Re: resend: Re: [Cooker] APM and time reset after suspend
Andrey Borzenkov kirjoitti viestissään (lähetysaika Tiistai 19. Elokuuta 2003 13:58): Juan please disable CONFIG_APM_RTC_IS_GMT. Kernel does the right thing when it is disabled ALWAYS. When it is enabled it just confuses it. Others who produce custom kernels may consider it as well :) Missed this mail before... (added to my kernel) Thomas
resend: Re: [Cooker] APM and time reset after suspend
[due to yesterday's cooker mailing list problems, I'm reposting this] Andrey Borzenkov wrote: you can just as well compile kernel under 9.1. it should have the same problem and the same fix. Hi, Andrey. Sorry about the long wait. For a while, demands at work made it impossible to play with this. And being enough of a novice, I avoid at all costs playing around with my production environment until after I test things (often in cooker). That's why I was waiting for a kernel that compiled under cooker. For others following along: putting my laptop into suspend resulted in the clock changing time. Andrey pointed out that the kernel assumes that the hardware clock is in GMT. If it's not, then there is a drift at wakeup time (in my case, at UTC+2, my clock moves forward by two hours upon every wakeup). Andrey, as you can expect, your theory was correct: it is indeed the GMT setting in kernel: standard 2.4.22-0.6mdk kernel (CONFIG_APM_RTC_IS_GMT=y): [EMAIL PROTECTED] jkeller]$ date Mon Aug 18 17:39:56 CEST 2003 (suspend, wake up) [EMAIL PROTECTED] jkeller]$ date Mon Aug 18 19:40:23 CEST 2003 modified kernel (# CONFIG_APM_RTC_IS_GMT is not set): [EMAIL PROTECTED] jkeller]$ date Mon Aug 18 17:47:05 CEST 2003 (suspend, wake up) [EMAIL PROTECTED] jkeller]$ date Mon Aug 18 17:47:40 CEST 2003 My fix has been to use UTC/GMT for my hardware clock. I am no longer dual-booting Windows on that machine, so this is a safe option. However, this obviously doesn't cover the lion's share of the desktop users who Mandrake targets as customers. These people will need to be able to leave the hardware clock in local time. However, I'm pretty sure that Mandrake won't be able to include a second kernel compiled with this switch off just to accommodate those people. For everyone, please note that I *do* have suspend-scripts installed (this was a common suggestion the last time I brought this up). However, hibernate doesn't enter sleep (it starts the process, then cancels it). I've never tried to fix /etc/suspend.conf because the command is incredibly slow compared to using the keyboard combo (Fn+F4) to sleep. Whatever script should be executed on wakeup, isn't. This is of course why the clock isn't set taking into account UTC= in /etc/sysconfig/clock. I use APM, so ACPI doesn't have a hand in any of these problems. That's my (long-delayed) report. I don't know what trail to follow, but I'm more than happy to try things out. I'll cross my fingers that you have ideas, Andrey. Thank you again for all the time you've devoted to this rather small problem. - John P.S. Andrey, I *think* I'm off your ISP's blacklist now. Knock on wood, this will make it to you.
Re: resend: Re: [Cooker] APM and time reset after suspend
Juan please disable CONFIG_APM_RTC_IS_GMT. Kernel does the right thing when it is disabled ALWAYS. When it is enabled it just confuses it. Others who produce custom kernels may consider it as well :) Andrey, as you can expect, your theory was correct: it is indeed the GMT setting in kernel: standard 2.4.22-0.6mdk kernel (CONFIG_APM_RTC_IS_GMT=y): [EMAIL PROTECTED] jkeller]$ date Mon Aug 18 17:39:56 CEST 2003 (suspend, wake up) [EMAIL PROTECTED] jkeller]$ date Mon Aug 18 19:40:23 CEST 2003 modified kernel (# CONFIG_APM_RTC_IS_GMT is not set): [EMAIL PROTECTED] jkeller]$ date Mon Aug 18 17:47:05 CEST 2003 (suspend, wake up) [EMAIL PROTECTED] jkeller]$ date Mon Aug 18 17:47:40 CEST 2003 background. when APM_RTC_IS_GMT is disabled, kernel computes the difference between CMOS and kernel time immediately before going to sleep and restores kernel time taking difference in account after resuming. It works irrespectively of whether your system is using GMT or local time - for GMT diff will be (close to) zero, that's all. When APM_RTC_IS_GMT is enabled kernel does nothing taking kernel time be equal to CMOS time. This breaks for the case system is using local time producing observed time jumps. I am not sure about hwclock --hctosys (that is done in APM scripts); I have vague recollection that it did break in this case as well but I do not remember anymore why I came to this conclusion. Anyway APM_RTC_IS_GMT is redundant for RTC == GMT, is buggy for RTC == local so it should be killed. [...] P.S. Andrey, I *think* I'm off your ISP's blacklist now. Knock on wood, this will make it to you. I got this just fine. Recently ISPs seem to go mad about blacklisting everything possible. Even more amusing is that whether my ISP is blacklisted or blacklists some address, they recommend contact the other server's admin in both cases :)
Re: [Cooker] APM and time reset after suspend
