Re: [git patches] IDE update
Bartlomiej Zolnierkiewicz wrote: Bartlomiej Zolnierkiewicz (11): ide: fix UDMA/MWDMA/SWDMA masks (v3) ide: rework the code for selecting the best DMA transfer mode (v3) ide: add ide_tune_dma() helper ide: make /proc/ide/ optional ide: split off ioctl handling from IDE settings (v2) ide: move IDE settings handling to ide-proc.c ide: cable detection fixes (take 2) ide: add initializing argument to ide_register_hw() ide: add ide_proc_register_port() ide: legacy PCI bus order probing fixes ide: fix PIO setup on resume for ATAPI devices [ sorry but the patch is too big to be included ] The limit was raised to 400K IIRC. Jeff - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
Bartlomiej Zolnierkiewicz wrote: * the code for selecting and programming the best DMA transfer mode has been reworked to be cleaner, more generic and more libata-like, ( 500 LOCs gone and this change allows the change described below) Bartlomiej Zolnierkiewicz (11): ide: rework the code for selecting the best DMA transfer mode (v3) Has this seen testing/exposure in -mm tree? Jeff - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
Andrew Morton wrote: Jeff Garzik [EMAIL PROTECTED] wrote: Has this seen testing/exposure in -mm tree? argh. If this was in a file called ide-rework-the-code-for-selecting-the-best-DMA-transfer-mode.patch then it would be so easy. logs into hera greps ah, it's hidden in ide-max-dma-mode-v3.patch. akpm:/usr/src/25 grep ide-max-dma-mode-v3 pc/*series pc/2.6.21-mm1-series:ide-ide-max-dma-mode-v3.patch pc/2.6.21-mm2-series:ide-ide-max-dma-mode-v3.patch pc/2.6.21-rc2-mm2-series:ide-ide-max-dma-mode-v3.patch pc/2.6.21-rc3-mm1-series:ide-ide-max-dma-mode-v3.patch pc/2.6.21-rc3-mm2-series:ide-ide-max-dma-mode-v3.patch pc/2.6.21-rc4-mm1-series:ide-ide-max-dma-mode-v3.patch pc/2.6.21-rc5-mm3-series:ide-ide-max-dma-mode-v3.patch pc/2.6.21-rc5-mm4-series:ide-ide-max-dma-mode-v3.patch pc/2.6.21-rc6-mm1-series:ide-ide-max-dma-mode-v3.patch pc/2.6.21-rc7-mm1-series:ide-ide-max-dma-mode-v3.patch pc/2.6.21-rc7-mm2-series:ide-ide-max-dma-mode-v3.patch No probs. Cool beans. I'm happy then. Jeff - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
Bartlomiej Zolnierkiewicz wrote: On Thursday 10 May 2007, Jeff Garzik wrote: The limit was raised to 400K IIRC. That's (good) news to me, here goes the actual 150K patch: Thanks. I did in fact receive copies from vger, so it went through. Jeff - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
From: Jeff Garzik [EMAIL PROTECTED] Date: Wed, 09 May 2007 18:46:16 -0400 Bartlomiej Zolnierkiewicz wrote: Bartlomiej Zolnierkiewicz (11): ide: fix UDMA/MWDMA/SWDMA masks (v3) ide: rework the code for selecting the best DMA transfer mode (v3) ide: add ide_tune_dma() helper ide: make /proc/ide/ optional ide: split off ioctl handling from IDE settings (v2) ide: move IDE settings handling to ide-proc.c ide: cable detection fixes (take 2) ide: add initializing argument to ide_register_hw() ide: add ide_proc_register_port() ide: legacy PCI bus order probing fixes ide: fix PIO setup on resume for ATAPI devices [ sorry but the patch is too big to be included ] The limit was raised to 400K IIRC. On lkml but not linux-ide, I'll go and bump it up there too. There, done... - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] ide update
On 8/19/05, Alan Cox [EMAIL PROTECTED] wrote: On Gwe, 2005-08-19 at 11:02 +0200, Bartlomiej Zolnierkiewicz wrote: lkml.org/lkml/2005/1/27/20 AFAIK CS5535 driver was never ported to 2.6.x. Somebody needs to port it to 2.6.x kernel, cleanup to match kernel coding standards and test. That was done some time ago and posted to various people. This is a good news that cs5530 driver was ported. BTW posting to various people is really not the best method of submitting kernel patches (in this case linux-ide@vger.kernel.org + cc: me is). Bartlomiej - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] ide update
On 8/18/05, Linus Torvalds [EMAIL PROTECTED] wrote: On Thu, 18 Aug 2005, Bartlomiej Zolnierkiewicz wrote: 3 obvious fixes + support for 2 new controllers (just new PCI IDs). Btw, things like this: +#define IDEFLOPPY_TICKS_DELAY HZ/20 /* default delay for ZIP 100 (50ms) */ are just bugs waiting to happen. snip Shouldn't this be msecs_to_jiffies(50) to avoid the rounding issues I mentioned on LKML a bit ago? Thanks, Nish - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] ide update
Linus Torvalds wrote: Btw, things like this: +#define IDEFLOPPY_TICKS_DELAY HZ/20 /* default delay for ZIP 100 (50ms) */ are just bugs waiting to happen. Needs parenthesis: ((HZ)/20) Or one could just use the msecs_to_jiffies() macro. Cheers - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
On Maw, 2005-07-05 at 20:14, Jens Axboe wrote: IDE still has much lower overhead per command than your average SCSI hardware. SATA with FIS even improves on this, definitely a good thing! But SCSI overlaps them while in PATA they are dead time. Thats why PATA is so demanding of large I/O block sizes - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
* Jens Axboe [EMAIL PROTECTED] wrote: But! I used hdparm -t solely, 2.6 was always ~5% faster than 2.4. But using -Tt slowed down the hd speed by about 30%. So it looks like some scheduler interaction, perhaps the memory timing loops gets it marked as batch or something? to check whether that could be the case, could you try: nice -n -20 hdparm -t /dev/hdc does that produce different results? Ingo - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
On Fri, 2005-07-08 at 10:06 +1000, Grant Coady wrote: I've not been able to get dual channel I/O speed faster than single interface speed, either as 'md' RAID0 or simultaneous reading or writing done the other day: Time to write or read 500MB file: summary 2.4.31-hf1 2.6.12.2 boxen \ time - w r w r --- ... peetoo 33 20 26.522 (simultaneuous 57 37.552 38.5) MB/s 2.4.31-hf1 2.6.12.2 w r w r single15 25 19 23 dual 17.527 19 26 These timings show very little happening in parallel, is that normal? me too ;-) signature.asc Description: This is a digitally signed message part
Re: [git patches] IDE update
On Fri, Jul 08 2005, Ingo Molnar wrote: * Jens Axboe [EMAIL PROTECTED] wrote: But! I used hdparm -t solely, 2.6 was always ~5% faster than 2.4. But using -Tt slowed down the hd speed by about 30%. So it looks like some scheduler interaction, perhaps the memory timing loops gets it marked as batch or something? to check whether that could be the case, could you try: nice -n -20 hdparm -t /dev/hdc does that produce different results? Same result, see my next mail, it turned out to be a read-ahead bug. -- Jens Axboe - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
