Ask help about this patch b951aaff5035 " mm: enable page allocation tagging"

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi, all

 

I found a 300ms~600ms IRQ off when writing 1Gb data to mmc device at I.MX7d SDB board at Linux-kernel-v6.14.

But I test the same case at Linux-kernel-v6.7, this longest IRQ off time is only 1ms~2ms. So the issue is introduced from v6.7~v6.14.

 

Run this cmd to test:

dd if=/dev/zero of=/dev/mmcblk2p4 bs=4096 seek=12500 count=256000 conv=fsync

 

This issue looks from blkdev_buffered_write() function. Because when I run this cmd with “oflag=direct” to use

blkdev_direct_write(), I can not see any long time IRQ off.

 

Then I use Ftrace irqoff tracer to trace the longest IRQ off event, I found some differences between v6.7 and v6.14:

In iomap_file_buffered_write(), __folio_alloc (in v6.7) is replaced by _folio_alloc_noprof (in v6.14) by this patch.

The spinlock disabled IRQ ~300ms+. It looks there are some fixes for this patch, but I still can see IRQ off 300ms+ at 6.14.0-rc7-next-20250319.

 

Do I trigger one bug? I know little about mem so I have to report it and hope I can get some help or guide.

I put my ftrace log at the mail tail to help trace and explain.

 

commit b951aaff503502a7fe066eeed2744ba8a6413c89

Author: Suren Baghdasaryan surenb@xxxxxxxxxx

Date:   Thu Mar 21 09:36:40 2024 -0700

 

    mm: enable page allocation tagging

 

    Redefine page allocators to record allocation tags upon their invocation.

    Instrument post_alloc_hook and free_pages_prepare to modify current

    allocation tag.

 

    [surenb@xxxxxxxxxx: undo _noprof additions in the documentation]

      Link: https://lkml.kernel.org/r/20240326231453.1206227-3-surenb@xxxxxxxxxx

    Link: https://lkml.kernel.org/r/20240321163705.3067592-19-surenb@xxxxxxxxxx

    Signed-off-by: Suren Baghdasaryan surenb@xxxxxxxxxx

    Co-developed-by: Kent Overstreet kent.overstreet@xxxxxxxxx

    Signed-off-by: Kent Overstreet kent.overstreet@xxxxxxxxx

    Reviewed-by: Kees Cook keescook@xxxxxxxxxxxx

    Tested-by: Kees Cook keescook@xxxxxxxxxxxx

    Cc: Alexander Viro viro@xxxxxxxxxxxxxxxxxx

    Cc: Alex Gaynor alex.gaynor@xxxxxxxxx

    Cc: Alice Ryhl aliceryhl@xxxxxxxxxx

    Cc: Andreas Hindborg a.hindborg@xxxxxxxxxxx

    Cc: Benno Lossin benno.lossin@xxxxxxxxx

    Cc: "Björn Roy Baron" bjorn3_gh@xxxxxxxxxxxxxx

    Cc: Boqun Feng boqun.feng@xxxxxxxxx

    Cc: Christoph Lameter cl@xxxxxxxxx

    Cc: Dennis Zhou dennis@xxxxxxxxxx

    Cc: Gary Guo gary@xxxxxxxxxxx

    Cc: Miguel Ojeda ojeda@xxxxxxxxxx

    Cc: Pasha Tatashin pasha.tatashin@xxxxxxxxxx

    Cc: Peter Zijlstra peterz@xxxxxxxxxxxxx

    Cc: Tejun Heo tj@xxxxxxxxxx

    Cc: Vlastimil Babka vbabka@xxxxxxx

    Cc: Wedson Almeida Filho wedsonaf@xxxxxxxxx

    Signed-off-by: Andrew Morton akpm@xxxxxxxxxxxxxxxxxxxx

 

 

Ftrace irqoff tracer shows detail:

At v6.14:

# tracer: irqsoff

#

# irqsoff latency trace v1.1.5 on 6.14.0-rc7-next-20250319

# --------------------------------------------------------------------

# latency: 279663 us, #21352/21352, CPU#0 | (M:NONE VP:0, KP:0, SP:0 HP:0 #P:2)

#    -----------------

#    | task: dd-805 (uid:0 nice:0 policy:0 rt_prio:0)

#    -----------------

#  => started at: __rmqueue_pcplist

#  => ended at:   _raw_spin_unlock_irqrestore

#

#

#                    _------=> CPU#           

#                   / _-----=> irqs-off/BH-disabled

#                  | / _----=> need-resched   

#                  || / _---=> hardirq/softirq

#                  ||| / _--=> preempt-depth  

#                  |||| / _-=> migrate-disable

#                  ||||| /     delay          

#  cmd     pid     |||||| time  |   caller    

#     \   /        ||||||  \    |    /      

      dd-805       0d....    1us : __rmqueue_pcplist

      dd-805       0d....    3us : _raw_spin_trylock <-__rmqueue_pcplist

      dd-805       0d....    7us : __mod_zone_page_state <-__rmqueue_pcplist

      dd-805       0d....   10us : __mod_zone_page_state <-__rmqueue_pcplist

      dd-805       0d....   12us : __mod_zone_page_state <-__rmqueue_pcplist

      dd-805       0d....   15us : __mod_zone_page_state <-__rmqueue_pcplist

      dd-805       0d....   17us : __mod_zone_page_state <-__rmqueue_pcplist

      dd-805       0d....   19us : __mod_zone_page_state <-__rmqueue_pcplist

   …

      dd-805       0d.... 1535us : find_suitable_fallback <-__rmqueue_pcplist

      dd-805       0d.... 1538us : find_suitable_fallback <-__rmqueue_pcplist

      dd-805       0d.... 1539us : find_suitable_fallback <-__rmqueue_pcplist

      dd-805       0d.... 1542us+: try_to_claim_block <-__rmqueue_pcplist

      dd-805       0d.... 1597us : find_suitable_fallback <-__rmqueue_pcplist

      dd-805       0d.... 1599us+: try_to_claim_block <-__rmqueue_pcplist

      dd-805       0d.... 1674us : find_suitable_fallback <-__rmqueue_pcplist

      dd-805       0d.... 1676us+: try_to_claim_block <-__rmqueue_pcplist

      dd-805       0d.... 1716us : find_suitable_fallback <-__rmqueue_pcplist

      dd-805       0d.... 1718us+: try_to_claim_block <-__rmqueue_pcplist

      dd-805       0d.... 1801us : find_suitable_fallback <-__rmqueue_pcplist

      dd-805       0d.... 1803us+: try_to_claim_block <-__rmqueue_pcplist

     dd-805       0d.... 279555us : find_suitable_fallback <-__rmqueue_pcplist

      dd-805       0d.... 279556us : find_suitable_fallback <-__rmqueue_pcplist

      dd-805       0d.... 279558us : find_suitable_fallback <-__rmqueue_pcplist

      dd-805       0d.... 279560us+: try_to_claim_block <-__rmqueue_pcplist

      dd-805       0d.... 279616us : find_suitable_fallback <-__rmqueue_pcplist

      dd-805       0d.... 279618us : __mod_zone_page_state <-__rmqueue_pcplist

      dd-805       0d.... 279620us : find_suitable_fallback <-__rmqueue_pcplist

      dd-805       0d.... 279658us : find_suitable_fallback <-__rmqueue_pcplist

      dd-805       0d.... 279660us : _raw_spin_unlock_irqrestore <-__rmqueue_pcplist

      dd-805       0d.... 279662us : _raw_spin_unlock_irqrestore

      dd-805       0d.... 279666us+: trace_hardirqs_on <-_raw_spin_unlock_irqrestore

      dd-805       0d.... 279712us : <stack trace>

=> get_page_from_freelist

=> __alloc_frozen_pages_noprof

=> __folio_alloc_noprof

=> __filemap_get_folio

=> iomap_write_begin

=> iomap_file_buffered_write

=> blkdev_write_iter

=> vfs_write

=> ksys_write

=> ret_fast_syscall

 

At v6.7:

# tracer: irqsoff

#

# irqsoff latency trace v1.1.5 on 6.7.0

# --------------------------------------------------------------------

# latency: 2477 us, #146/146, CPU#0 | (M:server VP:0, KP:0, SP:0 HP:0 #P:2)

#    -----------------

#    | task: dd-808 (uid:0 nice:0 policy:0 rt_prio:0)

#    -----------------

#  => started at: _raw_spin_lock_irqsave

#  => ended at:   _raw_spin_unlock_irqrestore

#

#

#                    _------=> CPU#           

#                   / _-----=> irqs-off/BH-disabled

#                  | / _----=> need-resched   

#                  || / _---=> hardirq/softirq

#                  ||| / _--=> preempt-depth  

#                  |||| / _-=> migrate-disable

#                  ||||| /     delay          

#  cmd     pid     |||||| time  |   caller    

#     \   /        ||||||  \    |    /      

      dd-808       0d....    1us!: _raw_spin_lock_irqsave

      dd-808       0d....  186us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  189us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  191us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  192us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  194us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  196us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  199us : steal_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  203us!: move_freepages_block <-steal_suitable_fallback

      dd-808       0d....  330us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  332us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  334us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  336us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  338us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  339us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  341us : steal_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  343us!: move_freepages_block <-steal_suitable_fallback

      dd-808       0d....  479us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  481us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  483us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  485us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  486us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  488us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  490us : steal_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  492us!: move_freepages_block <-steal_suitable_fallback

      dd-808       0d....  630us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  632us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  634us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  636us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  638us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  640us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  642us : steal_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  644us!: move_freepages_block <-steal_suitable_fallback

      dd-808       0d....  771us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  773us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  775us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  777us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  778us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  780us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  782us : steal_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  784us!: move_freepages_block <-steal_suitable_fallback

      dd-808       0d....  911us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  913us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  915us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  916us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  918us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  920us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  922us : steal_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d....  924us!: move_freepages_block <-steal_suitable_fallback

      dd-808       0d.... 1055us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1058us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1059us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1061us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1063us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1065us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1066us : steal_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1068us!: move_freepages_block <-steal_suitable_fallback

      dd-808       0d.... 1194us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1196us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1198us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1200us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1202us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1203us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1205us : steal_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1208us!: move_freepages_block <-steal_suitable_fallback

      dd-808       0d.... 1333us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1335us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1337us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1339us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1341us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1342us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1344us : steal_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1346us!: move_freepages_block <-steal_suitable_fallback

      dd-808       0d.... 1480us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1482us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1484us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1486us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1488us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1490us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1492us : steal_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1494us!: move_freepages_block <-steal_suitable_fallback

      dd-808       0d.... 1621us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1623us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1625us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1627us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1629us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1630us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1632us : steal_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1634us!: move_freepages_block <-steal_suitable_fallback

      dd-808       0d.... 1761us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1763us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1765us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1766us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1768us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1770us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1772us : steal_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1774us!: move_freepages_block <-steal_suitable_fallback

      dd-808       0d.... 1900us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1902us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1903us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1905us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1907us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1909us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1911us : steal_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 1913us!: move_freepages_block <-steal_suitable_fallback

      dd-808       0d.... 2038us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2040us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2042us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2044us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2046us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2047us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2049us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2051us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2053us : steal_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2055us!: move_freepages_block <-steal_suitable_fallback

      dd-808       0d.... 2175us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2176us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2178us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2180us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2182us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2183us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2185us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2187us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2189us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2191us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2192us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2194us : steal_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2196us!: move_freepages_block <-steal_suitable_fallback

      dd-808       0d.... 2323us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2325us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2327us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2328us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2330us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2332us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2334us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2335us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2337us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2339us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2341us : find_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2343us : steal_suitable_fallback <-__rmqueue_pcplist

      dd-808       0d.... 2345us!: move_freepages_block <-steal_suitable_fallback

      dd-808       0d.... 2470us : __mod_zone_page_state <-__rmqueue_pcplist

      dd-808       0d.... 2473us : _raw_spin_unlock_irqrestore <-__rmqueue_pcplist

      dd-808       0d.... 2476us : _raw_spin_unlock_irqrestore

      dd-808       0d.... 2479us+: tracer_hardirqs_on <-_raw_spin_unlock_irqrestore

      dd-808       0d.... 2520us : <stack trace>

=> get_page_from_freelist

=> __alloc_pages

=> __folio_alloc

=> __filemap_get_folio

=> iomap_write_begin

=> iomap_file_buffered_write

=> blkdev_write_iter

=> vfs_write

=> ksys_write

=> ret_fast_syscall

 

Best Regard

Carlos Song

 


[Index of Archives]     [Linux ARM Kernel]     [Linux ARM]     [Linux Omap]     [Fedora ARM]     [IETF Annouce]     [Bugtraq]     [Linux OMAP]     [Linux MIPS]     [eCos]     [Asterisk Internet PBX]     [Linux API]

  Powered by Linux