Re: Linux 2.2.18pre25
>Well, I've found that VM-global patch before, of course. Until now, the > last version was against pre18. Since I do not know the exact rules for > including new things into Alan's tree, I thought that VM-global patch was > already included in pre24. Sorry for my lack of experience. ;-)) I should > have checked it. >As I wrote before, I had no time recently to follow the mailing list > carefully and I didn't know exactly what VM-global patch is. > > # >It seems we need to return back to 2.2.13 for some time. :-( > # Definitely no, you only need to apply the above collection of bugfixes. > >Ok, I can try it, at least. >I will let you know about results. VM-global is currently on my 2.2.19pre pile of stuff. Im monitoring a few cases with interest before I commit to that decision however - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Re: Linux 2.2.18pre25
> "I'm sure" meaning "I didn't test it" ? absolutely, I believed that the driver was *exactly* the same as the previous release which didn't boot and needed the fix, but another fix has been applied and corrected it. Now I think it will work with a clean 2.2.18pre25. Anyway, I left a kernel compile behind me this evening, so I'll confirm this on monday as soon as I can reboot the server on a pre25. Cheers, Willy ___ Do You Yahoo!? -- Pour dialoguer en direct avec vos amis, Yahoo! Messenger : http://fr.messenger.yahoo.com - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Re: Linux 2.2.18pre25
> > Bad day, Alan? ;) > Umm no but having people _keep_ sending you do > nothing patches gets annoying after a while ;) Please accept all my apologies, Alan. When I quickly sent you the last patch, I didn't notice that some other broken code had been removed, what I discovered later back home and after comparing 2.2.18pre2[15] (what Miquel noticed too). Next time, I'll spend a little more of my time on carefully reading the patch before resending an old useless one. Cheers, Willy ___ Do You Yahoo!? -- Pour dialoguer en direct avec vos amis, Yahoo! Messenger : http://fr.messenger.yahoo.com - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Re: Linux 2.2.18pre25
On Fri, 8 Dec 2000, Andrea Arcangeli wrote: # On Fri, Dec 08, 2000 at 06:02:57PM +0100, Martin Kacer wrote: # >Is there any chance to get rid of these VMM failures? # You should apply this patch on top of 2.2.18pre25: # ftp://.../VM-global-2.2.18pre25-7.bz2 Well, I've found that VM-global patch before, of course. Until now, the last version was against pre18. Since I do not know the exact rules for including new things into Alan's tree, I thought that VM-global patch was already included in pre24. Sorry for my lack of experience. ;-)) I should have checked it. As I wrote before, I had no time recently to follow the mailing list carefully and I didn't know exactly what VM-global patch is. # >It seems we need to return back to 2.2.13 for some time. :-( # Definitely no, you only need to apply the above collection of bugfixes. Ok, I can try it, at least. I will let you know about results. Martin. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Re: Linux 2.2.18pre25
On Fri, Dec 08, 2000 at 10:47:46AM +0100, Willy Tarreau wrote: > |Bus 0, device 2, function 1: > | Unknown class: Intel OEM MegaRAID Controller (rev 5). > |Medium devsel. Fast back-to-back capable. BIST capable. IRQ 10. Master > Capable. Latency=64. > |Prefetchable 32 bit memory at 0xf000 [0xf008]. > > as you see, the board is found at 0xf008, but used aligned to 0xf000. No. It's found at 0xf000, and has 8 bytes of MMIO space. > my server currently works with that patch, but I'm sure it won't boot anymore > if I apply this 2.2.18pre25 alone. "I'm sure" meaning "I didn't test it" ? - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Re: Linux 2.2.18pre25