On 7/6/05, Bill Davidsen [EMAIL PROTECTED] wrote: Ondrej Zary wrote: Jens Axboe wrote: On Tue, Jul 05 2005, Ondrej Zary wrote: André Tomt wrote: Al Boldi wrote: Bartlomiej Zolnierkiewicz wrote: { On 7/4/05, Al Boldi [EMAIL PROTECTED] wrote: Hdparm -tT gives 38mb/s in 2.4.31 Cat /dev/hda /dev/null gives 2% user 33% sys 65% idle Hdparm -tT gives 28mb/s in 2.6.12 Cat /dev/hda /dev/null gives 2% user 25% sys 0% idle 73% IOWAIT The hdparm doesn't get as high scores as in 2.4 is a old discussed to death problem on LKML. So far nobody has been able to show it affects anything but that pretty useless quasi-benchmark. No, it's not a problem with hdparm. hdparm only shows that there is _really_ a problem: 2.6.12 [EMAIL PROTECTED]:/home/rainbow# time dd if=/dev/hda of=/dev/null bs=512 count=1048576 1048576+0 records in 1048576+0 records out real0m32.339s user0m1.500s sys 0m14.560s 2.4.26 [EMAIL PROTECTED]:/home/rainbow# time dd if=/dev/hda of=/dev/null bs=512 count=1048576 1048576+0 records in 1048576+0 records out real0m23.858s user0m1.750s sys 0m15.180s Perhaps some read-ahead bug. What happens if you use bs=128k for instance? Nothing - it's still the same. [EMAIL PROTECTED]:/home/rainbow# time dd if=/dev/hda of=/dev/null bs=128k count=4096 4096+0 records in 4096+0 records out real0m32.832s user0m0.040s sys 0m15.670s Why is the system time so high? I tried that test here, and got: oddball:root time dd if=/dev/hda of=/dev/null bs=128k count=4096 4096+0 records in 4096+0 records out real0m37.927s user0m0.025s sys 0m6.547s oddball:root uname -rn oddball.prodigy.com 2.6.11ac7 Now this is one of the slowest CPUs still in use (which I why I test responsiveness on it), and it uses far less CPU time. cat /proc/cpuinfo processor : 0 vendor_id : GenuineIntel cpu family : 6 model : 5 model name : Pentium II (Deschutes) stepping: 1 cpu MHz : 348.507 cache size : 512 KB fdiv_bug: no hlt_bug : no f00f_bug: no coma_bug: no fpu : yes fpu_exception : yes cpuid level : 2 wp : yes flags : fpu vme de tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 mmx fxsr bogomips: 686.08 The first post said it felt like running PIO, it certainly is using CPU like it as well. Now here's some dmesg from this system... PIIX4: IDE controller at PCI slot :00:07.1 PIIX4: chipset revision 1 PIIX4: not 100% native mode: will probe irqs later ide0: BM-DMA at 0x1080-0x1087, BIOS settings: hda:DMA, hdb:pio ide1: BM-DMA at 0x1088-0x108f, BIOS settings: hdc:DMA, hdd:pio Probing IDE interface ide0... hda: Maxtor 90845D4, ATA DISK drive hdb: WDC AC31600H, ATA DISK drive hdb: Disabling (U)DMA for WDC AC31600H (blacklisted) ide0 at 0x1f0-0x1f7,0x3f6 on irq 14 Probing IDE interface ide1... hdc: NEC CD-ROM DRIVE:28C, ATAPI CD/DVD-ROM drive ide1 at 0x170-0x177,0x376 on irq 15 hda: max request size: 128KiB hda: 16514064 sectors (8455 MB) w/512KiB Cache, CHS=16383/16/63, UDMA(33) hda: cache flushes not supported hda: hda1 hda2 hda3 hda4 hda5 hdb: max request size: 128KiB hdb: 3173184 sectors (1624 MB) w/128KiB Cache, CHS=3148/16/63 hdb: cache flushes not supported hdb: hdb1 hdb2 hdb3 hdc: ATAPI 32X CD-ROM drive, 128kB Cache, UDMA(33) Uniform CD-ROM driver Revision: 3.20 And indeed it does show hda as dma, and hdb as pio (older versions of the kernel let me set hdb to dma and it worked fine...). But in the posted demsg the BIOS settings show pio for hda. Is this in any way relevant, given that UDA(33) appears later? BIOS setting is irrelevant and ~14MB/s for UDMA33 is OK. CPU cycles are wasted somewhere else... - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
Bill Davidsen wrote: Ondrej Zary wrote: Jens Axboe wrote: On Tue, Jul 05 2005, Ondrej Zary wrote: André Tomt wrote: Al Boldi wrote: Bartlomiej Zolnierkiewicz wrote: { On 7/4/05, Al Boldi [EMAIL PROTECTED] wrote: Hdparm -tT gives 38mb/s in 2.4.31 Cat /dev/hda /dev/null gives 2% user 33% sys 65% idle Hdparm -tT gives 28mb/s in 2.6.12 Cat /dev/hda /dev/null gives 2% user 25% sys 0% idle 73% IOWAIT The hdparm doesn't get as high scores as in 2.4 is a old discussed to death problem on LKML. So far nobody has been able to show it affects anything but that pretty useless quasi-benchmark. No, it's not a problem with hdparm. hdparm only shows that there is _really_ a problem: 2.6.12 [EMAIL PROTECTED]:/home/rainbow# time dd if=/dev/hda of=/dev/null bs=512 count=1048576 1048576+0 records in 1048576+0 records out real0m32.339s user0m1.500s sys 0m14.560s 2.4.26 [EMAIL PROTECTED]:/home/rainbow# time dd if=/dev/hda of=/dev/null bs=512 count=1048576 1048576+0 records in 1048576+0 records out real0m23.858s user0m1.750s sys 0m15.180s Perhaps some read-ahead bug. What happens if you use bs=128k for instance? Nothing - it's still the same. [EMAIL PROTECTED]:/home/rainbow# time dd if=/dev/hda of=/dev/null bs=128k count=4096 4096+0 records in 4096+0 records out real0m32.832s user0m0.040s sys 0m15.670s Why is the system time so high? I tried that test here, and got: oddball:root time dd if=/dev/hda of=/dev/null bs=128k count=4096 4096+0 records in 4096+0 records out real0m37.927s user0m0.025s sys 0m6.547s oddball:root uname -rn oddball.prodigy.com 2.6.11ac7 Now this is one of the slowest CPUs still in use (which I why I test responsiveness on it), and it uses far less CPU time. cat /proc/cpuinfo processor : 0 vendor_id : GenuineIntel cpu family : 6 model : 5 model name : Pentium II (Deschutes) stepping: 1 cpu MHz : 348.507 cache size : 512 KB fdiv_bug: no hlt_bug : no f00f_bug: no coma_bug: no fpu : yes fpu_exception : yes cpuid level : 2 wp : yes flags : fpu vme de tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 mmx fxsr bogomips: 686.08 The first post said it felt like running PIO, it certainly is using CPU like it as well. Now here's some dmesg from this system... PIIX4: IDE controller at PCI slot :00:07.1 PIIX4: chipset revision 1 PIIX4: not 100% native mode: will probe irqs later ide0: BM-DMA at 0x1080-0x1087, BIOS settings: hda:DMA, hdb:pio ide1: BM-DMA at 0x1088-0x108f, BIOS settings: hdc:DMA, hdd:pio Probing IDE interface ide0... hda: Maxtor 90845D4, ATA DISK drive hdb: WDC AC31600H, ATA DISK drive hdb: Disabling (U)DMA for WDC AC31600H (blacklisted) ide0 at 0x1f0-0x1f7,0x3f6 on irq 14 Probing IDE interface ide1... hdc: NEC CD-ROM DRIVE:28C, ATAPI CD/DVD-ROM drive ide1 at 0x170-0x177,0x376 on irq 15 hda: max request size: 128KiB hda: 16514064 sectors (8455 MB) w/512KiB Cache, CHS=16383/16/63, UDMA(33) hda: cache flushes not supported hda: hda1 hda2 hda3 hda4 hda5 hdb: max request size: 128KiB hdb: 3173184 sectors (1624 MB) w/128KiB Cache, CHS=3148/16/63 hdb: cache flushes not supported hdb: hdb1 hdb2 hdb3 hdc: ATAPI 32X CD-ROM drive, 128kB Cache, UDMA(33) Uniform CD-ROM driver Revision: 3.20 And indeed it does show hda as dma, and hdb as pio (older versions of the kernel let me set hdb to dma and it worked fine...). But in the posted demsg the BIOS settings show pio for hda. Is this in any way relevant, given that UDA(33) appears later? I've had AC31600H drive too and it used to run fine in DMA mode for ages (at least in Windows). No idea why it is blacklisted. Ondrei: PIIX4: IDE controller at PCI slot 00:07.1 PIIX4: chipset revision 1 PIIX4: not 100% native mode: will probe irqs later ide0: BM-DMA at 0xf000-0xf007, BIOS settings: hda:pio, hdb:pio ide1: BM-DMA at 0xf008-0xf00f, BIOS settings: hdc:pio, hdd:pio hda: WDC WD300BB-00AUA1, ATA DISK drive blk: queue c03b3360, I/O limit 4095Mb (mask 0x) hdd: MSI CD-RW MS-8340S, ATAPI CD/DVD-ROM drive ide0 at 0x1f0-0x1f7,0x3f6 on irq 14 ide1 at 0x170-0x177,0x376 on irq 15 hda: attached ide-disk driver. hda: host protected area = 1 hda: 58633344 sectors (30020 MB) w/2048KiB Cache, CHS=3649/255/63, UDMA(33) Unless Ondrei's CPU is slower than a PI-350, maybe it is running pio for some reason, in spite of hdparm -i showing udma2 mode. I just mentioned it because the CPU usage doesn't seem compatible with DMA operation. Yes, the CPU is slower. It's Cyrix MII PR300 (225MHz): [EMAIL PROTECTED]:/home/rainbow# cat /proc/cpuinfo processor : 0 vendor_id : CyrixInstead cpu family : 6 model : 2 model name : M II 3x Core/Bus Clock stepping: 4 cpu MHz : 225.004 fdiv_bug: no hlt_bug : no f00f_bug: no coma_bug: no
Re: [git patches] IDE update
