Re: blk-13B
> You could try instrumenting the paths in bdflush and kswapd, to see > what really happens when they run crazy. This would help a lot. :-( My knowledge of the kernel is too poor and I have too few time to study it. I tried the suggestion of A.Arcangeli about size interval. The test is always the same large files cp. At the beginning it swapped out ~3MB and during that time (~5 sec.) the system was slow and sluggish. Then it worked fine for about 10 minutes. At a certain point it started swapping out without any apparent reason (I was just looking at the screen) and the system was hardly usable, then the swapping stopped and everything but disk writes remained frozen for about 20 sec. Now it's all ok and I have about 20MB of the swap. I don't think the change in ll_rw_blk caused the swapping, it simply didn't see it before (I use 2.2.18 most of the time). procsmemoryswap io system cpu r b w swpd free buff cache si sobibo incs us sy id 3 0 0 3512 1048 1776 126288 0 0 6682 768 162 981 12 25 63 3 0 0 3512 1048 1800 126264 0 0 6526 3200 161 1281 23 28 49 3 0 0 3512 1048 1816 126248 0 0 7373 3776 198 1149 23 21 55 2 0 0 3512 1048 1816 126248 0 0 7172 3584 206 799 8 26 66 1 0 0 3512 1048 1820 126256 0 0 7052 3520 153 917 11 29 60 1 0 1 3512 1048 1820 126256 0 0 7296 1218 177 3136 3 24 73 2 1 1 3512 1048 1840 126252 0 68 5654 145 101 3316 1 44 55 1 1 2 3504 1048 1848 126336 0 0 462134 69 2505 1 63 36 4 0 2 3600 1428 1856 128476 0 120 283254 191 1747 13 75 12 2 0 2 3920 1048 1740 129288 200 288 2876 202 135 3262 8 77 15 1 0 2 3980 1048 1548 129564 0 0 161510 143 4388 4 92 4 3 1 2 4832 1048 1264 130652 64 776 2013 276 268 3048 2 91 8 3 1 2 4832 1048 1264 130652 64 776 2013 276 268 3048 2 91 8 0 2 2 6560 1048 956 132696 0 1340 1783 567 150 1973 2 86 12 5 0 2 8288 1052 968 134400 632 2288 1426 898 242 1452 2 38 60 0 1 1 10912 1048 984 137004 388 1504 2997 546 127 2510 4 48 48 6 1 1 10924 1048 1008 136924 192 1088 2368 656 57 1737 4 59 37 1 0 2 11424 1096 968 137504 104 504 2066 302 42 1815 3 81 16 0 1 1 14160 1048 888 140372 24 2488 2649 1146 38 1770 2 57 41 1 2 1 16556 1048 908 142656 700 1512 2512 750 52 1787 0 53 47 1 0 1 18228 1048 852 144260 348 2260 2657 905 59 2465 10 49 42 0 1 2 19704 1048 704 145808 0 1772 2827 953 39 1977 2 55 43 0 2 2 20592 1048 716 146612 304 1076 3087 594 51 2177 1 58 41 6 1 2 21532 1048 712 147512 0 940 1615 366 31 1961 3 83 14 2 0 2 23056 1048 716 148972 264 688 3334 316 46 2406 3 66 31 4 0 2 24676 1048 732 150488 0 2096 2178 837 43 1357 1 52 47 0 1 1 27116 1048 736 152832 128 1972 2544 869 44 1812 2 43 55 5 0 2 27768 1048 744 153372 52 1308 2668 777 43 1772 2 61 37 0 2 1 28360 1048 752 153900 332 564 2311 955 49 2081 1 68 31 1 7 2 28356 1048 752 153708 3936 0 2175 29091 494 27348 0 1 99 1 0 2 28356 1048 792 153656 172 0 7166 0 144 838 4 17 80 1 0 2 28356 1048 836 153592 0 0 7453 0 77 668 0 18 82 0 1 2 28356 1048 800 153036 0 0 7196 0 63 3981 0 24 76 I wanted to fiddle with /proc/sys/vm/freepages but it's not writable in 2.4 Bye. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Re: blk-13B