On Friday 01 August 2003 20:40, John Keller wrote: Andrey Borzenkov wrote: is gcc still broken? does the problem still exist? Sorry, I've been tied up otherwise. I've been intending to try this again, and hope to get to it next week. I don't currently have a cooker install (had to wipe my drive for other reasons). From what I've seen, there may be issues with kernel compiles and clean cooker installs. So while I'd originally intended to get to it last weekend, I put it off yet again in hopes of avoiding hiccups. you can just as well compile kernel under 9.1. it should have the same problem and the same fix. I'll be sure to let you know as soon as I get back in action. :) - John
Re: [Cooker] APM and time reset after suspend
Andrey Borzenkov wrote: is gcc still broken? does the problem still exist? Sorry, I've been tied up otherwise. I've been intending to try this again, and hope to get to it next week. I don't currently have a cooker install (had to wipe my drive for other reasons). From what I've seen, there may be issues with kernel compiles and clean cooker installs. So while I'd originally intended to get to it last weekend, I put it off yet again in hopes of avoiding hiccups. I'll be sure to let you know as soon as I get back in action. :) - John
Re: [Cooker] APM and time reset after suspend
is gcc still broken? does the problem still exist? -Original Message- Andrey Borzenkov wrote: Well, I'm not sure but I think the latest version of gcc has messed things up. I install kernel-source, change the one line in .config to n, this should be commented out (usually). I am not sure what effect n has (it must work as well, but I never tried it). Yeah, I'd tried it commented out, but in my tests I also tried n. The effect seems the same, except that n would explicitly set something off even if the default was y or m. # CONFIG_APM_RTC_IS_GMT is not set (this is from 2.5 but you got an idea) Yup; thanks. It's always good to have confirmation. started off with make oldconfig. wrong. cp .config .config.old make mrproper cp .config.old .config I'd done that in one of my variations, but then thought maybe since it's from an rpm, it's better *not* to do a mrproper. Thanks for the correction. edit Makefile and set EXTRAVERSION to something different to just use make EXTRAVERSION=xxx ... on command line Oops; had missed that. do not do it in /usr/src, you will screw up module compilation for vanilla mdk kernel. Copy sources somwehere in your home. Do not change order. Yeah, I'd seen both do and don't instructions about doing it in /usr/src. My mistake was deciding to go the simple route (i.e. no copies), as outlined in MandrakeUser: http://www.mandrakeuser.org/docs/resource/res_aolm-faq.html#4d Oops. :) I should've listened to the readme, which warns against this very thing. Apparently this time I got it just fine. Anyway, you can keep list informed just the same :) Well, thanks to the list for tolerating what amounts to n00b pollution. I realize that cooker isn't the place for tutorials, and I very much appreciate your help and others' tolerance. I try to research things on my own, but I seem to have reached the limits of what I could do that way. Here's where I'm at now: I reinstalled cooker from scratch last night, since I wanted to be positive that my having done a compile in /usr/src didn't mess things up too much. Besides, it was time for a clean cooker. I followed your steps to the letter. Unfortunately, make stops with an error (make bzImage, by the way, is what is documented in the readme and gives the same error as what follows): - sched.c: In function `schedule': sched.c:714: internal compiler error: in merge_assigned_reloads, at reload1.c:6134 Please submit a full bug report, with preprocessed source if appropriate. See URL:https://qa.mandrakesoft.com/ for instructions. make[2]: *** [sched.o] Error 1 make[1]: *** [first_rule] Error 2 make: *** [_dir_kernel] Error 2 - This is exactly the error that I was getting before my last message. I haven't changed any settings other than the APM/GMT one, and this is a fresh install from scratch. Of course, these messages just indicate where the problem is, not what it is. I'd be happy to do some exploring, but I'm wondering if we're just at a moment in cooker where the kernel can't be compiled? Anyone care to give hope by saying they've successfully compiled in the last week or so? Anyway, there's that. Thanks again for all your time, Andrey. - John
Re: [Cooker] APM and time reset after suspend
Andrey Borzenkov wrote: Well, I'm not sure but I think the latest version of gcc has messed things up. I install kernel-source, change the one line in .config to n, this should be commented out (usually). I am not sure what effect n has (it must work as well, but I never tried it). Yeah, I'd tried it commented out, but in my tests I also tried n. The effect seems the same, except that n would explicitly set something off even if the default was y or m. # CONFIG_APM_RTC_IS_GMT is not set (this is from 2.5 but you got an idea) Yup; thanks. It's always good to have confirmation. started off with make oldconfig. wrong. cp .config .config.old make mrproper cp .config.old .config I'd done that in one of my variations, but then thought maybe since it's from an rpm, it's better *not* to do a mrproper. Thanks for the correction. edit Makefile and set EXTRAVERSION to something different to just use make EXTRAVERSION=xxx ... on command line Oops; had missed that. do not do it in /usr/src, you will screw up module compilation for vanilla mdk kernel. Copy