Gunter Ohrner wrote: > I did something as simple as > > dd if=/dev/zero of=test.null bs=4M count=1024 ; rm test.null > This looked like a fun test, so I decided to do it myself. Especially since I've been having problems playing back videos from my hard drive lately. Linux roam 2.6.23.15-137.fc8 #1 SMP Sun Feb 10 17:03:13 EST 2008 x86_64 x86_64 x86_64 GNU/Linux It appears at first glance that the limiting factor here was disk I/O, since my hard drive light was continuously lit during the write operation. This was conducted on a laptop with a Samsung HS122JC 1.8" drive; the write of 4GB of data seemed to go about as fast as the drive can physically handle. error@roam /tmp $ time dd if=/dev/zero of=test.null bs=4M count=1024 ; rm test.null 1024+0 records in 1024+0 records out 4294967296 bytes (4.3 GB) copied, 559.611 s, 7.7 MB/s real 9m39.652s user 0m0.008s sys 0m20.271s While it was running, my music stopped playing for almost a minute at a time, several times. I could still use the box; only disk I/O was affected. My load average came up to 7.94. And the CPU wasn't at all stressed; this is what top looked like at a few points during the operation: (note that at several points only one of the kcryptd threads was running) PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 510 root 15 -5 0 0 0 D 9 0.0 1:01.09 kcryptd/0 511 root 15 -5 0 0 0 D 9 0.0 1:31.48 kcryptd/1 PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 511 root 15 -5 0 0 0 R 19 0.0 1:35.64 kcryptd/1 PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 15801 error 20 0 88308 4700 4608 D 46 0.2 0:12.09 dd 510 root 15 -5 0 0 0 D 7 0.0 1:17.67 kcryptd/0 511 root 15 -5 0 0 0 D 7 0.0 1:58.84 kcryptd/1 PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 510 root 15 -5 0 0 0 S 16 0.0 1:35.29 kcryptd/0 My first suspicion would be that Linux can't get any other I/O off the disk when both kcryptd threads are tied up. It was only when one of the two threads appeared in top that my music started playing again. Clearly I have enough CPU to handle both reading and writing encrypted data at once (it's only a 1.2GHz Intel Core 2 Duo). So perhaps only one thread should be handling these sorts of sustained write operations, so that the other thread can fetch my music off the disk? -- Homeland Stupidity <http://www.homelandstupidity.us/> --------------------------------------------------------------------- dm-crypt mailing list - http://www.saout.de/misc/dm-crypt/ To unsubscribe, e-mail: dm-crypt-unsubscribe@xxxxxxxx For additional commands, e-mail: dm-crypt-help@xxxxxxxx