Ondrej Zary wrote: Jens Axboe wrote: On Tue, Jul 05 2005, Ondrej Zary wrote: André Tomt wrote: Al Boldi wrote: Bartlomiej Zolnierkiewicz wrote: { On 7/4/05, Al Boldi [EMAIL PROTECTED] wrote: Hdparm -tT gives 38mb/s in 2.4.31 Cat /dev/hda /dev/null gives 2% user 33% sys 65% idle Hdparm -tT gives 28mb/s in 2.6.12 Cat /dev/hda /dev/null gives 2% user 25% sys 0% idle 73% IOWAIT The hdparm doesn't get as high scores as in 2.4 is a old discussed to death problem on LKML. So far nobody has been able to show it affects anything but that pretty useless quasi-benchmark. No, it's not a problem with hdparm. hdparm only shows that there is _really_ a problem: 2.6.12 [EMAIL PROTECTED]:/home/rainbow# time dd if=/dev/hda of=/dev/null bs=512 count=1048576 1048576+0 records in 1048576+0 records out real0m32.339s user0m1.500s sys 0m14.560s 2.4.26 [EMAIL PROTECTED]:/home/rainbow# time dd if=/dev/hda of=/dev/null bs=512 count=1048576 1048576+0 records in 1048576+0 records out real0m23.858s user0m1.750s sys 0m15.180s Perhaps some read-ahead bug. What happens if you use bs=128k for instance? Nothing - it's still the same. [EMAIL PROTECTED]:/home/rainbow# time dd if=/dev/hda of=/dev/null bs=128k count=4096 4096+0 records in 4096+0 records out real0m32.832s user0m0.040s sys 0m15.670s Why is the system time so high? I tried that test here, and got: oddball:root time dd if=/dev/hda of=/dev/null bs=128k count=4096 4096+0 records in 4096+0 records out real0m37.927s user0m0.025s sys 0m6.547s oddball:root uname -rn oddball.prodigy.com 2.6.11ac7 Now this is one of the slowest CPUs still in use (which I why I test responsiveness on it), and it uses far less CPU time. cat /proc/cpuinfo processor : 0 vendor_id : GenuineIntel cpu family : 6 model : 5 model name : Pentium II (Deschutes) stepping: 1 cpu MHz : 348.507 cache size : 512 KB fdiv_bug: no hlt_bug : no f00f_bug: no coma_bug: no fpu : yes fpu_exception : yes cpuid level : 2 wp : yes flags : fpu vme de tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 mmx fxsr bogomips: 686.08 The first post said it felt like running PIO, it certainly is using CPU like it as well. Now here's some dmesg from this system... PIIX4: IDE controller at PCI slot :00:07.1 PIIX4: chipset revision 1 PIIX4: not 100% native mode: will probe irqs later ide0: BM-DMA at 0x1080-0x1087, BIOS settings: hda:DMA, hdb:pio ide1: BM-DMA at 0x1088-0x108f, BIOS settings: hdc:DMA, hdd:pio Probing IDE interface ide0... hda: Maxtor 90845D4, ATA DISK drive hdb: WDC AC31600H, ATA DISK drive hdb: Disabling (U)DMA for WDC AC31600H (blacklisted) ide0 at 0x1f0-0x1f7,0x3f6 on irq 14 Probing IDE interface ide1... hdc: NEC CD-ROM DRIVE:28C, ATAPI CD/DVD-ROM drive ide1 at 0x170-0x177,0x376 on irq 15 hda: max request size: 128KiB hda: 16514064 sectors (8455 MB) w/512KiB Cache, CHS=16383/16/63, UDMA(33) hda: cache flushes not supported hda: hda1 hda2 hda3 hda4 hda5 hdb: max request size: 128KiB hdb: 3173184 sectors (1624 MB) w/128KiB Cache, CHS=3148/16/63 hdb: cache flushes not supported hdb: hdb1 hdb2 hdb3 hdc: ATAPI 32X CD-ROM drive, 128kB Cache, UDMA(33) Uniform CD-ROM driver Revision: 3.20 And indeed it does show hda as dma, and hdb as pio (older versions of the kernel let me set hdb to dma and it worked fine...). But in the posted demsg the BIOS settings show pio for hda. Is this in any way relevant, given that UDA(33) appears later? Ondrei: PIIX4: IDE controller at PCI slot 00:07.1 PIIX4: chipset revision 1 PIIX4: not 100% native mode: will probe irqs later ide0: BM-DMA at 0xf000-0xf007, BIOS settings: hda:pio, hdb:pio ide1: BM-DMA at 0xf008-0xf00f, BIOS settings: hdc:pio, hdd:pio hda: WDC WD300BB-00AUA1, ATA DISK drive blk: queue c03b3360, I/O limit 4095Mb (mask 0x) hdd: MSI CD-RW MS-8340S, ATAPI CD/DVD-ROM drive ide0 at 0x1f0-0x1f7,0x3f6 on irq 14 ide1 at 0x170-0x177,0x376 on irq 15 hda: attached ide-disk driver. hda: host protected area = 1 hda: 58633344 sectors (30020 MB) w/2048KiB Cache, CHS=3649/255/63, UDMA(33) Unless Ondrei's CPU is slower than a PI-350, maybe it is running pio for some reason, in spite of hdparm -i showing udma2 mode. I just mentioned it because the CPU usage doesn't seem compatible with DMA operation. - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
Note: hdparm can also use O_DIRECT for the -t timing test. Eg. hdparm --direct -t /dev/hda - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
On Thu, 07 Jul 2005 18:32:52 -0400, Mark Lord [EMAIL PROTECTED] wrote: hdparm can also use O_DIRECT for the -t timing test. I've not been able to get dual channel I/O speed faster than single interface speed, either as 'md' RAID0 or simultaneous reading or writing done the other day: Time to write or read 500MB file: summary2.4.31-hf1 2.6.12.2 boxen \ time - w r w r --- ... peetoo 33 20 26.522 (simultaneuous 57 37.552 38.5) MB/s2.4.31-hf1 2.6.12.2 w r w r single 15 25 19 23 dual17.527 19 26 These timings show very little happening in parallel, is that normal? Thanks, --Grant. - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
André Tomt wrote: Al Boldi wrote: Bartlomiej Zolnierkiewicz wrote: { On 7/4/05, Al Boldi [EMAIL PROTECTED] wrote: Hdparm -tT gives 38mb/s in 2.4.31 Cat /dev/hda /dev/null gives 2% user 33% sys 65% idle Hdparm -tT gives 28mb/s in 2.6.12 Cat /dev/hda /dev/null gives 2% user 25% sys 0% idle 73% IOWAIT The hdparm doesn't get as high scores as in 2.4 is a old discussed to death problem on LKML. So far nobody has been able to show it affects anything but that pretty useless quasi-benchmark. No, it's not a problem with hdparm. hdparm only shows that there is _really_ a problem: 2.6.12 [EMAIL PROTECTED]:/home/rainbow# time dd if=/dev/hda of=/dev/null bs=512 count=1048576 1048576+0 records in 1048576+0 records out real0m32.339s user0m1.500s sys 0m14.560s 2.4.26 [EMAIL PROTECTED]:/home/rainbow# time dd if=/dev/hda of=/dev/null bs=512 count=1048576 1048576+0 records in 1048576+0 records out real0m23.858s user0m1.750s sys 0m15.180s -- Ondrej Zary - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
Jens Axboe wrote: On Tue, Jul 05 2005, Ondrej Zary wrote: André Tomt wrote: Al Boldi wrote: Bartlomiej Zolnierkiewicz wrote: { On 7/4/05, Al Boldi [EMAIL PROTECTED] wrote: Hdparm -tT gives 38mb/s in 2.4.31 Cat /dev/hda /dev/null gives 2% user 33% sys 65% idle Hdparm -tT gives 28mb/s in 2.6.12 Cat /dev/hda /dev/null gives 2% user 25% sys 0% idle 73% IOWAIT The hdparm doesn't get as high scores as in 2.4 is a old discussed to death problem on LKML. So far nobody has been able to show it affects anything but that pretty useless quasi-benchmark. No, it's not a problem with hdparm. hdparm only shows that there is _really_ a problem: 2.6.12 [EMAIL PROTECTED]:/home/rainbow# time dd if=/dev/hda of=/dev/null bs=512 count=1048576 1048576+0 records in 1048576+0 records out real0m32.339s user0m1.500s sys 0m14.560s 2.4.26 [EMAIL PROTECTED]:/home/rainbow# time dd if=/dev/hda of=/dev/null bs=512 count=1048576 1048576+0 records in 1048576+0 records out real0m23.858s user0m1.750s sys 0m15.180s Perhaps some read-ahead bug. What happens if you use bs=128k for instance? Nothing - it's still the same. [EMAIL PROTECTED]:/home/rainbow# time dd if=/dev/hda of=/dev/null bs=128k count=4096 4096+0 records in 4096+0 records out real0m32.832s user0m0.040s sys 0m15.670s -- Ondrej Zary - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
On Tue, Jul 05 2005, Ondrej Zary wrote: Jens Axboe wrote: On Tue, Jul 05 2005, Ondrej Zary wrote: André Tomt wrote: Al Boldi wrote: Bartlomiej Zolnierkiewicz wrote: { On 7/4/05, Al Boldi [EMAIL PROTECTED] wrote: Hdparm -tT gives 38mb/s in 2.4.31 Cat /dev/hda /dev/null gives 2% user 33% sys 65% idle Hdparm -tT gives 28mb/s in 2.6.12 Cat /dev/hda /dev/null gives 2% user 25% sys 0% idle 73% IOWAIT The hdparm doesn't get as high scores as in 2.4 is a old discussed to death problem on LKML. So far nobody has been able to show it affects anything but that pretty useless quasi-benchmark. No, it's not a problem with hdparm. hdparm only shows that there is _really_ a problem: 2.6.12 [EMAIL PROTECTED]:/home/rainbow# time dd if=/dev/hda of=/dev/null bs=512 count=1048576 1048576+0 records in 1048576+0 records out real0m32.339s user0m1.500s sys 0m14.560s 2.4.26 [EMAIL PROTECTED]:/home/rainbow# time dd if=/dev/hda of=/dev/null bs=512 count=1048576 1048576+0 records in 1048576+0 records out real0m23.858s user0m1.750s sys 0m15.180s Perhaps some read-ahead bug. What happens if you use bs=128k for instance? Nothing - it's still the same. [EMAIL PROTECTED]:/home/rainbow# time dd if=/dev/hda of=/dev/null bs=128k count=4096 4096+0 records in 4096+0 records out real0m32.832s user0m0.040s sys 0m15.670s Can you post full dmesg of 2.4 and 2.6 kernel boot? What does hdparm -I/-i say for both kernels? -- Jens Axboe - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