On Fri, Dec 08, 2000 at 06:02:57PM +0100, Martin Kacer wrote: >Is there any chance to get rid of these VMM failures? You should apply this patch on top of 2.2.18pre25: ftp://ftp.us.kernel.org/pub/linux/kernel/people/andrea/patches/v2.2/2.2.18pre25/VM-global-2.2.18pre25-7.bz2 >It seems we need to return back to 2.2.13 for some time. :-( Definitely no, you only need to apply the above collection of bugfixes. Andrea - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Re: Linux 2.2.18pre25
On Fri, 8 Dec 2000, Alan Cox wrote: # >Is there any chance to get rid of these VMM failures? # By finding them. :-) I am not so familiar with MM in Linux. :^( And do not have enough time for intensive study... Although I would probably like that work... # Are you confident you are not running out of memory. Well, almost sure. This is the log with load records: (according to /proc/meminfo) FTPusers SMBusr load free memfree swap Fri Dec 8 14:35:05 CET 2000 61 35 6.173068 kB 128932 kB Fri Dec 8 14:40:04 CET 2000 59 36 5.052280 kB 130320 kB Fri Dec 8 14:45:03 CET 2000 59 36 5.972896 kB 131448 kB Fri Dec 8 14:50:03 CET 2000 59 35 6.592908 kB 133140 kB Fri Dec 8 14:55:04 CET 2000 53 36 8.822380 kB 133952 kB Fri Dec 8 15:00:03 CET 2000 53 40 6.422728 kB 135064 kB Fri Dec 8 15:05:03 CET 2000 48 39 5.472264 kB 135684 kB Fri Dec 8 15:10:03 CET 2000 48 41 3.903204 kB 135928 kB Fri Dec 8 15:15:03 CET 2000 51 41 5.932628 kB 135700 kB Fri Dec 8 15:20:03 CET 2000 50 45 6.502124 kB 135828 kB Fri Dec 8 15:25:03 CET 2000 56 44 7.922192 kB 136080 kB Fri Dec 8 15:30:03 CET 2000 49 45 10.892072 kB 136176 kB Fri Dec 8 15:35:03 CET 2000 51 42 6.322960 kB 136156 kB Fri Dec 8 15:40:04 CET 2000 54 44 6.922364 kB 136220 kB Fri Dec 8 15:45:03 CET 2000 54 44 6.632852 kB 136348 kB Fri Dec 8 15:50:04 CET 2000 53 46 3.632248 kB 136420 kB Fri Dec 8 15:55:03 CET 2000 59 48 6.513060 kB 136312 kB (crashed during the next 5 minutes) Doesn't seem to have consumed all of swap space. I will try to determine more info the next time - I promise... # Presumably since 2.2.13 works you are 8) I didn't tell it worked. It had worked a long time ago. It is still not tested now. Unfortunately, due to the absence of raid0 module the bootup process destroyed our 140GB partition. It will take some time to make the system running again. :-( Thank for your answer anyway... Martin. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Re: Linux 2.2.18pre25
>We aplied 2.2.18pre25 patch yesterday hoping it could solve it. The > only difference is that the server reached several hours uptime instead of > 40 minutes (with pre24). After two hours of load between 6.00 and 15.00 > the console was flooded with those unpopular messages ("VM: ..."). The > system was taken down by generation of these messages so quickly, that > even none of the messages appeared in syslog! No response to Ctrl-Alt-Del, > of course... :-( Just trashing... > >Our bug can generate them. :-( Maybe it's a different one? ;-) Quite possibly. >Is there any chance to get rid of these VMM failures? By finding them. Are you confident you are not running out of memory. Presumably since 2.2.13 works you are 8) - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Re: Linux 2.2.18pre25
> > Some days I don't know why I bother > Bad day, Alan? ;) Umm no but having people _keep_ sending you do nothing patches gets annoying after a while ;) > reading the patch, it makes sense. It probably does about the same > as Willy's patch, but the "right" way by using pci_resource_start() > which the one in pre18 only did for kernels > 2.3.0 I suspect what actually happened is that someone fixed pci_resource_start() looking over the change set, and that fixed the megaraid driver - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Re: Linux 2.2.18pre25
