linux-kernel - Re: [PATCHv4 11/12] dmapool: link blocks across pages

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:   Sun, 26 Feb 2023 16:54:45 -0800
From:   Guenter Roeck <linux@...ck-us.net>
To:     Keith Busch <kbusch@...a.com>
Cc:     linux-mm@...ck.org, linux-kernel@...r.kernel.org,
        willy@...radead.org, hch@....de, tonyb@...ernetics.com,
        akpm@...ux-foundation.org, kernel-team@...a.com,
        Keith Busch <kbusch@...nel.org>
Subject: Re: [PATCHv4 11/12] dmapool: link blocks across pages

Hi,

On Thu, Jan 26, 2023 at 01:51:24PM -0800, Keith Busch wrote:
> From: Keith Busch <kbusch@...nel.org>
> 
> The allocated dmapool pages are never freed for the lifetime of the
> pool. There is no need for the two level list+stack lookup for finding a
> free block since nothing is ever removed from the list. Just use a
> simple stack, reducing time complexity to constant.
> 
> The implementation inserts the stack linking elements and the dma handle
> of the block within itself when freed. This means the smallest possible
> dmapool block is increased to at most 16 bytes to accomodate these
> fields, but there are no exisiting users requesting a dma pool smaller
> than that anyway.
> 
> Removing the list has a significant change in performance. Using the
> kernel's micro-benchmarking self test:
> 
> Before:
> 
>   # modprobe dmapool_test
>   dmapool test: size:16   blocks:8192   time:57282
>   dmapool test: size:64   blocks:8192   time:172562
>   dmapool test: size:256  blocks:8192   time:789247
>   dmapool test: size:1024 blocks:2048   time:371823
>   dmapool test: size:4096 blocks:1024   time:362237
> 
> After:
> 
>   # modprobe dmapool_test
>   dmapool test: size:16   blocks:8192   time:24997
>   dmapool test: size:64   blocks:8192   time:26584
>   dmapool test: size:256  blocks:8192   time:33542
>   dmapool test: size:1024 blocks:2048   time:9022
>   dmapool test: size:4096 blocks:1024   time:6045
> 
> The module test allocates quite a few blocks that may not accurately
> represent how these pools are used in real life. For a more marco level
> benchmark, running fio high-depth + high-batched on nvme, this patch
> shows submission and completion latency reduced by ~100usec each, 1%
> IOPs improvement, and perf record's time spent in dma_pool_alloc/free
> were reduced by half.
> 

With this patch in linux-next, I see a boot failure when trying to boot
a powernv qemu emulation from the SCSI MEGASAS controller.

Qemu command line is

qemu-system-ppc64 -M powernv -cpu POWER9 -m 2G \
	-kernel arch/powerpc/boot/zImage.epapr \
	-snapshot \
	-device megasas,id=scsi,bus=pcie.0 -device scsi-hd,bus=scsi.0,drive=d0 \
	-drive file=rootfs-el.ext2,format=raw,if=none,id=d0 \
	-device i82557a,netdev=net0,bus=pcie.1 -netdev user,id=net0 \
	-nographic -vga none -monitor null -no-reboot \
	--append "root=/dev/sda console=tty console=hvc0"

Reverting this patch together with "dmapool: create/destroy cleanup"
fixes the problem.

Bisect log is attached for reference.

Guenter

---
# bad: [8232539f864ca60474e38eb42d451f5c26415856] Add linux-next specific files for 20230225
# good: [c9c3395d5e3dcc6daee66c6908354d47bf98cb0c] Linux 6.2
git bisect start 'HEAD' 'v6.2'
# good: [fe3130bc4df0b1303de4321af2bc4dcee5d7db2f] cifs: reuse cifs_match_ipaddr for comparison of dstaddr too
git bisect good fe3130bc4df0b1303de4321af2bc4dcee5d7db2f
# good: [8138ddac3c324feb92cc30f6d0d3a1bba51345a9] Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input.git
git bisect good 8138ddac3c324feb92cc30f6d0d3a1bba51345a9
# bad: [2a15ddbcd09ca3a7843a48832884e37e703eaf83] Merge branch 'master' of git://linuxtv.org/media_tree.git
git bisect bad 2a15ddbcd09ca3a7843a48832884e37e703eaf83
# bad: [a7d241d71cf464413307df69177ae2dec8481d37] Merge branch 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux.git
git bisect bad a7d241d71cf464413307df69177ae2dec8481d37
# bad: [446eb7f1f4aec9232d4b10222123a4566a8b1a95] Merge branch 'for-next' of git://git.kernel.org/pub/scm/linux/kernel/git/tmlind/linux-omap.git
git bisect bad 446eb7f1f4aec9232d4b10222123a4566a8b1a95
# good: [14c61d2100377dde2f6338395325b4090279d6a7] soc: document merges
git bisect good 14c61d2100377dde2f6338395325b4090279d6a7
# bad: [cb26c07e8a8acaecb43228181e1eae68ece8db0e] Merge branch 'for-next' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild.git
git bisect bad cb26c07e8a8acaecb43228181e1eae68ece8db0e
# bad: [d37d53a39d853fcc2121770fd3b61f274985d594] Merge branch 'mm-everything' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
git bisect bad d37d53a39d853fcc2121770fd3b61f274985d594
# bad: [708a06c601945c3415240ed0950e37fe27dd8e60] mm/userfaultfd: support WP on multiple VMAs
git bisect bad 708a06c601945c3415240ed0950e37fe27dd8e60
# good: [beb78ba6c0dbed73b38d5ed74bf47aa2c65fafa7] dmapool: move debug code to own functions
git bisect good beb78ba6c0dbed73b38d5ed74bf47aa2c65fafa7
# bad: [a2cb3f101b06f78258cf0c6813b3a17bd1ec846a] zsmalloc: remove insert_zspage() ->inuse optimization
git bisect bad a2cb3f101b06f78258cf0c6813b3a17bd1ec846a
# good: [e637ac603aec2b0a73e50fd8031481c6e55bf139] dmapool: don't memset on free twice
git bisect good e637ac603aec2b0a73e50fd8031481c6e55bf139
# bad: [8f5073712e32685dfeb4925f13a95c6eb9f10cd8] dmapool: create/destroy cleanup
git bisect bad 8f5073712e32685dfeb4925f13a95c6eb9f10cd8
# bad: [28b0a0c64bc658e176368f9270dc8085aa469c63] dmapool: link blocks across pages
git bisect bad 28b0a0c64bc658e176368f9270dc8085aa469c63
# first bad commit: [28b0a0c64bc658e176368f9270dc8085aa469c63] dmapool: link blocks across pages