[Bug 493201] New: liveCD performance is poor
http://bugzilla.novell.com/show_bug.cgi?id=493201 Summary: liveCD performance is poor Classification: openSUSE Product: openSUSE 11.2 Version: Factory Platform: Other OS/Version: Other Status: NEW Severity: Major Priority: P5 - None Component: Kernel AssignedTo: teheo@novell.com ReportedBy: coolo@novell.com QAContact: qa@suse.de Found By: Development Created an attachment (id=284754) --> (http://bugzilla.novell.com/attachment.cgi?id=284754) boot.msg Reading from CD when installed is ok, but when booting the live cd, it takes _forever_ and the CD reading performance is bad then too. The CD driver reads 105MB in 55,45s (1.9MB/s) when running from an installed factory, but needs 224.26s (468KB/s) when booted from live cd. The drive details: /dev/sr0: ATAPI CD-ROM, with removable media Model Number: MATSHITADVD-RAM UJ-842S Serial Number: Firmware Revision: 1.10 Standards: Likely used CD-ROM ATAPI-1 Configuration: DRQ response: 50us. Packet size: 12 bytes cache/buffer size = unknown Capabilities: LBA, IORDY(can be disabled) DMA: sdma0 sdma1 sdma2 mdma0 mdma1 *mdma2 Cycle time: min=120ns recommended=120ns PIO: pio0 pio1 pio2 pio3 pio4 Cycle time: no flow control=240ns IORDY flow control=120ns HW reset results: CBLID- below Vih Device num = 1 Attached is boot.msg when booting from live cd. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User coolo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c1 --- Comment #1 from Stephan Kulow <coolo@novell.com> 2009-04-08 06:46:12 MDT --- Created an attachment (id=284757) --> (http://bugzilla.novell.com/attachment.cgi?id=284757) boot.msg from hdd boot -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User coolo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c2 --- Comment #2 from Stephan Kulow <coolo@novell.com> 2009-04-08 06:51:28 MDT --- please note that it's a i686 livecd while my installation is 64bit. So there are some differences derived from that. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User coolo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c3 --- Comment #3 from Stephan Kulow <coolo@novell.com> 2009-04-08 06:57:07 MDT --- just tested the dd speed from Fedora10 live cd: 55.4s - 2.6.27.5 -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User teheo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c4 Tejun Heo <teheo@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |NEEDINFO Info Provider| |coolo@novell.com --- Comment #4 from Tejun Heo <teheo@novell.com> 2009-04-08 15:00:31 MDT --- Hmmm... I couldn't find anything which indicate that the two devices are configured differently. Can you please try the followings? 1. pata_acpi is being loaded for live CD. It fails during initialization and detaches but it might as well be that one of the acpi calls are messing up the configuration. pata_acpi is blacklisted by default for !live installations but it seems kiwi didn't pick it up. Can you please exclude pata_acpi and see what happens? 2. If that doesn't help, can you please test x86_64 live CD too? Thanks. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User coolo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c5 Stephan Kulow <coolo@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |ms@novell.com, | |snwint@novell.com --- Comment #5 from Stephan Kulow <coolo@novell.com> 2009-04-09 04:10:07 MDT --- interesting. the live cd does whatever hwinfo --storage tells it to call. And it has pata_acpi as first module. So the module order is pata_acpi ata_piix uchi-hcd ehci-hcd So the blacklist isn't taken into account. If I remove the module from the initrd, I get 922KB/s later into the boot - and the module is for whatever reason still loaded . Which is a strange value. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User snwint@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c6 --- Comment #6 from Steffen Winterfeldt <snwint@novell.com> 2009-04-09 04:52:41 MDT --- IIRC Marcus kept hwinfo for module loading because there were problems with udevd loading them. Maybe time to change that (or respect the blacklist). -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User coolo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c7 Stephan Kulow <coolo@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |NEW Info Provider|coolo@novell.com | --- Comment #7 from Stephan Kulow <coolo@novell.com> 2009-04-09 05:37:50 MDT ---
From the changelog, this seems to be the reason: https://bugzilla.novell.com/show_bug.cgi?id=354278
-- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User teheo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c8 --- Comment #8 from Tejun Heo <teheo@novell.com> 2009-04-09 06:04:22 MDT --- Aieee... kiwi using completely separate initrd hurts. We pretty much have to load all matching drivers (sans blacklisted ones) in the order specified by modules.order which udev/modprobe does fine. This is for the following two reasons. * Some controllers can switch between two different modes of operation but the selection is made by the bios (ata_piix/ahci for intel ICHs). In this case, the initrd _should_ carry both drivers and load them both in case the user switched the mode. * Some controllers can't be fully matched using the pci device ID mechanism used for module selection. Different generations of HPT drivers share the same PCI IDs but need to use different drivers. Only the driver init function knows whether it can attach to the controller or not, so unless all the matching drivers are loaded, some controllers won't get detected properly. By having two separate initrds, we're again facing problems which were solved quite some time ago. Is there any way not to do this? Thanks. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 Stephan Kulow <coolo@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |teheo@novell.com AssignedTo|teheo@novell.com |ms@novell.com -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User snwint@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c9 --- Comment #9 from Steffen Winterfeldt <snwint@novell.com> 2009-04-09 06:26:38 MDT --- I think the bug coolo mentioned was before modules.order. Which, correct me if I'm wrong, the kernel people got right just in time for 11.1. So, it might be a bit unfair to call kiwi that much backward. :-) But in principle you're right and we should move to udevd module loading here. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User teheo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c10 --- Comment #10 from Tejun Heo <teheo@novell.com> 2009-04-09 06:46:35 MDT --- I wasn't trying to criticize kiwi. I apologize if it came out like that. What I'm concerned about is having two completely separate initrds. There have been and will continue to be fixes which make for certain releases just in time. That's what the release process is for after all. The thing is that we're now basically doing that twice. initrd is a fairly complex and important part of the whole system and it's also a place where a lot of workarounds and black magics happen. It just doesn't make any sense to do it twice. If kiwi initrd is the way to go, let's standardize on that. If not, let's drop it and expand the original initrd to cover kiwi cases. Keeping both around is the worst choice we can make. Thanks. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User snwint@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c11 --- Comment #11 from Steffen Winterfeldt <snwint@novell.com> 2009-04-09 07:21:27 MDT --- Tejun, keep cool and note the smiley in my comment. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User teheo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c12 --- Comment #12 from Tejun Heo <teheo@novell.com> 2009-04-09 07:25:53 MDT --- Yeah, still quite jet lagged. Sleep deprivation does wonders to sense of humor. :-) -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User ms@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c13 Marcus Schaefer <ms@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |NEEDINFO Info Provider| |snwint@novell.com Summary|liveCD performance is poor |kiwi: module black list not | |used --- Comment #13 from Marcus Schaefer <ms@novell.com> 2009-04-20 03:46:23 MDT --- In reply to comment #5 I was told calling hwinfo --storage and selecting "Driver Info #0" is a safe way to load the correct module. You said there is a blacklist I should handle ? What's this where can I find it and why does hwinfo not use that blacklist ? Thanks -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 Marcus Schaefer <ms@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Severity|Major |Normal -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User coolo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c14 --- Comment #14 from Stephan Kulow <coolo@novell.com> 2009-04-20 06:29:40 MDT --- it's not just about the blacklist but also about modules.order btw. I guess hwinfo should either know all or nothing about modules. The current situation is bad. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User snwint@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c15 Steffen Winterfeldt <snwint@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |NEW Info Provider|snwint@novell.com | --- Comment #15 from Steffen Winterfeldt <snwint@novell.com> 2009-04-20 06:55:12 MDT --- The mentioned list is /etc/modprobe.d/blacklist. hwinfo just reports the driver situation. It does not load modules and for that reason does not look at any blacklists. With modules.order there is no reason left not to let udev handle the module setup. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User ms@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c16 Marcus Schaefer <ms@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |NEEDINFO Info Provider| |snwint@novell.com --- Comment #16 from Marcus Schaefer <ms@novell.com> 2009-04-20 07:04:39 MDT --- does that mean udev takes care for the blacklist and also for the module load order now ? in other words if there is more than one alternative udev now picks the best solution ? Is that correct, really correct ? :) ... for 11.2 ? thanks -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User coolo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c17 --- Comment #17 from Stephan Kulow <coolo@novell.com> 2009-04-20 07:45:58 MDT --- actually for 11.1 already -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User snwint@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c18 Steffen Winterfeldt <snwint@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |NEW Info Provider|snwint@novell.com | --- Comment #18 from Steffen Winterfeldt <snwint@novell.com> 2009-04-20 07:48:27 MDT --- yes -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User ms@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c19 Marcus Schaefer <ms@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |NEEDINFO Info Provider| |snwint@novell.com --- Comment #19 from Marcus Schaefer <ms@novell.com> 2009-04-20 08:29:51 MDT --- ok, I will change that. To support older versions I need a switch Is it ok if I call udevd --version
= 140 ---> let udev load all modules
< 140 ---> search modules via hwinfo --storage -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User snwint@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c20 Steffen Winterfeldt <snwint@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |NEW Info Provider|snwint@novell.com | --- Comment #20 from Steffen Winterfeldt <snwint@novell.com> 2009-04-20 09:03:48 MDT --- I wouldn't relate this to udev. Checking for modules.order might be more logical. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User ms@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c21 --- Comment #21 from Marcus Schaefer <ms@novell.com> 2009-04-20 09:26:29 MDT --- yep good point will do it that way. Thanks -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 Marcus Schaefer <ms@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Priority|P5 - None |P2 - High Status|NEW |ASSIGNED -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User ms@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c22 Marcus Schaefer <ms@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |RESOLVED Resolution| |FIXED --- Comment #22 from Marcus Schaefer <ms@novell.com> 2009-04-21 06:58:59 MDT --- fixed in kiwi v3.43 -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User coolo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c23 Stephan Kulow <coolo@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|RESOLVED |REOPENED Resolution|FIXED | AssignedTo|ms@novell.com |teheo@novell.com --- Comment #23 from Stephan Kulow <coolo@novell.com> 2009-04-22 03:16:02 MDT --- now that kiwi does no longer use hwinfo, the pata_acpi module is not loaded (according to boot.kiwi and lsmod). But I still only get 470K/s - for the Milestone1 KDE-Live-i586 -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User coolo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c24 --- Comment #24 from Stephan Kulow <coolo@novell.com> 2009-04-22 03:29:19 MDT --- The list of modules loaded is different between the i586 live (t2) and the x86_64 installation (t1): Some are expected, but what about e.g. ide_cd_mod ? --- t1 2009-04-22 11:26:41.000000000 +0200 +++ t2 2009-04-22 11:26:46.000000000 +0200 @@ -1,14 +1,16 @@ ac acpi_cpufreq af_packet +agpgart +ahci arc4 ata_generic ata_piix auth_rpcgss -autofs4 battery bluetooth bnep +brd bridge btusb button @@ -18,23 +20,26 @@ cpufreq_conservative cpufreq_powersave cpufreq_userspace -crc16 crc_t10dif +dm_mod drm e1000e ecb edd ehci_hcd -ext4 +ext3 fan fat fuse +hid hp_accel hwmon i2c_algo_bit i2c_core i915 +ide_cd_mod ide_core +ide_gd_mod ide_pci_generic