> as soon as I can reboot it, I promise I will test the > kernel with and without the patch to be really sure. > but before that, if people who have problems with > megaraid/netraid could give it a try, that would be > cool. Also, it would be nice if people for which the > normal megaraid driver works would accept to check > this > doesn't break anything. Your patch changes the mask on both IO and memory ports to be MEM mask, which is obviously incorrect. It wont actually bite you because all the masking has already been done by pci_resource_start() so you are masking already zero bits. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Re: Linux 2.2.18pre25
On Thu, 7 Dec 2000, Alan Cox wrote: # Ok we believe the VM crash looping printing error messages is now fixed. # Marcelo finally figured it out and my 8Mb 486 has been running 2.2.18pre # with that fix and stably[1]. Unfortunately, I don't think it is fixed. We maintain a heavy loaded FTP/Samba server here (120+ active connections with very long data transfers in rush hours) and it had the "VM: do_try_to_free_pages failed" problem since 2.2.17 was first installed (there was FreeBSD before that). We aplied 2.2.18pre25 patch yesterday hoping it could solve it. The only difference is that the server reached several hours uptime instead of 40 minutes (with pre24). After two hours of load between 6.00 and 15.00 the console was flooded with those unpopular messages ("VM: ..."). The system was taken down by generation of these messages so quickly, that even none of the messages appeared in syslog! No response to Ctrl-Alt-Del, of course... :-( Just trashing... On Fri, 8 Dec 2000, Andrea Arcangeli wrote: # > Ok we believe the VM crash looping printing error messages is now fixed. # Such bug can't generate crashes. Did you ever reproduced crashes on your 8Mb # 486 with 2.2.18pre24? Our bug can generate them. :-( Maybe it's a different one? ;-) Is there any chance to get rid of these VMM failures? Sorry if I've missed something important recently mentioned here. I had not enough time to follow the lk list carefully. Is there any reliable solution? It seems we need to return back to 2.2.13 for some time. :-( Martin. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Re: Linux 2.2.18pre25
According to Alan Cox: > > my server currently works with that patch, but I'm sure it won't boot anymore > > if I apply this 2.2.18pre25 alone. > > Some days I don't know why I bother Bad day, Alan? ;) > > just in case, here it is again. > It doesnt even apply Hmm, it did apply for me. Do newer versions of patch have the -l option on by default? Anyway. I just threw together a testmachine with a megaraid card. With 2.2.18pre18, it doesn't boot. With 2.2.18pre18 + Willy's patch, it does boot. And with 2.2.18pre25 without any extra patches, it magically works. So I took the plunge and compiled 2.2.18pre25 on the production machine with the megaraid. And well, it's coming up as I write this. I see that another patch _has_ been applied between pre18 and pre25 that tooks out some forward/backwards-compat logic with LINUX_VERSION_CODE magic (beneath /* Read the base port and IRQ from PCI */). And reading the patch, it makes sense. It probably does about the same as Willy's patch, but the "right" way by using pci_resource_start() which the one in pre18 only did for kernels > 2.3.0 So, it looks like pre25 has a working megaraid driver. Thanks Alan. Mike. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Re: Linux 2.2.18pre25
> It doesnt even apply sorry Alan, I think it's because I had to copy/paste it with my mouse under X into my browser (I don't have smtp access here at work), and it applies here with a -12 lines offset... Here it is attached for 2.2.18pre25, but since the raid server is running now (under 2.2.18pre20+patch), I won't be able to test it till next week, but I'm a bit confident since it will do the same as the one which currently allows this server to boot. as soon as I can reboot it, I promise I will test the kernel with and without the patch to be really sure. but before that, if people who have problems with megaraid/netraid could give it a try, that would be cool. Also, it would be nice if people for which the normal megaraid driver works would accept to check this doesn't break anything. Regards, Willy ___ Do You Yahoo!? -- Pour dialoguer en direct avec vos amis, Yahoo! Messenger : http://fr.messenger.yahoo.com patch-megaraid-fix