You could try instrumenting the paths in bdflush and kswapd, to see what really happens when they run crazy. This would help a lot. :-( My knowledge of the kernel is too poor and I have too few time to study it. I tried the suggestion of A.Arcangeli about size interval. The test is always the same large files cp. At the beginning it swapped out ~3MB and during that time (~5 sec.) the system was slow and sluggish. Then it worked fine for about 10 minutes. At a certain point it started swapping out without any apparent reason (I was just looking at the screen) and the system was hardly usable, then the swapping stopped and everything but disk writes remained frozen for about 20 sec. Now it's all ok and I have about 20MB of the swap. I don't think the change in ll_rw_blk caused the swapping, it simply didn't see it before (I use 2.2.18 most of the time). procsmemoryswap io system cpu r b w swpd free buff cache si sobibo incs us sy id 3 0 0 3512 1048 1776 126288 0 0 6682 768 162 981 12 25 63 3 0 0 3512 1048 1800 126264 0 0 6526 3200 161 1281 23 28 49 3 0 0 3512 1048 1816 126248 0 0 7373 3776 198 1149 23 21 55 2 0 0 3512 1048 1816 126248 0 0 7172 3584 206 799 8 26 66 1 0 0 3512 1048 1820 126256 0 0 7052 3520 153 917 11 29 60 1 0 1 3512 1048 1820 126256 0 0 7296 1218 177 3136 3 24 73 2 1 1 3512 1048 1840 126252 0 68 5654 145 101 3316 1 44 55 1 1 2 3504 1048 1848 126336 0 0 462134 69 2505 1 63 36 4 0 2 3600 1428 1856 128476 0 120 283254 191 1747 13 75 12 2 0 2 3920 1048 1740 129288 200 288 2876 202 135 3262 8 77 15 1 0 2 3980 1048 1548 129564 0 0 161510 143 4388 4 92 4 3 1 2 4832 1048 1264 130652 64 776 2013 276 268 3048 2 91 8 3 1 2 4832 1048 1264 130652 64 776 2013 276 268 3048 2 91 8 0 2 2 6560 1048 956 132696 0 1340 1783 567 150 1973 2 86 12 5 0 2 8288 1052 968 134400 632 2288 1426 898 242 1452 2 38 60 0 1 1 10912 1048 984 137004 388 1504 2997 546 127 2510 4 48 48 6 1 1 10924 1048 1008 136924 192 1088 2368 656 57 1737 4 59 37 1 0 2 11424 1096 968 137504 104 504 2066 302 42 1815 3 81 16 0 1 1 14160 1048 888 140372 24 2488 2649 1146 38 1770 2 57 41 1 2 1 16556 1048 908 142656 700 1512 2512 750 52 1787 0 53 47 1 0 1 18228 1048 852 144260 348 2260 2657 905 59 2465 10 49 42 0 1 2 19704 1048 704 145808 0 1772 2827 953 39 1977 2 55 43 0 2 2 20592 1048 716 146612 304 1076 3087 594 51 2177 1 58 41 6 1 2 21532 1048 712 147512 0 940 1615 366 31 1961 3 83 14 2 0 2 23056 1048 716 148972 264 688 3334 316 46 2406 3 66 31 4 0 2 24676 1048 732 150488 0 2096 2178 837 43 1357 1 52 47 0 1 1 27116 1048 736 152832 128 1972 2544 869 44 1812 2 43 55 5 0 2 27768 1048 744 153372 52 1308 2668 777 43 1772 2 61 37 0 2 1 28360 1048 752 153900 332 564 2311 955 49 2081 1 68 31 frozen 1 7 2 28356 1048 752 153708 3936 0 2175 29091 494 27348 0 1 99 1 0 2 28356 1048 792 153656 172 0 7166 0 144 838 4 17 80 1 0 2 28356 1048 836 153592 0 0 7453 0 77 668 0 18 82 0 1 2 28356 1048 800 153036 0 0 7196 0 63 3981 0 24 76 I wanted to fiddle with /proc/sys/vm/freepages but it's not writable in 2.4 Bye. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
blk-13B
Hi! I found the time to make some tests with 2.4 and the blk-13B patch. It performs very well, no process starvation, no missed merges, etc., but sometimes it happens dbflush and kswapd start eating 100% CPU for about 20-30 secs.: 11:02am up 18 min, 5 users, load average: 2.74, 1.97, 1.15 66 processes: 60 sleeping, 6 running, 0 zombie, 0 stopped CPU states: 2.7% user, 97.2% system, 0.0% nice, 0.0% idle Mem: 189568K av, 188520K used, 1048K free, 209144K shrd, 1168K buff Swap: 262136K av, 0K used, 262136K free122764K cached PID USER PRI NI SIZE RSS SHARE STAT LIB %CPU %MEM TIME COMMAND 3 root20 0 00 0 SW 0 45.1 0.0 0:21 kswapd 5 root14 0 00 0 RW 0 44.5 0.0 0:29 bdflush 606 Giu 9 0 464 464 352 R 0 3.6 0.2 0:02 cp 543 Giu 9 0 604 604 472 S 0 1.9 0.3 0:14 vmstat vmstat doesn't show anything strange at the same time: procsmemoryswap io system cpu r b w swpd free buff cache si sobibo incs