ieee1394 intel_agp @@ -44,10 +49,7 @@ iwlagn iwlcore jbd -jbd2 joydev -kvm -kvm_intel l2cap led_class libata @@ -62,6 +64,7 @@ nfs_acl nls_cp437 nls_iso8859_1 +nls_utf8 ohci1394 output pcmcia @@ -96,7 +99,10 @@ snd_seq_device snd_timer soundcore +speedstep_lib +squashfs sr_mod +st stp sunrpc thermal @@ -107,6 +113,7 @@ uhci_hcd uinput usbcore +usbhid usb_storage vfat video -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User coolo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c25 --- Comment #25 from Stephan Kulow <coolo@novell.com> 2009-04-22 05:16:03 MDT --- fascinating. If I boot the (x86_64) initrd from a usb stick and then do dd, I get 1.9MB/s. I'll burn the x86_64 ISO to CD now. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User coolo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c26 Stephan Kulow <coolo@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|REOPENED |NEEDINFO Info Provider| |coolo@novell.com --- Comment #26 from Stephan Kulow <coolo@novell.com> 2009-04-22 05:57:41 MDT --- ok, it's not the architecture. x86_64 ISO is just as slow, slower even. But I checked boot.msg differences between live and installation and there is at least one problem that I need to fix first: crc_t10dif is missing in the build. lib/* didn't catch it for some reason unknown to me yet. No idea why the usb initrd would fix that ;( -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User coolo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c27 Stephan Kulow <coolo@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |REOPENED Info Provider|coolo@novell.com | --- Comment #27 from Stephan Kulow <coolo@novell.com> 2009-04-22 08:36:41 MDT --- ok, that wasn't the problem. But I know now how to gain 1.9MB/s even when booting from CD: --- initrd.orig//include 2009-04-22 14:50:14.000000000 +0200 +++ initrd/include 2009-04-22 16:27:02.000000000 +0200 @@ -1765,7 +1765,7 @@ local count=0 local h=/usr/sbin/hwinfo if [ $HAVE_MODULES_ORDER = 0 ];then - for module in sg sd_mod sr_mod cdrom ide-cd BusLogic vfat; do + for module in BusLogic vfat; do /sbin/modprobe $module done fi No idea which of these modules triggers the problem, but it doesn't sound wise to mix udev and explicit modprobes of ide-cd -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User ms@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c28 --- Comment #28 from Marcus Schaefer <ms@novell.com> 2009-04-22 08:53:15 MDT --- I don't mix any modprobes. Did you see the if [ $HAVE_MODULES_ORDER = 0 ];then ... fi HAVE_MODULES_ORDER is set to 1 if lib/modules/*/modules.order exists and only in that case I let udev load the modules and the manual probes are prevented there are only the following modules which are loaded explicitly # /.../ # older systems require ide-disk to be present at any time # for details on this crappy call see bug: #250241 # ---- modprobe ide-disk &>/dev/null # /.../ # default loading of modules not loaded on demand # ---- for i in rd brd edd dm-mod xennet xenblk;do modprobe $i &>/dev/null done does one of these cause the problem ? -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User coolo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c29 --- Comment #29 from Stephan Kulow <coolo@novell.com> 2009-04-22 09:20:38 MDT --- hmm, interesting. The logs prove that the code patched isn't run :) But I actually had all modprobes commented out, so also the ide-disk. rd == brd in stable and edd and xen* aren't even in the isoboot config. I'll test in more detail tomorrow (I ordered CD-RW :) -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User coolo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c30 --- Comment #30 from Stephan Kulow <coolo@novell.com> 2009-04-22 09:30:22 MDT --- two more modprobes are in boot.kiwi: usbhid and usb-storage. usbhid is not loaded for installs, but I can't see a reason why it's needed or why it should hurt. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User coolo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c31 --- Comment #31 from Stephan Kulow <coolo@novell.com> 2009-04-23 09:19:56 MDT --- ok, ide-disk is the one to blame for one part of the problem. with ide-disk: 470K, without it: 950K. I can remove all other modprobes and even all hwinfo calls without a further change. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User coolo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c32 --- Comment #32 from Stephan Kulow <coolo@novell.com> 2009-04-23 09:22:24 MDT --- one thing that pops into my eye is this "Sense Key : Illegal Request [current]" - also in #0's boot.msg -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User teheo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c33 Tejun Heo <teheo@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|REOPENED |NEEDINFO CC| |jack@novell.com Info Provider| |jack@novell.com --- Comment #33 from Tejun Heo <teheo@novell.com> 2009-04-23 09:54:24 MDT --- cc'ing Jan. The illegal requests are probably from recent iso9660 probe update. IIUC, new code may try to access sectors near the end of the recorded area which tend to fail with certain hardware. It could be that those errors are somehow scaring the firmware into slower operation (it's ATAPI after all, everything is possible). Jan, is there a way to suppress those probings near the end of the media so that we can determine whether those illegal request failures are relevant to the slower performance? Thanks. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User jack@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c34 Jan Kara <jack@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |REOPENED Info Provider|jack@novell.com | --- Comment #34 from Jan Kara <jack@novell.com> 2009-04-23 14:37:20 MDT --- Tejun, the changes in probe code were in UDF, there are no changes to ISO9660 and a quick check of isofs_fill_super() seems to show that it reads blocks only at the beginning of the media. But the messages could actually come from UDF code if something tries to mount the CD first as UDF and after that as ISO9660. So how is the filesystem mounted? -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User teheo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c35 --- Comment #35 from Tejun Heo <teheo@novell.com> 2009-04-23 16:11:50 MDT --- (In reply to comment #34)
Tejun, the changes in probe code were in UDF, there are no changes to ISO9660 and a quick check of isofs_fill_super() seems to show that it reads blocks only at the beginning of the media.
Oops, right. Sorry about the confusion.
But the messages could actually come from UDF code if something tries to mount the CD first as UDF and after that as ISO9660. So how is the filesystem mounted?
Yeap, something is triggering those errors. Trying w/o udf sounds like a good idea. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User coolo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c36 --- Comment #36 from Stephan Kulow <coolo@novell.com> 2009-04-24 02:21:59 MDT --- there is no udf in boot.msg and mount -t iso9660 doesn't change anything. The boot.msg shows this pattern: <6>usb bla blah <7>ISO 9660 Extensions: <3>end_request: I/O error <7>ISO 9660 Extensions <6>loop: module loaded <6>squashfs: version 4.0 So the IO errors happen in between two ISO mounts. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User coolo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c37 --- Comment #37 from Stephan Kulow <coolo@novell.com> 2009-04-24 02:25:17 MDT --- interesting. #0's boot.msg has the I/O errors before the two ISOs. And I just checked - there is another mount without -t -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User coolo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c38 --- Comment #38 from Stephan Kulow <coolo@novell.com> 2009-04-24 03:25:00 MDT --- having all 4 mounts in the initrd changed: 1.9MB/s - bingo! I'm not 100% sure I can classify this behaviour of mount -t auto as bug though. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User teheo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c39 --- Comment #39 from Tejun Heo <teheo@novell.com> 2009-04-24 04:45:09 MDT --- mount is legitimately accessing the media to determine fs type and as it happens the device can't access near the end of the file (don't know specifically why but it often happens with some devices, padding at the end to nice boundaries sometimes seem to help) which trigger command errors, which are fine too. However, the firmware feels that something is wrong and speed down (I don't think this is a common behavior, it might not even be an intended one). Jan, will this also be fixed by updating udf to not access near the end of the media when avoidable? Thanks. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User coolo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c40 --- Comment #40 from Stephan Kulow <coolo@novell.com> 2009-04-27 02:43:03 MDT --- this drive seems to be really, really easy to scare. The new ISOs do not show the errors anymore, but it's still slow. The ISO I built on friday had full speed, but the difference then was that I also avoided all hwinfo calls. So I need to test again ;( -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User coolo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c41 --- Comment #41 from Stephan Kulow <coolo@novell.com> 2009-04-27 05:56:38 MDT --- If avoiding hwinfo --cdrom (and simply taking /dev/cdrom if it exists), I could trick my drive to behave: --- /tmp/initrd.orig/include 2009-04-27 10:37:10.000000000 +0200 +++ /tmp/initrd/include 2009-04-27 13:24:35.000000000 +0200 @@ -1816,9 +1816,13 @@ for module in sg sd_mod sr_mod cdrom ide-cd BusLogic vfat; do /sbin/modprobe $module done + else + if test -r /dev/cdrom; then + cddev=/dev/cdrom + fi fi Echo -n "Waiting for CD/DVD device(s) to appear..." - while true;do + while test -z "$cddev";do cddevs=`$h --cdrom | grep "Device File:"|sed -e"s@(.*)@@" | cut -f2 -d:` cddevs=`echo $cddevs` for i in $cddevs;do -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User jack@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c42 --- Comment #42 from Jan Kara <jack@novell.com> 2009-04-27 06:06:52 MDT --- Tejun, if this really comes from someone trying to mount the disk as UDF (which I think we have not shown), then I don't think this will get fixed since the CD isn't UDF. So we really have to go through all locations as specified in the standard to find out none of them contains UDF anchor block. And some of these locations are near the end of the disk. When mount is used without filesystem type (or with -t auto), then it uses libvolume_id to identify the filesystem first and then calls kernel to mount the filesystem. Only if libvolume_id is unable to determine filesystem type, it calls kernel to try to mount the fs one filesystem after another - it might be good to check via vol_id utility whether libvolume_id recognizes the medium or not (and whether it causes those IO errors or not). BTW: I've checked and libvolume_id does not load blocks near the end of the filesystem to identify UDF (which makes it unreliable but that's not our problem for now). -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User snwint@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c43 --- Comment #43 from Steffen Winterfeldt <snwint@novell.com> 2009-04-27 06:31:58 MDT --- Hm, if it's really libvolume_id it need not be iso/udf related at all. See bugs 352055, 419195. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User coolo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c44 --- Comment #44 from Stephan Kulow <coolo@novell.com> 2009-04-27 08:01:39 MDT --- vol_id /dev/cdrom triggers the errors in dmesg -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User coolo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c45 --- Comment #45 from Stephan Kulow <coolo@novell.com> 2009-04-27 08:22:17 MDT --- and after that the drive is in slowmo -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User jack@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c46 --- Comment #46 from Jan Kara <jack@novell.com> 2009-04-27 08:29:44 MDT --- OK, so at least that is explained. The strange accesses are caused by libvolume_id. Since this is reported against 11.2 I guess it's quite recent, isn't it? Probably report that to Kay Sievers... -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=493201 User coolo@novell.com added comment http://bugzilla.novell.com/show_bug.cgi?id=493201#c47 Stephan Kulow <coolo@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |kasievers@novell.com --- Comment #47 from Stephan Kulow <coolo@novell.com> 2009-04-27 08:35:58 MDT --- yes, it's recent. I haven't tried any other distribution's livecd that uses such recent udev. But 11.1 and F10 were ok. But I'm not sure if vol_id is wrong in seeking or the kernel in tricking the drive into slowmo. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=493201 https://bugzilla.novell.com/show_bug.cgi?id=493201#c49 Jeff Mahoney <jeffm@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|REOPENED |NEEDINFO CC| |jeffm@novell.com InfoProvider| |coolo@novell.com --- Comment #49 from Jeff Mahoney <jeffm@novell.com> 2010-11-09 19:36:33 UTC --- Does this issue still exist? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=493201 https://bugzilla.novell.com/show_bug.cgi?id=493201#c50 Stephan Kulow <coolo@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |RESOLVED InfoProvider|coolo@novell.com | Resolution| |NORESPONSE --- Comment #50 from Stephan Kulow <coolo@novell.com> 2010-11-10 09:37:42 CET --- I don't have this laptop anymore, so I can't say. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
participants (1)
-
bugzilla_noreply＠novell.com