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  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:   Thu, 7 Nov 2019 10:38:19 -0500
From:   "Theodore Y. Ts'o" <>
To:     Dmitry Monakhov <>
Subject: Re: [PATCH] ext4: fix extent_status fragmentation for plain files

On Wed, Nov 06, 2019 at 12:25:02PM +0000, Dmitry Monakhov wrote:
> It is appeared that extent are not cached for inodes with depth == 0
> which result in suboptimal extent status populating inside ext4_map_blocks()
> by map's result where size requested is usually smaller than extent size so
> cache becomes fragmented
> # Example: I have plain file:
> File size of /mnt/test is 33554432 (8192 blocks of 4096 bytes)
>  ext:     logical_offset:        physical_offset: length:   expected: flags:
>    0:        0..    8191:      40960..     49151:   8192:             last,eof
> $ perf record -e 'ext4:ext4_es_*' /root/bin/fio --name=t --direct=0 --rw=randread --bs=4k --filesize=32M --size=32M --filename=/mnt/test
> $ perf script | grep ext4_es_insert_extent | head -n 10
>              fio   131 [000]    13.975421:           ext4:ext4_es_insert_extent: dev 253,0 ino 12 es [494/1) mapped 41454 status W
>              fio   131 [000]    13.976467:           ext4:ext4_es_insert_extent: dev 253,0 ino 12 es [6907/1) mapped 47867 status W

So this is certainly bad behavior, but the original intent was to not
cached extents that were in the inode's i_blocks[] array because the
information was already in the inode cache, and so we could save
memory but just pulling the information out of the i_blocks away and
there was no need to cache the extent in the es cache.

There are cases where we do need to track the extent in the es cache
--- for example, if we are writing the file and we need to track its
delayed allocation status.

So I wonder if we might be better off defining a new flag
EXT4_MAP_INROOT, which gets set by ext4_ext_map_blocks() and
ext4_ind_map_blocks() if the mapping is exclusively found in the
i_blocks array, and if EXT4_MAP_INROOT is set, and we don't need to
set EXTENT_STATUS_DELAYED, we skip the call to

What do you think?  This should significantly reduce the memory
utilization of the es_cache, which would be good for low-memory
workloads, and those where there are a large number of inodes that fit
in the es_cache, which is probably true for most desktops, especially
those belonging kernel developers.  :-)

						- Ted

Powered by blists - more mailing lists