Jens Axboe wrote: On Tue, Jul 05 2005, Ondrej Zary wrote: Jens Axboe wrote: On Tue, Jul 05 2005, Ondrej Zary wrote: André Tomt wrote: Al Boldi wrote: Bartlomiej Zolnierkiewicz wrote: { On 7/4/05, Al Boldi [EMAIL PROTECTED] wrote: Hdparm -tT gives 38mb/s in 2.4.31 Cat /dev/hda /dev/null gives 2% user 33% sys 65% idle Hdparm -tT gives 28mb/s in 2.6.12 Cat /dev/hda /dev/null gives 2% user 25% sys 0% idle 73% IOWAIT The hdparm doesn't get as high scores as in 2.4 is a old discussed to death problem on LKML. So far nobody has been able to show it affects anything but that pretty useless quasi-benchmark. No, it's not a problem with hdparm. hdparm only shows that there is _really_ a problem: 2.6.12 [EMAIL PROTECTED]:/home/rainbow# time dd if=/dev/hda of=/dev/null bs=512 count=1048576 1048576+0 records in 1048576+0 records out real0m32.339s user0m1.500s sys 0m14.560s 2.4.26 [EMAIL PROTECTED]:/home/rainbow# time dd if=/dev/hda of=/dev/null bs=512 count=1048576 1048576+0 records in 1048576+0 records out real0m23.858s user0m1.750s sys 0m15.180s Perhaps some read-ahead bug. What happens if you use bs=128k for instance? Nothing - it's still the same. [EMAIL PROTECTED]:/home/rainbow# time dd if=/dev/hda of=/dev/null bs=128k count=4096 4096+0 records in 4096+0 records out real0m32.832s user0m0.040s sys 0m15.670s Can you post full dmesg of 2.4 and 2.6 kernel boot? What does hdparm -I/-i say for both kernels? The 2.4.26 kernel is the one from Slackware 10.0 bootable install CD. dmesg outputs attached, hdparm -i and hdparm -I shows the same in both kernels (compared using diff) - attached too. -- Ondrej Zary Linux version 2.4.26 ([EMAIL PROTECTED]) (gcc version 3.3.4) #6 Mon Jun 14 19:07:27 PDT 2004 BIOS-provided physical RAM map: BIOS-e820: - 0009fc00 (usable) BIOS-e820: 0009fc00 - 000a (reserved) BIOS-e820: 000f - 0010 (reserved) BIOS-e820: 0010 - 07ff (usable) BIOS-e820: 07ff - 07ff3000 (ACPI NVS) BIOS-e820: 07ff3000 - 0800 (ACPI data) BIOS-e820: - 0001 (reserved) 127MB LOWMEM available. On node 0 totalpages: 32752 zone(0): 4096 pages. zone(1): 28656 pages. zone(2): 0 pages. Kernel command line: BOOT_IMAGE=/kernels/bare.i/bzImage initrd=initrd.img load_ramdisk=1 prompt_ramdisk=0 ramdisk_size=6464 rw root=/dev/ram SLACK_KERNEL=bare.i Initializing CPU#0 Detected 225.001 MHz processor. Console: colour VGA+ 80x25 Calibrating delay loop... 448.92 BogoMIPS Memory: 123608k/131008k available (1844k kernel code, 7012k reserved, 618k data, 120k init, 0k highmem) Checking if this processor honours the WP bit even in supervisor mode... Ok. Dentry cache hash table entries: 16384 (order: 5, 131072 bytes) Inode cache hash table entries: 8192 (order: 4, 65536 bytes) Mount cache hash table entries: 512 (order: 0, 4096 bytes) Buffer cache hash table entries: 4096 (order: 2, 16384 bytes) Page-cache hash table entries: 32768 (order: 5, 131072 bytes) CPU: After generic, caps: 0080a135 0004 CPU: Common caps: 0080a135 0004 CPU: Cyrix M II 3x Core/Bus Clock stepping 04 Checking 'hlt' instruction... OK. POSIX conformance testing by UNIFIX mtrr: v1.40 (20010327) Richard Gooch ([EMAIL PROTECTED]) mtrr: detected mtrr type: Cyrix ARR PCI: PCI BIOS revision 2.10 entry at 0xfb020, last bus=0 PCI: Using configuration type 1 PCI: Probing PCI hardware PCI: Probing PCI hardware (bus 00) PCI: Using IRQ router PIIX/ICH [8086/7110] at 00:07.0 PCI: Device 00:3a not found by BIOS PCI: Device 00:3b not found by BIOS Limiting direct PCI/PCI transfers. Linux NET4.0 for Linux 2.4 Based upon Swansea University Computer Society NET3.039 Initializing RT netlink socket Starting kswapd VFS: Disk quotas vdquot_6.5.1 Journalled Block Device driver loaded pty: 512 Unix98 ptys configured Serial driver version 5.05c (2001-07-08) with HUB-6 MANY_PORTS MULTIPORT SHARE_IRQ SERIAL_PCI enabled ttyS00 at 0x03f8 (irq = 4) is a 16550A ttyS01 at 0x02f8 (irq = 3) is a 16550A Real Time Clock Driver v1.10f Floppy drive(s): fd0 is 1.44M FDC 0 is a post-1991 82077 RAMDISK driver initialized: 16 RAM disks of 6464K size 1024 blocksize loop: loaded (max 8 devices) Uniform Multi-Platform E-IDE driver Revision: 7.00beta4-2.4 ide: Assuming 33MHz system bus speed for PIO modes; override with idebus=xx PIIX4: IDE controller at PCI slot 00:07.1 PIIX4: chipset revision 1 PIIX4: not 100% native mode: will probe irqs later ide0: BM-DMA at 0xf000-0xf007, BIOS settings: hda:pio, hdb:pio ide1: BM-DMA at 0xf008-0xf00f, BIOS settings: hdc:pio, hdd:pio hda: WDC WD300BB-00AUA1, ATA DISK drive blk: queue c03b3360, I/O limit 4095Mb (mask 0x) hdd: MSI CD-RW MS-8340S, ATAPI CD/DVD-ROM drive ide0 at 0x1f0-0x1f7,0x3f6 on irq 14 ide1 at 0x170-0x177,0x376 on irq 15 hda:
Re: [git patches] IDE update
On Tue, 2005-07-05 at 14:35 +0200, Ondrej Zary wrote: 2.4.26 [EMAIL PROTECTED]:/home/rainbow# time dd if=/dev/hda of=/dev/null bs=512 count=1048576 1048576+0 records in 1048576+0 records out real0m23.858s user0m1.750s sys 0m15.180s Perhaps some read-ahead bug. What happens if you use bs=128k for instance? Nothing - it's still the same. [EMAIL PROTECTED]:/home/rainbow# time dd if=/dev/hda of=/dev/null bs=128k count=4096 4096+0 records in 4096+0 records out real0m32.832s user0m0.040s sys 0m15.670s Can you post full dmesg of 2.4 and 2.6 kernel boot? What does hdparm -I/-i say for both kernels? The 2.4.26 kernel is the one from Slackware 10.0 bootable install CD. dmesg outputs attached, hdparm -i and hdparm -I shows the same in both kernels (compared using diff) - attached too. Ok, looks alright for both. Your machine is quite slow, perhaps that is showing the slower performance. Can you try and make HZ 100 in 2.6 and test again? 2.6.13-recent has it as a config option, otherwise edit include/asm/param.h appropriately. -- Jens Axboe [EMAIL PROTECTED] - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
Jens Axboe wrote: On Tue, 2005-07-05 at 14:35 +0200, Ondrej Zary wrote: 2.4.26 [EMAIL PROTECTED]:/home/rainbow# time dd if=/dev/hda of=/dev/null bs=512 count=1048576 1048576+0 records in 1048576+0 records out real0m23.858s user0m1.750s sys 0m15.180s Perhaps some read-ahead bug. What happens if you use bs=128k for instance? Nothing - it's still the same. [EMAIL PROTECTED]:/home/rainbow# time dd if=/dev/hda of=/dev/null bs=128k count=4096 4096+0 records in 4096+0 records out real0m32.832s user0m0.040s sys 0m15.670s Can you post full dmesg of 2.4 and 2.6 kernel boot? What does hdparm -I/-i say for both kernels? The 2.4.26 kernel is the one from Slackware 10.0 bootable install CD. dmesg outputs attached, hdparm -i and hdparm -I shows the same in both kernels (compared using diff) - attached too. Ok, looks alright for both. Your machine is quite slow, perhaps that is showing the slower performance. Can you try and make HZ 100 in 2.6 and test again? 2.6.13-recent has it as a config option, otherwise edit include/asm/param.h appropriately. I forgot to write that my 2.6.12 kernel is already compiled with HZ 100 (it makes the system more responsive). I've just tried 2.6.8.1 with HZ 1000 and there is no difference in HDD performance comparing to 2.6.12. -- Ondrej Zary - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