us sy id 2 0 00 1048 1304 122564 0 0 8960 4486 156 771 11 25 64 3 0 00 1048 1304 122564 0 0 9856 4996 123 622 3 30 67 1 0 10 1048 1304 122564 0 0 4096 1867 62 431 1 76 23 2 0 10 1048 1304 122560 0 0 1920 963 26 130 1 99 0 4 1 20 1048 1304 122564 0 0 1536 647 164 271 3 97 0 2 0 20 1048 1304 122564 0 0 760 516 139 482 6 94 0 1 1 20 1048 1304 122560 0 0 896 461 275 932 10 90 0 1 0 20 1048 1304 122552 0 0 672 452 130 988 7 93 0 1 0 20 1048 1304 122552 0 0 896 520 180 1905 11 89 0 2 0 20 1052 1304 122552 0 0 512 515 26 462 4 96 0 1 1 20 1048 1304 122556 0 0 740 519 109 428 1 99 0 3 0 10 1048 1228 122624 0 0 1052 516 23 175 4 96 0 The scsi driver reported it was writing large contiguous blocks and, 30secs later, a few 10s of smaller non mergeable block. It's ok IMO. ps doesn't show anything useful: [Giu@Jay Giu]$ ps -xao cmd,wchan CMD WCHAN init do_select [keventd]context_thread [kswapd] - [kreclaimd] kreclaimd [bdflush]- [kupdate]kupdate [khubd] usb_hub_thread The test is a large file copy from the HD to my MO. During the cp and now there's nothing on the swap. My config: PowerPC 750, 192MB ram, 256MB swap, kernel 2.4.0-bitkeeper (the official 2.4 doesn't compile) + blk-13B, Adaptec 2930U, gcc 2.95.3. The rest is a 2.2.x distribution. Bye. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
blk-13B
Hi! I found the time to make some tests with 2.4 and the blk-13B patch. It performs very well, no process starvation, no missed merges, etc., but sometimes it happens dbflush and kswapd start eating 100% CPU for about 20-30 secs.: 11:02am up 18 min, 5 users, load average: 2.74, 1.97, 1.15 66 processes: 60 sleeping, 6 running, 0 zombie, 0 stopped CPU states: 2.7% user, 97.2% system, 0.0% nice, 0.0% idle Mem: 189568K av, 188520K used, 1048K free, 209144K shrd, 1168K buff Swap: 262136K av, 0K used, 262136K free122764K cached PID USER PRI NI SIZE RSS SHARE STAT LIB %CPU %MEM TIME COMMAND 3 root20 0 00 0 SW 0 45.1 0.0 0:21 kswapd 5 root14 0 00 0 RW 0 44.5 0.0 0:29 bdflush 606 Giu 9 0 464 464 352 R 0 3.6 0.2 0:02 cp 543 Giu 9 0 604 604 472 S 0 1.9 0.3 0:14 vmstat vmstat doesn't show anything strange at the same time: procsmemoryswap io system cpu r b w swpd free buff cache si sobibo incs us sy id 2 0 00 1048 1304 122564 0 0 8960 4486 156 771 11 25 64 3 0 00 1048 1304 122564 0 0 9856 4996 123 622 3 30 67 1 0 10 1048 1304 122564 0 0 4096 1867 62 431 1 76 23 2 0 10 1048 1304 122560 0 0 1920 963 26 130 1 99 0 4 1 20 1048 1304 122564 0 0 1536 647 164 271 3 97 0 2 0 20 1048 1304 122564 0 0 760 516 139 482 6 94 0 1 1 20 1048 1304 122560 0 0 896 461 275 932 10 90 0 1 0 20 1048 1304 122552 0 0 672 452 130 988 7 93 0 1 0 20 1048 1304 122552 0 0 896 520 180 1905 11 89 0 2 0 20 1052 1304 122552 0 0 512 515 26 462 4 96 0 1 1 20 1048 1304 122556 0 0 740 519 109 428 1 99 0 3 0 10 1048 1228 122624 0 0 1052 516 23 175 4 96 0 The scsi driver reported it was writing large contiguous blocks and, 30secs later, a few 10s of smaller non mergeable block. It's ok IMO. ps doesn't show anything useful: [Giu@Jay Giu]$ ps -xao cmd,wchan CMD WCHAN init do_select [keventd]context_thread [kswapd] - [kreclaimd] kreclaimd [bdflush]- [kupdate]kupdate [khubd] usb_hub_thread The test is a large file copy from the HD to my MO. During the cp and now there's nothing on the swap. My config: PowerPC 750, 192MB ram, 256MB swap, kernel 2.4.0-bitkeeper (the official 2.4 doesn't compile) + blk-13B, Adaptec 2930U, gcc 2.95.3. The rest is a 2.2.x distribution. Bye. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
