.
thanks. _________________ this is a strange strange world. |
|
Back to top |
|
|
bingyuan n00b
Joined: 20 Feb 2011 Posts: 13
|
Posted: Sat Feb 26, 2011 9:37 am Post subject: |
|
|
@kernelOfTruth,
thx, everything goes well! |
|
Back to top |
|
|
bingyuan n00b
Joined: 20 Feb 2011 Posts: 13
|
Posted: Sat Feb 26, 2011 2:36 pm Post subject: |
|
|
@kernelOfTruth, get an error/bug(with 2.6.37.2_plus_v1_BFS_compaction_zcache_new.patch, 32bit):
When I build the kernel with
Code: | [*] Allow for memory compaction | Then I can't access my windows patitions, the error info Quote: | bingyuan@Tux ~ $ ls /media/sda6
ls: reading directory /media/sda6: Cannot allocate memory | but this windows patition is mounted Quote: | bingyuan@Tux ~ $ mount | grep sda6
/dev/sda6 on /media/sda6 type fuseblk (rw,nosuid,nodev,allow_other,blksize=4096) |
Then I try this(leave other configs unchanged) Code: | [ ] Allow for memory compaction | now everything goes well! |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
Posted: Sat Feb 26, 2011 3:24 pm Post subject: |
|
|
bingyuan wrote: | @kernelOfTruth, get an error/bug(with 2.6.37.2_plus_v1_BFS_compaction_zcache_new.patch, 32bit):
When I build the kernel with
Code: | [*] Allow for memory compaction | Then I can't access my windows patitions, the error info Quote: | bingyuan@Tux ~ $ ls /media/sda6
ls: reading directory /media/sda6: Cannot allocate memory | but this windows patition is mounted Quote: | bingyuan@Tux ~ $ mount | grep sda6
/dev/sda6 on /media/sda6 type fuseblk (rw,nosuid,nodev,allow_other,blksize=4096) |
Then I try this(leave other configs unchanged) Code: | [ ] Allow for memory compaction | now everything goes well! |
thanks for reporting, bingyuan !
I haven't tested ntfs-3g yet, will see if that happens, too, on 64 bit
it seems, it's a known problem that 32bit + memory compaction don't play well at the moment
this will get fixed upstream with 2.6.39 (?) when they are merged into mainline
afaik Steven from zen-kernel has reverted memory compaction for this very reason from zen
I haven't looked through the code but to be sure that your memory doesn't get heavily fragmented when not using memory compaction (and having those modifications in your source):
could you please use:
"2.6.37.2_plus_v1_BFS_no-compaction_zcache.patch" ?
instead of 2.6.37.2_plus_v1_BFS_compaction_zcache_new.patch
that's the reason why I created that no-compaction (no memory compaction) patches for you 32bit folks
and that's also the only difference between those - so you won't lose any other features _________________ https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa
Hardcore Gentoo Linux user since 2004 |
|
Back to top |
|
|
Ant P. Watchman
Joined: 18 Apr 2009 Posts: 6920
|
Posted: Sat Feb 26, 2011 5:26 pm Post subject: |
|
|
Is there any chance of adding the transparent-hugepages patches? I dunno if it's a safe idea, what with all the other memory patches here already, but I wanted to try it... |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
|
Back to top |
|
|
Ant P. Watchman
Joined: 18 Apr 2009 Posts: 6920
|
Posted: Sat Feb 26, 2011 8:36 pm Post subject: |
|
|
Trying to build with the no-compaction patch (on its own) and this happened:
Code: | GEN .version
CHK include/generated/compile.h
UPD include/generated/compile.h
CC init/version.o
LD init/built-in.o
LD .tmp_vmlinux1
kernel/built-in.o: In function `cpu_stop_cpu_callback':
stop_machine.c:(.cpuinit.text+0xc22): undefined reference to `sched_set_stop_task'
make: *** [.tmp_vmlinux1] Error 1 |
.config |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
Posted: Sat Feb 26, 2011 9:05 pm Post subject: |
|
|
Ant_P wrote: | Trying to build with the no-compaction patch (on its own) and this happened:
Code: | GEN .version
CHK include/generated/compile.h
UPD include/generated/compile.h
CC init/version.o
LD init/built-in.o
LD .tmp_vmlinux1
kernel/built-in.o: In function `cpu_stop_cpu_callback':
stop_machine.c:(.cpuinit.text+0xc22): undefined reference to `sched_set_stop_task'
make: *** [.tmp_vmlinux1] Error 1 |
.config |
for reference:
so you're trying to compile 2.6.37.2_plus_v1_BFS_no-compaction_zcache.patch :
any modifications beside that ? _________________ https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa
Hardcore Gentoo Linux user since 2004 |
|
Back to top |
|
|
|