Jens Axboe wrote: On Tue, 2005-07-05 at 15:02 +0200, Ondrej Zary wrote: Ok, looks alright for both. Your machine is quite slow, perhaps that is showing the slower performance. Can you try and make HZ 100 in 2.6 and test again? 2.6.13-recent has it as a config option, otherwise edit include/asm/param.h appropriately. I forgot to write that my 2.6.12 kernel is already compiled with HZ 100 (it makes the system more responsive). I've just tried 2.6.8.1 with HZ 1000 and there is no difference in HDD performance comparing to 2.6.12. OK, interesting. You could try and boot with profile=2 and do # readprofile -r # dd if=/dev/hda of=/dev/null bs=128k # readprofile prof_output for each kernel and post it here, so we can see if anything sticks out. Here are the profiles (used dd with count=4096) from 2.4.26 and 2.6.12 (nothing from 2.6.8.1 because I don't have the .map file anymore). -- Ondrej Zary 448 default_idle 5.6000 4 __switch_to0.0192 1 ret_from_sys_call 0.0588 3 handle_IRQ_event 0.0268 6 schedule 0.0072 8 __wake_up 0.0833 2 __constant_memcpy 0.0078 4 add_page_to_hash_queue 0.0625 2 __remove_inode_page0.0156 6 add_to_page_cache_unique 0.0375 7 page_cache_read0.0398 2 unlock_page0.0179 7 generic_file_readahead 0.0168 11 do_generic_file_read 0.0101 1366 file_read_actor8.5375 2 generic_file_read 0.0048 1 kmem_slab_destroy 0.0042 3 kmem_cache_grow0.0057 2 kmem_cache_alloc 0.0625 28 kmem_cache_free0.8750 2 kmem_cache_reap0.0042 12 __kmem_cache_alloc 0.0536 4 lru_cache_add 0.0357 1 __lru_cache_del0.0078 3 delta_nr_inactive_pages0.0312 5 delta_nr_cache_pages 0.0521 17 shrink_cache 0.0180 1 refill_inactive0.0031 16 __free_pages_ok0.0200 20 rmqueue0.0284 6 __alloc_pages 0.0091 1 __free_pages 0.0208 3 sys_read 0.0117 1 sys_write 0.0039 5 __remove_from_lru_list 0.0391 3 __remove_from_queues 0.0625 7 __put_unused_buffer_head 0.0729 3 get_unused_buffer_head 0.0234 24 set_bh_page0.2143 32 create_buffers 0.1429 4 try_to_release_page0.0500 4 create_empty_buffers 0.0417 24 block_read_full_page 0.0375 27 try_to_free_buffers0.1055 10 max_block 0.0694 9 blkdev_get_block 0.1406 2 blkdev_direct_IO 0.0417 1 blkdev_readpage0.0312 2 init_buffer_head 0.0312 25 __constant_c_and_count_memset 0.1736 2 write_profile 0.0417 114 __make_request 0.0642 27 generic_make_request 0.0844 25 submit_bh 0.0977 5 ide_inb0.3125 7 ide_outb 0.4375 2 ide_outl 0.1250 1 ide_execute_command0.0078 2 SELECT_DRIVE 0.0312 4 ide_start_request 0.0083 2 ide_do_request 0.0050 2 ide_get_queue 0.0312 4 ide_intr 0.0139 6 ide_dma_intr 0.0312 30 ide_build_sglist 0.0481 5 ide_build_dmatable 0.0120 1 __ide_dma_read 0.0042 1 __ide_dma_count0.0312 3 __constant_c_and_count_memset 0.0208 67 idedisk_end_request0.3490 2 __generic_copy_to_user 0.0312 0 *unknown* 2499 total
Re: [git patches] IDE update
On Tue, Jul 05 2005, Ondrej Zary wrote: Jens Axboe wrote: On Tue, 2005-07-05 at 15:02 +0200, Ondrej Zary wrote: Ok, looks alright for both. Your machine is quite slow, perhaps that is showing the slower performance. Can you try and make HZ 100 in 2.6 and test again? 2.6.13-recent has it as a config option, otherwise edit include/asm/param.h appropriately. I forgot to write that my 2.6.12 kernel is already compiled with HZ 100 (it makes the system more responsive). I've just tried 2.6.8.1 with HZ 1000 and there is no difference in HDD performance comparing to 2.6.12. OK, interesting. You could try and boot with profile=2 and do # readprofile -r # dd if=/dev/hda of=/dev/null bs=128k # readprofile prof_output for each kernel and post it here, so we can see if anything sticks out. Here are the profiles (used dd with count=4096) from 2.4.26 and 2.6.12 (nothing from 2.6.8.1 because I don't have the .map file anymore). Looks interesting, 2.6 spends oodles of times copying to user space. Lets check if raw reads perform ok, please try and time this app in 2.4 and 2.6 as well. # gcc -Wall -O2 -o oread oread.c # time ./oread /dev/hda -- Jens Axboe #include stdio.h #include unistd.h #define __USE_GNU #include fcntl.h #include stdlib.h #define BS (131072) #define BLOCKS (4096) #define ALIGN(buf) (char *) (((unsigned long) (buf) + 4095) ~(4095)) int main(int argc, char *argv[]) { char *buffer; int fd, i; if (argc 2) { printf(%s: device\n, argv[0]); return 1; } fd = open(argv[1], O_RDONLY | O_DIRECT); if (fd == -1) { perror(open); return 2; } buffer = ALIGN(malloc(BS + 4095)); for (i = 0; i BLOCKS; i++) { int ret = read(fd, buffer, BS); if (!ret) break; else if (ret 0) { perror(read infile); break; } } return 0; }
Re: [git patches] IDE update
Jens Axboe wrote: On Tue, Jul 05 2005, Ondrej Zary wrote: Jens Axboe wrote: On Tue, 2005-07-05 at 15:02 +0200, Ondrej Zary wrote: Ok, looks alright for both. Your machine is quite slow, perhaps that is showing the slower performance. Can you try and make HZ 100 in 2.6 and test again? 2.6.13-recent has it as a config option, otherwise edit include/asm/param.h appropriately. I forgot to write that my 2.6.12 kernel is already compiled with HZ 100 (it makes the system more responsive). I've just tried 2.6.8.1 with HZ 1000 and there is no difference in HDD performance comparing to 2.6.12. OK, interesting. You could try and boot with profile=2 and do # readprofile -r # dd if=/dev/hda of=/dev/null bs=128k # readprofile prof_output for each kernel and post it here, so we can see if anything sticks out. Here are the profiles (used dd with count=4096) from 2.4.26 and 2.6.12 (nothing from 2.6.8.1 because I don't have the .map file anymore). Looks interesting, 2.6 spends oodles of times copying to user space. Lets check if raw reads perform ok, please try and time this app in 2.4 and 2.6 as well. # gcc -Wall -O2 -o oread oread.c # time ./oread /dev/hda oread is faster than dd, but still not as fast as 2.4. In 2.6.12, HDD led is blinking, in 2.4 it's solid on during the read. 2.6.12: [EMAIL PROTECTED]:/home/rainbow# time ./oread /dev/hda real0m25.082s user0m0.000s sys 0m0.680s 2.4.26: [EMAIL PROTECTED]:/home/rainbow# time ./oread /dev/hda real0m23.513s user0m0.000s sys 0m2.360s -- Ondrej Zary 2016 default_idle 25.2000 1 system_call0.0156 2 handle_IRQ_event 0.0179 1 schedule 0.0012 1 __run_task_queue 0.0104 4 follow_page0.0312 3 get_user_pages 0.0063 4 mark_dirty_kiobuf 0.0357 1 handle_mm_fault0.0048 2 generic_file_direct_IO 0.0026 1 lru_cache_add 0.0089 2 init_buffer0.0625 12 set_bh_page0.1071 3 create_buffers 0.0134 30 generic_direct_IO 0.0457 17 brw_kiovec 0.0186 11 max_block 0.0764 9 blkdev_get_block 0.1406 2 blkdev_direct_IO 0.0417 1 write_profile 0.0208 2 generic_unplug_device 0.0312 93 __make_request 0.0524 25 generic_make_request 0.0781 20 submit_bh 0.0781 5 ide_inb0.3125 12 ide_outb 0.7500 1 ide_outl 0.0625 1 ide_wait_stat 0.0030 4 ide_execute_command0.0312 2 ide_do_request 0.0050 4 ide_get_queue 0.0625 2 ide_intr 0.0069 2 ide_dma_intr 0.0104 10 ide_build_sglist 0.0160 1 ide_build_dmatable 0.0024 1 __ide_dma_read 0.0042 1 __constant_c_and_count_memset 0.0069 1 __ide_do_rw_disk 0.0007 42 idedisk_end_request0.2188 4 __rdtsc_delay 0.1250 0 *unknown* 2356 total 0.0013 1 sched_clock0.0057 3 cond_resched_lock 0.0312 2 write_profile 0.0312 1 __wake_up_bit 0.0208 2 __generic_file_aio_read0.0038 1 generic_file_read 0.0052 1 mempool_alloc 0.0037 1 set_page_dirty 0.0125 3 set_page_dirty_lock0.0625 4 __follow_page 0.0227 7 get_user_pages 0.0072 2 do_wp_page 0.0025 8 __bio_add_page 0.0227 5 bio_add_page 0.1562 1 bio_set_map_data 0.0208 1 update_atime 0.0057 1 dio_get_page 0.0156 1 dio_bio_submit 0.0069 2 dio_bio_complete
Re: [git patches] IDE update
