Re: [PATCH] ext4: Do not zeroout uninitialized extents beyond i_size

Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]
From: Dmitry Monakhov
Date: Thursday, April 15, 2010 - 2:20 am

tytso@mit.edu writes:

Yepp. Indeed function are almost identical, and sadly i've missed 
may_zeroout condition update in second one :(. see later in the text
Other useful test is to run fsstress with boosted write/fallocate
probability on small(1Gb)  filesystem
xfstests/ltp/fsstress -p10 -z -f mkdir=1 -f creat=1 -f write=10 \
                      -f resrvsp=10 -n999999999 &
sleep 100
killall -9 fsstress
#umount/e2fsck
We have to update may_zeroout since orig extent was changed.
+               may_zeroout = ee_block + ee_len <= eof_block;  
I've attached incremental patch for you.
In fact, this not result in any serious bugs but increase our changes to
use zeroout optimization and helps us to overcome ENOSPC if possible.
Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]

Messages in current thread:
Re: [PATCH] ext4: Do not zeroout uninitialized extents bey ..., Dmitry Monakhov, (Thu Apr 15, 2:20 am)
Re: [PATCH] ext4: Do not zeroout uninitialized extents bey ..., Aneesh Kumar K. V, (Tue Apr 27, 9:40 pm)
Re: [PATCH] ext4: Do not zeroout uninitialized extents bey ..., Aneesh Kumar K. V, (Thu May 27, 10:19 am)