Re: Linux 2.2.18pre25
> my server currently works with that patch, but I'm sure it won't boot anymore > if I apply this 2.2.18pre25 alone. Some days I don't know why I bother > just in case, here it is again. It doesnt even apply > - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Re: Linux 2.2.18pre25
> I asked people to explain why it was needed. I am still waiting. It is a > patch that does nothing. I will not put random deep magic into the > kernel. Alan, I replied to you a few weeks ago (pre20 times) when you asked me why I was sending you this patch. (perhaps you didn't receive my email). What I observed was that my netraid card had a 0x8 base address and the patch aligned that address to 16 bytes : |Bus 0, device 2, function 1: | Unknown class: Intel OEM MegaRAID Controller (rev 5). |Medium devsel. Fast back-to-back capable. BIST capable. IRQ 10. Master Capable. Latency=64. |Prefetchable 32 bit memory at 0xf000 [0xf008]. as you see, the board is found at 0xf008, but used aligned to 0xf000. my server currently works with that patch, but I'm sure it won't boot anymore if I apply this 2.2.18pre25 alone. just in case, here it is again. Cheers, Willy --- 18pre/drivers/scsi/megaraid.c Wed Nov 8 16:02:45 2000 +++ 18pre+megaraid/drivers/scsi/megaraid.c Fri Nov 10 12:03:05 2000 @@ -1920,10 +1920,14 @@ pciIdx++; -if (flag & BOARD_QUARTZ) +if (flag & BOARD_QUARTZ) { + megaBase &= PCI_BASE_ADDRESS_IO_MASK; megaBase = (long) ioremap (megaBase, 128); -else +} +else { + megaBase &= PCI_BASE_ADDRESS_MEM_MASK; megaBase += 0x10; +} /* Initialize SCSI Host structure */ host = scsi_register (pHostTmpl, sizeof (mega_host_config)); - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Re: Linux 2.2.18pre25
In article <[EMAIL PROTECTED]>, Alan Cox <[EMAIL PROTECTED]> wrote: > >Excellent. I've been trying to avoid VM fixes for 2.2.18 to stop stuff getting >muddled together and hard to debug. Running with page aging convinces me that >2.2.19 we need to sort some of the vm issues out badly, and make it faster than >2.4test 8) Ahh.. The challenge is out! You and me. Mano a mano. Linus - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Re: Linux 2.2.18pre25
> (note: the above is outdated so it's not anymore suggested for inclusion of > course) > > I sumbitted most of the not-feature-oriented stuff at pre2 time and I plan to > re-submit after 2.2.18 is released. Excellent. I've been trying to avoid VM fixes for 2.2.18 to stop stuff getting muddled together and hard to debug. Running with page aging convinces me that 2.2.19 we need to sort some of the vm issues out badly, and make it faster than 2.4test 8) - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Re: Linux 2.2.18pre25
On Fri, Dec 08, 2000 at 12:27:58AM +, Alan Cox wrote: > The problem is its hard to know which of your patches depend on what, and > the complete set is large to say the least. That's why I use a `proposed' directory that only contains patches that can be applied to your tree, in this case it was: ftp://ftp.us.kernel.org/pub/linux/kernel/people/andrea/proposed/v2.2/2.2.18pre2/VM-global-2.2.18pre2-6.bz2 (note: the above is outdated so it's not anymore suggested for inclusion of course) I sumbitted most of the not-feature-oriented stuff at pre2 time and I plan to re-submit after 2.2.18 is released. Andrea - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Re: Linux 2.2.18pre25