On Tue, 5 Jul 2005, Jens Axboe wrote: Looks interesting, 2.6 spends oodles of times copying to user space. Lets check if raw reads perform ok, please try and time this app in 2.4 and 2.6 as well. I think it's just that 2.4.x used to allow longer command queues. I think MAX_NR_REQUESTS is 1024 in 2.4.x, and just 128 in 2.6.x or something like that. Also, the congestion thresholds are questionable: we consider a queue congested if it is within 12% of full, but then we consider it uncongested whenever it falls to within 18% of full, which I bet means that for some streaming loads we have just a 6% window that we keep adding new requests to (we wait when we're almost full, but then we start adding requests again when we're _still_ almost full). Jens, we talked about this long ago, but I don't think we ever did any timings. Making things worse, things like this are only visible on stupid hardware that has long latencies to get started (many SCSI controllers used to have horrid latencies), so you'll never even see any difference on a lot of hardware. It's probably worth testing with a bigger request limit. I forget what the /proc interfaces are (and am too lazy to look it up), Jens can tell us ;) Linus - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
On Tue, Jul 05 2005, Linus Torvalds wrote: On Tue, 5 Jul 2005, Jens Axboe wrote: Looks interesting, 2.6 spends oodles of times copying to user space. Lets check if raw reads perform ok, please try and time this app in 2.4 and 2.6 as well. I think it's just that 2.4.x used to allow longer command queues. I think MAX_NR_REQUESTS is 1024 in 2.4.x, and just 128 in 2.6.x or something like that. But for this case, you only have one command in flight. hdparm is highly synchronous, my oread case is as well. Also, the congestion thresholds are questionable: we consider a queue congested if it is within 12% of full, but then we consider it uncongested whenever it falls to within 18% of full, which I bet means that for some streaming loads we have just a 6% window that we keep adding new requests to (we wait when we're almost full, but then we start adding requests again when we're _still_ almost full). Jens, we talked about this long ago, but I don't think we ever did any timings. In theory, the ioc batching should handle that case. But as you can see from recent commits, I'm not very happy with how this currently works. It should not impact this testing, though. Making things worse, things like this are only visible on stupid hardware that has long latencies to get started (many SCSI controllers used to have horrid latencies), so you'll never even see any difference on a lot of hardware. IDE still has much lower overhead per command than your average SCSI hardware. SATA with FIS even improves on this, definitely a good thing! It's probably worth testing with a bigger request limit. I forget what the /proc interfaces are (and am too lazy to look it up), Jens can tell us ;) It's /sys/block/device/queue/nr_requests now, can be changed at will. -- Jens Axboe - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
On Tue, Jul 05 2005, Ondrej Zary wrote: Jens Axboe wrote: On Tue, Jul 05 2005, Ondrej Zary wrote: Jens Axboe wrote: On Tue, 2005-07-05 at 15:02 +0200, Ondrej Zary wrote: Ok, looks alright for both. Your machine is quite slow, perhaps that is showing the slower performance. Can you try and make HZ 100 in 2.6 and test again? 2.6.13-recent has it as a config option, otherwise edit include/asm/param.h appropriately. I forgot to write that my 2.6.12 kernel is already compiled with HZ 100 (it makes the system more responsive). I've just tried 2.6.8.1 with HZ 1000 and there is no difference in HDD performance comparing to 2.6.12. OK, interesting. You could try and boot with profile=2 and do # readprofile -r # dd if=/dev/hda of=/dev/null bs=128k # readprofile prof_output for each kernel and post it here, so we can see if anything sticks out. Here are the profiles (used dd with count=4096) from 2.4.26 and 2.6.12 (nothing from 2.6.8.1 because I don't have the .map file anymore). Looks interesting, 2.6 spends oodles of times copying to user space. Lets check if raw reads perform ok, please try and time this app in 2.4 and 2.6 as well. # gcc -Wall -O2 -o oread oread.c # time ./oread /dev/hda oread is faster than dd, but still not as fast as 2.4. In 2.6.12, HDD led is blinking, in 2.4 it's solid on during the read. 2.6.12: [EMAIL PROTECTED]:/home/rainbow# time ./oread /dev/hda real0m25.082s user0m0.000s sys 0m0.680s 2.4.26: [EMAIL PROTECTED]:/home/rainbow# time ./oread /dev/hda real0m23.513s user0m0.000s sys 0m2.360s Hmm, still not as fast, not so good. 2.6 shows more idle time than 2.4, about 20% more. I seem to remember Ken Chen saying that 2.6 direct io was still a little slower than 2.4, your really slow hardware could be showing this to a much greater effect. I'll try and play with this tomorrow! -- Jens Axboe - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
On Tue, Jul 05 2005, Ondrej Zary wrote: oread is faster than dd, but still not as fast as 2.4. In 2.6.12, HDD led is blinking, in 2.4 it's solid on during the read. Oh, and please do test 2.6 by first setting the deadline scheduler for hda. I can see you are using the 'as' scheduler right now. # echo deadline /sys/block/hda/queue/scheduler Thanks! -- Jens Axboe - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
Jens Axboe wrote: On Tue, Jul 05 2005, Ondrej Zary wrote: oread is faster than dd, but still not as fast as 2.4. In 2.6.12, HDD led is blinking, in 2.4 it's solid on during the read. Oh, and please do test 2.6 by first setting the deadline scheduler for hda. I can see you are using the 'as' scheduler right now. # echo deadline /sys/block/hda/queue/scheduler No change, still 25 seconds. -- Ondrej Zary - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
Jens Axboe wrote: On Tue, Jul 05 2005, Linus Torvalds wrote: On Tue, 5 Jul 2005, Jens Axboe wrote: Looks interesting, 2.6 spends oodles of times copying to user space. Lets check if raw reads perform ok, please try and time this app in 2.4 and 2.6 as well. I think it's just that 2.4.x used to allow longer command queues. I think MAX_NR_REQUESTS is 1024 in 2.4.x, and just 128 in 2.6.x or something like that. But for this case, you only have one command in flight. hdparm is highly synchronous, my oread case is as well. Also, the congestion thresholds are questionable: we consider a queue congested if it is within 12% of full, but then we consider it uncongested whenever it falls to within 18% of full, which I bet means that for some streaming loads we have just a 6% window that we keep adding new requests to (we wait when we're almost full, but then we start adding requests again when we're _still_ almost full). Jens, we talked about this long ago, but I don't think we ever did any timings. In theory, the ioc batching should handle that case. But as you can see from recent commits, I'm not very happy with how this currently works. It should not impact this testing, though. Making things worse, things like this are only visible on stupid hardware that has long latencies to get started (many SCSI controllers used to have horrid latencies), so you'll never even see any difference on a lot of hardware. IDE still has much lower overhead per command than your average SCSI hardware. SATA with FIS even improves on this, definitely a good thing! It's probably worth testing with a bigger request limit. I forget what the /proc interfaces are (and am too lazy to look it up), Jens can tell us ;) It's /sys/block/device/queue/nr_requests now, can be changed at will. Tested with default 128, 1024 and 4 (minimum) and no change. -- Ondrej Zary - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
