Distribution: Fedora8 Hardware Environment: attached lspci, dmesg Software Environment: Problem Description: [root@mailhub mail]# sync; time { dd if=/dev/zero of=F1 bs=1M count=102; sync; } 102+0 records in 102+0 records out 106954752 bytes (107 MB) copied, 34.7124 s, 3.1 MB/s real 2m23.034s user 0m0.000s sys 0m0.275s [root@mailhub mail]# hdparm -t /dev/mapper/mail /dev/mapper/mail: Timing buffered disk reads: 344 MB in 3.01 seconds = 114.28 MB/sec [root@mailhub mail]# uname -a Linux mailhub 2.6.24.4-64.fc8 #1 SMP Sat Mar 29 09:15:49 EDT 2008 x86_64 x86_64 x86_64 GNU/Linux very same happens on 2.6.25-gentoo-r1 It's probably related to http://bugzilla.kernel.org/show_bug.cgi?id=8020 but I was advised to create new bug entry
Created attachment 15833 [details] dmesg
Created attachment 15834 [details] lspci
Are you still seeing this? If so, try to gather a considerably more-complete set of data, varying one thing at a time and eliminating as much as possible from the problem. So do the tests in single user mode. Test varying quantities of data. Boot with a minimum amount of system memory and no swap. Test reads and writes. Flush the caches independently throughout the device stack between tests - don't just rely on 'sync'. Vary the configuration - compare different encryption methods and without any encryption, readhead etc. etc.
Closing old bug report.