> Such bug can't generate crashes. Did you ever reproduced crashes on your 8Mb > 486 with 2.2.18pre24? Yes. Every 20 minutes or so quite reliably. With that change it has yet to crash (its actually running that + page aging + another minor tweak so it doesnt return success on page aging until we have a clump of free pages. With just the page aging patch it performed way better but still hung. > >ftp://ftp.us.kernel.org/pub/linux/kernel/people/andrea/kernels/v2.2/2.2.18pre24aa1/00_account-failed-buffer-tries-1 > Oh well ;) > account-failed-buffer-tries-1 is included in VM-global-7 and it was > described in the 2.2.18pre21aa2 email to l-k (CC'ed you) in date Fri, 17 Nov > 2000 18:54:43 +0100: The problem is its hard to know which of your patches depend on what, and the complete set is large to say the least. Alan - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Re: Linux 2.2.18pre25
On Thu, Dec 07, 2000 at 08:03:00PM +, Alan Cox wrote: > > Ok we believe the VM crash looping printing error messages is now fixed. Such bug can't generate crashes. Did you ever reproduced crashes on your 8Mb 486 with 2.2.18pre24? > Marcelo finally figured it out and my 8Mb 486 has been running 2.2.18pre > with that fix and stably[1]. diff -urN 2.2.18pre24/mm/filemap.c 2.2.18pre25/mm/filemap.c --- 2.2.18pre24/mm/filemap.cWed Nov 29 19:28:29 2000 +++ 2.2.18pre25/mm/filemap.cFri Dec 8 00:41:45 2000 @@ -220,8 +220,10 @@ * throttling. */ - if (!try_to_free_buffers(page, wait)) + if (!try_to_free_buffers(page, wait)) { + if(--count < 0) break; goto refresh_clock; + } return 1; } ftp://ftp.us.kernel.org/pub/linux/kernel/people/andrea/kernels/v2.2/2.2.18pre24aa1/00_account-failed-buffer-tries-1 --- 2.2.17pre19/mm/filemap.cTue Aug 22 14:54:13 2000 +++ /tmp/filemap.c Thu Aug 24 01:05:50 2000 @@ -179,6 +179,8 @@ if ((gfp_mask & __GFP_DMA) && !PageDMA(page)) continue; + count--; + /* * Is it a page swap page? If so, we want to * drop it if it is no longer used, even if it @@ -224,7 +226,7 @@ return 1; } - } while (--count > 0); + } while (count > 0); return 0; } lftp> pwd ftp://ftp.kernel.org/pub/linux/kernel/people/andrea/patches/v2.2/2.2.17pre19 ^^^ lftp> ls -l account-failed-buffer-tries-1 -rw-r--r-- 1 korg korg 407 Sep 5 22:43 account-failed-buffer-tries-1 ^^ lftp> Only difference is that pre25 keeps decreasing `count' for locked, mapped and out-of-zone pages and that means it will still fail to shrink the cache when it looks at the unlucky part of the physical memory while the account-failed-buffer-tries-1 intentionally doesn't decrease `count' in that cases to avoid failing in such unlucky cases. account-failed-buffer-tries-1 is included in VM-global-7 and it was described in the 2.2.18pre21aa2 email to l-k (CC'ed you) in date Fri, 17 Nov 2000 18:54:43 +0100: [..] 00_account-failed-buffer-tries-1 Account also the failed buffer tries during shrink_mmap. (me) (this is included in the VM-global that I maintain against vanilla 2.2.x btw) [..] Andrea - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Re: Linux 2.2.18pre25
> Megaraid still needs fixing. I sent you the patch twice, so have > other people, but it still isn't fixed. The I asked people to explain why it was needed. I am still waiting. It is a patch that does nothing. I will not put random deep magic into the kernel. I have no reason to believe the current driver in 2.2.18pre24 does not work, have you tried that specific kernel ? - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Re: Linux 2.2.18pre25
In article <[EMAIL PROTECTED]>, Alan Cox <[EMAIL PROTECTED]> wrote: >So I figure this is it for 2.2.18, subject to evidence to the contrary Megaraid still needs fixing. I sent you the patch twice, so have other people, but it still isn't fixed. The megaBase &= PCI_BASE_ADDRESS_MEM_MASK; ... megaBase &= PCI_BASE_ADDRESS_IO_MASK; is removed by the 2.2.18 version (read the patch) and that breaks older megaraid cards. Existing megaraid system with 2.2.x kernels WILL break with 2.2.18 Mike. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/