On Tue, 5 Jul 2005 16:21:26 +0200, Jens Axboe [EMAIL PROTECTED] wrote: # gcc -Wall -O2 -o oread oread.c # time ./oread /dev/hda Executive Summary `` Comparing 'oread' with hdparm -tT on latest 2.4 vs 2.6 stable on various x86 boxen. Performance drops for 2.6, sometimes: 2.4.31-hf1 2.6.12.2 peetoo: 26MB/s - 20MB/s, oread similar120GB + 80GB silly: 22MB/s - 8.5MB/s, oread similar13GB tosh: 35MB/s - 23MB/s, oread similar40GB 2.5 pooh: 17MB/s - 14MB/s, oread 30 - 24 40GB menace: 11.5MB/s - 13MB/s, oread similar 6GB 2.5 --Grant The details, config, dmesg, etc from linked resources: Intel SE440BX-2 with pII 400/100/512/2.0 CPU/FSB/L2/Vccp 512MB SDRAM on 440BX + PIIX4 http://scatter.mine.nu/test/linux-2.6/peetoo/ peetoo:~$ uname -r 2.4.31-hf1 peetoo:~$ time /home/share/install/oread /dev/hda real0m20.065s user0m0.010s sys 0m0.810s peetoo:~$ time /home/share/install/oread /dev/hdc real0m18.484s user0m0.000s sys 0m0.800s peetoo:~$ hdparm -tT /dev/hda /dev/hda: Timing cached reads: 444 MB in 2.00 seconds = 222.00 MB/sec Timing buffered disk reads: 84 MB in 3.06 seconds = 27.45 MB/sec peetoo:~$ hdparm -tT /dev/hda /dev/hda: Timing cached reads: 456 MB in 2.01 seconds = 226.87 MB/sec Timing buffered disk reads: 76 MB in 3.00 seconds = 25.33 MB/sec peetoo:~$ hdparm -tT /dev/hdc /dev/hdc: Timing cached reads: 464 MB in 2.01 seconds = 230.85 MB/sec Timing buffered disk reads: 76 MB in 3.00 seconds = 25.33 MB/sec peetoo:~$ hdparm -tT /dev/hdc /dev/hdc: Timing cached reads: 460 MB in 2.00 seconds = 230.00 MB/sec Timing buffered disk reads: 82 MB in 3.00 seconds = 27.33 MB/sec - - - peetoo:~$ uname -r 2.6.12.2b peetoo:~$ time /home/share/install/oread /dev/hda real0m21.428s user0m0.003s sys 0m0.436s peetoo:~$ time /home/share/install/oread /dev/hdc real0m17.588s user0m0.001s sys 0m0.455s peetoo:~$ cat /sys/block/hda/queue/scheduler noop [deadline] peetoo:~$ cat /sys/block/hdc/queue/scheduler noop [deadline] peetoo:~$ time /home/share/install/oread /dev/hda real0m21.417s user0m0.005s sys 0m0.462s peetoo:~$ time /home/share/install/oread /dev/hda real0m18.626s user0m0.006s sys 0m0.440s peetoo:~$ time /home/share/install/oread /dev/hdc real0m17.555s user0m0.005s sys 0m0.422s peetoo:~$ df Filesystem 1K-blocks Used Available Use% Mounted on /dev/hda3 2586348563000 2023348 22% / /dev/hdc3 2586348 1582716 1003632 62% /usr /dev/hdc6 2586348 1217568 1368780 48% /usr/src /dev/hda9 20562504 10821500 9741004 53% /home/install /dev/hdc9 20562504 3634700 16927804 18% /home/public /dev/hda1041446344 39676256 1770088 96% /home/archive deltree:/home/share2064256 1042952 1021304 51% /home/share peetoo:~$ hdparm -tT /dev/hda /dev/hda: Timing cached reads: 448 MB in 2.00 seconds = 223.81 MB/sec Timing buffered disk reads: 62 MB in 3.07 seconds = 20.17 MB/sec peetoo:~$ hdparm -tT /dev/hda /dev/hda: Timing cached reads: 472 MB in 2.01 seconds = 234.40 MB/sec Timing buffered disk reads: 54 MB in 3.01 seconds = 17.97 MB/sec peetoo:~$ hdparm -tT /dev/hdc /dev/hdc: Timing cached reads: 456 MB in 2.01 seconds = 226.56 MB/sec Timing buffered disk reads: 62 MB in 3.08 seconds = 20.11 MB/sec peetoo:~$ hdparm -tT /dev/hdc o o o EPoX EP-61LXA-M: Intel 440LX chipset with pentiumII/266 on 66MHz FSB (4 x 66) with 128MB SDRAM http://scatter.mine.nu/test/linux-2.6/silly/ [EMAIL PROTECTED]:~# uname -r 2.4.31-hf1 [EMAIL PROTECTED]:~# time /home/share/install/oread /dev/hda real0m23.657s user0m0.000s sys 0m1.300s [EMAIL PROTECTED]:~# hdparm -tT /dev/hda /dev/hda: Timing cached reads: 356 MB in 2.02 seconds = 176.24 MB/sec Timing buffered disk reads: 68 MB in 3.04 seconds = 22.37 MB/sec [EMAIL PROTECTED]:~# hdparm -tT /dev/hda /dev/hda: Timing cached reads: 344 MB in 2.00 seconds = 172.00 MB/sec Timing buffered disk reads: 68 MB in 3.03 seconds = 22.44 MB/sec - - - [EMAIL PROTECTED]:~# uname -r 2.6.12.2a [EMAIL PROTECTED]:~# time /home/share/install/oread /dev/hda real0m23.569s user0m0.005s sys 0m0.563s [EMAIL PROTECTED]:~# hdparm -tT /dev/hda /dev/hda: Timing cached reads: 336 MB in 2.00 seconds = 167.77 MB/sec Timing buffered disk reads: 50 MB in 3.05 seconds = 16.37 MB/sec [EMAIL PROTECTED]:~# hdparm -tT /dev/hda /dev/hda: Timing cached reads: 340 MB in 2.01 seconds = 169.01 MB/sec Timing buffered disk reads: 28 MB in 3.22 seconds = 8.70 MB/sec [EMAIL PROTECTED]:~# hdparm -tT /dev/hda /dev/hda: Timing cached reads: 336 MB in 2.01 seconds = 167.44 MB/sec Timing buffered disk reads: 26 MB in 3.05 seconds = 8.52 MB/sec [EMAIL PROTECTED]:~# echo deadline /sys/block/hda/queue/scheduler [EMAIL PROTECTED]:~# hdparm -tT /dev/hda
RE: [git patches] IDE update
Linus Torvalds wrote: { On Wed, 6 Jul 2005, Grant Coady wrote: Executive Summary Btw, can you try this same thing (or at least a subset) with a large file on a filesystem? Does that show the same pattern, or is it always just the raw device? } Linus, Cat /dev/hda /dev/null and cat /tmp/tst.dsk /dev/null show the same symptoms. The problem shows most when the cpu is slow and the hd is fast. When the cpu is fast and the hd is slow the cpu will make up for lost cycles and the problem will not show! - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
On Tue, 5 Jul 2005 17:51:50 -0700 (PDT), Linus Torvalds [EMAIL PROTECTED] wrote: Btw, can you try this same thing (or at least a subset) with a large file on a filesystem? Does that show the same pattern, or is it always just the raw device? Sure, take a while longer to vary by block size. One effect seems to be wrong is interaction between /dev/hda and /dev/hdc in 'peetoo', the IDE channels not independent? write: time $(dd if=/dev/zero bs=1M count=500 of=/zeroes; sync) read: time dd of=/dev/null bs=1M if=/zeroes summary 2.4.31-hf1 2.6.12.2 boxen \ time - w r w r --- menace 58.550 57 47.5 pooh24 24 22.527 peetoo 33 20 26.522 (simultaneuous 57 37.552 38.5) silly 54 24 49 25 tosh30 19.527 19.5 filesystem: reiserfs 3.6, distro: slackware-10.1 + updates hardware config, etc: http://scatter.mine.nu/test/ --Grant the long story: [EMAIL PROTECTED]:~# uname -r 2.4.31-hf1 [EMAIL PROTECTED]:~# df Filesystem 1K-blocks Used Available Use% Mounted on /dev/hda3 1991992 1074516917476 54% / deltree:/home/share2064256 1042968 1021288 51% /home/share [EMAIL PROTECTED]:~# time $(dd if=/dev/zero bs=1M count=500 of=/zeroes; sync) 500+0 records in 500+0 records out real0m57.830s user0m0.050s sys 0m20.940s [EMAIL PROTECTED]:~# time $(dd if=/dev/zero bs=1M count=500 of=/zeroes; sync) 500+0 records in 500+0 records out real0m59.041s user0m0.030s sys 0m21.780s [EMAIL PROTECTED]:~# time dd of=/dev/null bs=1M if=/zeroes 500+0 records in 500+0 records out real0m49.963s user0m0.000s sys 0m15.510s - - - [EMAIL PROTECTED]:~# uname -r 2.6.12.2a [EMAIL PROTECTED]:~# time $(dd if=/dev/zero bs=1M count=500 of=/zeroes; sync) 500+0 records in 500+0 records out real0m57.199s user0m0.022s sys 0m15.040s [EMAIL PROTECTED]:~# time $(dd if=/dev/zero bs=1M count=500 of=/zeroes; sync) 500+0 records in 500+0 records out real0m56.825s user0m0.024s sys 0m14.893s [EMAIL PROTECTED]:~# time dd of=/dev/null bs=1M if=/zeroes 500+0 records in 500+0 records out real0m47.560s user0m0.017s sys 0m15.533s [EMAIL PROTECTED]:~# time dd of=/dev/null bs=1M if=/zeroes 500+0 records in 500+0 records out real0m47.842s user0m0.012s sys 0m15.647s o o o [EMAIL PROTECTED]:~# df Filesystem 1K-blocks Used Available Use% Mounted on /dev/hda3 3084380 2018240 1066140 66% / /dev/hda6 2056220 1049544 1006676 52% /usr/src /dev/hda7 256996 34260222736 14% /usr/local /dev/hda8 256996 33896223100 14% /home /dev/hda1420562536 32840 20529696 1% /home/pooh deltree:/home/share2064256 1042968 1021288 51% /home/share [EMAIL PROTECTED]:~# uname -r 2.4.31-hf1 [EMAIL PROTECTED]:~# time $(dd if=/dev/zero bs=1M count=500 of=/zeroes; sync) 500+0 records in 500+0 records out real0m23.912s user0m0.010s sys 0m19.820s [EMAIL PROTECTED]:~# time dd of=/dev/null bs=1M if=/zeroes 500+0 records in 500+0 records out real0m24.302s user0m0.020s sys 0m16.760s - - - [EMAIL PROTECTED]:~# uname -r 2.6.12.2a [EMAIL PROTECTED]:~# time $(dd if=/dev/zero bs=1M count=500 of=/zeroes; sync) 500+0 records in 500+0 records out real0m22.449s user0m0.017s sys 0m13.576s [EMAIL PROTECTED]:~# time dd of=/dev/null bs=1M if=/zeroes 500+0 records in 500+0 records out real0m26.780s user0m0.010s sys 0m13.398s o o o peetoo:~$ df Filesystem 1K-blocks Used Available Use% Mounted on /dev/hda3 2586348 1075652 1510696 42% / /dev/hdc3 2586348 2044228542120 80% /usr /dev/hdc6 2586348 1217568 1368780 48% /usr/src /dev/hda9 20562504 10821500 9741004 53% /home/install /dev/hdc9 20562504 4329320 16233184 22% /home/public /dev/hda1041446344 39676256 1770088 96% /home/archive deltree:/home/share2064256 1042968 1021288 51% /home/share peetoo:~$ uname -r 2.4.31-hf1 peetoo:~$ time $(dd if=/dev/zero bs=1M count=500 of=/zeroes; sync) 500+0 records in 500+0 records out real0m33.322s user0m0.000s sys 0m13.650s peetoo:~$ time $(dd if=/dev/zero bs=1M count=500 of=/zeroes; sync) 500+0 records in 500+0 records out real0m32.983s user0m0.010s sys 0m13.740s peetoo:~$ time $(dd if=/dev/zero bs=1M count=500 of=/usr/zeroes; sync) 500+0 records in 500+0 records out real0m30.775s user0m0.000s sys 0m13.600s peetoo:~$ time $(dd if=/dev/zero bs=1M count=500 of=/usr/zeroes; sync) 500+0 records in 500+0 records out real0m33.077s user0m0.010s sys 0m13.740s peetoo:~$ time dd of=/dev/null bs=1M if=/zeroes 500+0 records in 500+0 records out real
RE: [git patches] IDE update
Bartlomiej Zolnierkiewicz wrote: { Please pull from: rsync://rsync.kernel.org/pub/scm/linux/kernel/git/bart/ide-2.6.git } Does it fix the idedriver int/dma problem? Al - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
On 7/4/05, Al Boldi [EMAIL PROTECTED] wrote: Bartlomiej Zolnierkiewicz wrote: { Please pull from: rsync://rsync.kernel.org/pub/scm/linux/kernel/git/bart/ide-2.6.git } Does it fix the idedriver int/dma problem? What is the int/dma problem? - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
RE: [git patches] IDE update
Bartlomiej Zolnierkiewicz wrote: { What is the int/dma problem? } Hdparm -tT gives 38mb/s in 2.4.31 Cat /dev/hda /dev/null gives 2% user 33% sys 65% idle Hdparm -tT gives 28mb/s in 2.6.12 Cat /dev/hda /dev/null gives 2% user 25% sys 0% idle 73% IOWAIT It feels like DMA is not being applied properly in 2.6.12. Your comments please. Thanks! - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
On 7/4/05, Al Boldi [EMAIL PROTECTED] wrote: Bartlomiej Zolnierkiewicz wrote: { What is the int/dma problem? } Hdparm -tT gives 38mb/s in 2.4.31 Cat /dev/hda /dev/null gives 2% user 33% sys 65% idle Hdparm -tT gives 28mb/s in 2.6.12 Cat /dev/hda /dev/null gives 2% user 25% sys 0% idle 73% IOWAIT It feels like DMA is not being applied properly in 2.6.12. Your comments please. Are earlier 2.6.x kernels okay? dmesg output? Bartlomiej - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
RE: [git patches] IDE update
Bartlomiej Zolnierkiewicz wrote: { On 7/4/05, Al Boldi [EMAIL PROTECTED] wrote: Hdparm -tT gives 38mb/s in 2.4.31 Cat /dev/hda /dev/null gives 2% user 33% sys 65% idle Hdparm -tT gives 28mb/s in 2.6.12 Cat /dev/hda /dev/null gives 2% user 25% sys 0% idle 73% IOWAIT It feels like DMA is not being applied properly in 2.6.12. Are earlier 2.6.x kernels okay? dmesg output? } Same on 2.6.10,11,12. No errors though, only sluggish system. - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
Al Boldi wrote: Bartlomiej Zolnierkiewicz wrote: { On 7/4/05, Al Boldi [EMAIL PROTECTED] wrote: Hdparm -tT gives 38mb/s in 2.4.31 Cat /dev/hda /dev/null gives 2% user 33% sys 65% idle Hdparm -tT gives 28mb/s in 2.6.12 Cat /dev/hda /dev/null gives 2% user 25% sys 0% idle 73% IOWAIT It feels like DMA is not being applied properly in 2.6.12. Are earlier 2.6.x kernels okay? dmesg output? } Same on 2.6.10,11,12. No errors though, only sluggish system. Something like this http://lkml.org/lkml/2005/6/13/1 ? -- Ondrej Zary - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
On 7/4/05, Ondrej Zary [EMAIL PROTECTED] wrote: Al Boldi wrote: Bartlomiej Zolnierkiewicz wrote: { On 7/4/05, Al Boldi [EMAIL PROTECTED] wrote: Hdparm -tT gives 38mb/s in 2.4.31 Cat /dev/hda /dev/null gives 2% user 33% sys 65% idle Hdparm -tT gives 28mb/s in 2.6.12 Cat /dev/hda /dev/null gives 2% user 25% sys 0% idle 73% IOWAIT It feels like DMA is not being applied properly in 2.6.12. Are earlier 2.6.x kernels okay? dmesg output? } Same on 2.6.10,11,12. No errors though, only sluggish system. What about earlier kernels? Please try to narrow down the problem to a specific kernel version. Something like this http://lkml.org/lkml/2005/6/13/1 ? It doesn't seem like IDE regression but... Bartlomiej - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
On 7/4/05, Al Boldi [EMAIL PROTECTED] wrote: Bartlomiej Zolnierkiewicz wrote: { On 7/4/05, Al Boldi [EMAIL PROTECTED] wrote: Hdparm -tT gives 38mb/s in 2.4.31 Cat /dev/hda /dev/null gives 2% user 33% sys 65% idle Hdparm -tT gives 28mb/s in 2.6.12 Cat /dev/hda /dev/null gives 2% user 25% sys 0% idle 73% IOWAIT It feels like DMA is not being applied properly in 2.6.12. Same on 2.6.10,11,12. No errors though, only sluggish system. What about earlier kernels? Please try to narrow down the problem to a specific kernel version. } Don't know about 2.6.0-2.6.9, but 2.4.31 is ok. 2.4 - 2.6 means zillions of changes. Bartlomiej, When you compare 2.4.31 with 2.6.12 don't you see this problem on your machine? Unfortunately I've never encountered this problem on any machine. If you have a fast system the slowdown won't show, but your IOWAIT will be higher anyway! AFAIK high iowait is a known block layer problem It feels like running on PIO instead of DMA. - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [git patches] IDE update
Al Boldi wrote: Bartlomiej Zolnierkiewicz wrote: { On 7/4/05, Al Boldi [EMAIL PROTECTED] wrote: Hdparm -tT gives 38mb/s in 2.4.31 Cat /dev/hda /dev/null gives 2% user 33% sys 65% idle Hdparm -tT gives 28mb/s in 2.6.12 Cat /dev/hda /dev/null gives 2% user 25% sys 0% idle 73% IOWAIT The hdparm doesn't get as high scores as in 2.4 is a old discussed to death problem on LKML. So far nobody has been able to show it affects anything but that pretty useless quasi-benchmark. It feels like DMA is not being applied properly in 2.6.12. Same on 2.6.10,11,12. No errors though, only sluggish system. Really sluggish or just benchmark-sluggish? If the former, try selecting a different IO elevator/sheduler. If the latter it doesn't matter much, at least not with the very simple hdparm test :-) What about earlier kernels? Please try to narrow down the problem to a specific kernel version. } Don't know about 2.6.0-2.6.9, but 2.4.31 is ok. Bartlomiej, When you compare 2.4.31 with 2.6.12 don't you see this problem on your machine? If you have a fast system the slowdown won't show, but your IOWAIT will be higher anyway! Nothing wrong with 73% iowait, I'd even consider it very low while putting load on a harddrive. Its just time spent waiting for data to be returned from disk, and thus I usually expect no lower than ~98-99% while stressing any disk. Harddisks are _slow as snails_ compared to cpu cycles ;-) Beware 2.4 didn't export that statistic at all to userspace, so 0% iowait gets reported from most 2.6-ready reporting tools on 2.4. - To unsubscribe from this list: send the line unsubscribe linux-ide in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html