sources somwehere in your home. Do not change order. Yeah, I'd seen both do and don't instructions about doing it in /usr/src. My mistake was deciding to go the simple route (i.e. no copies), as outlined in MandrakeUser: http://www.mandrakeuser.org/docs/resource/res_aolm-faq.html#4d Oops. :) I should've listened to the readme, which warns against this very thing. Apparently this time I got it just fine. Anyway, you can keep list informed just the same :) Well, thanks to the list for tolerating what amounts to n00b pollution. I realize that cooker isn't the place for tutorials, and I very much appreciate your help and others' tolerance. I try to research things on my own, but I seem to have reached the limits of what I could do that way. Here's where I'm at now: I reinstalled cooker from scratch last night, since I wanted to be positive that my having done a compile in /usr/src didn't mess things up too much. Besides, it was time for a clean cooker. I followed your steps to the letter. Unfortunately, make stops with an error (make bzImage, by the way, is what is documented in the readme and gives the same error as what follows): - sched.c: In function `schedule': sched.c:714: internal compiler error: in merge_assigned_reloads, at reload1.c:6134 Please submit a full bug report, with preprocessed source if appropriate. See URL:https://qa.mandrakesoft.com/ for instructions. make[2]: *** [sched.o] Error 1 make[1]: *** [first_rule] Error 2 make: *** [_dir_kernel] Error 2 - This is exactly the error that I was getting before my last message. I haven't changed any settings other than the APM/GMT one, and this is a fresh install from scratch. Of course, these messages just indicate where the problem is, not what it is. I'd be happy to do some exploring, but I'm wondering if we're just at a moment in cooker where the kernel can't be compiled? Anyone care to give hope by saying they've successfully compiled in the last week or so? Anyway, there's that. Thanks again for all your time, Andrey. - John
Re: [Cooker] APM and time reset after suspend
Michael Reinsch wrote: What is in your /etc/sysconfig/clock? UTC should not be set to true. Nope, it's set to false. While poor at kernel compiling, I've got my config files more or less down pat. :) In my opinion, if this is something that's broken at kernel config level, no band-aid fixes with scripts (whether it works or not) is the correct solution. Well, as this stuff can be handled in user space, why handle them in kernel space? The kernel apparently doesn't know enough to handle this correctly. That at least is my opinion. Well, it can handle it. The problem is that the kernel -- at *compile* time -- has things set on or off. That's reflective of server thinking, not what's needed for desktop users (laptop for now, but most systems have some sort of power savings). I can hardly blame them, but it seems that it would be better to have it at a module level. Besides, in essence this is for people who have multiple (mixed) boots. I'm soon going to be Linux-only on my laptop, so I'll simply move to GMT on the hardware clock. But it doesn't really matter for the user, so let's try and get this fixed the easiest way. I definitely agree. I guess I'm not a fan of a script correcting something whose cause could be avoided. On the other hand, I'm beginning to think that it might be the best way available at the moment. Otherwise, the Mandrake installer/DrakConf would have to keep two copies of the kernel on hand for the moment the user changes this setting. This is hardly optimal, and compilation solutions carried to their extreme lead to Gentoo... The main question currently imho is: why does this script not set correct time. Maybe there is a bug in hwclock? Could you maybe try executing hwclock the same way it is executed by the script? Maybe you could also add a line logging the way hwclock is executed by the script. The execution of hwclock is pretty much vanilla. I can manually execute the installed hibernate, but the machine doesn't turn off. But on the good side, the verbose output confirms that the system clock is reset from the hwclock with non-GMT. If I didn't need for my machine to turn off, that would be prefect. :) Using my suspend key on the keyboard is how I go into the mode, and waking up seems to properly execute *something*. I'm beginning to wonder if some detail(s) is/are missing in the messages sent to the system by the BIOS. You wrote some time back, but I wanted to take a moment now that I had one to reply. In the end, I'm beginning to think I'll take the path of least resistance and just solve it by doing what I intended once Linux-only and set the hwclock to UTC. - John
Re: [Cooker] APM and time reset after suspend
On Thursday 12 June 2003 13:55, John Keller wrote: I followed your steps to the letter. Unfortunately, make stops with an error (make bzImage, by the way, is what is documented in the readme and gives the same error as what follows): - sched.c: In function `schedule': sched.c:714: internal compiler error: in merge_assigned_reloads, at reload1.c:6134 Please submit a full bug report, with preprocessed source if appropriate. See URL:https://qa.mandrakesoft.com/ for instructions. make[2]: *** [sched.o] Error 1 make[1]: *** [first_rule] Error 2 make: *** [_dir_kernel] Error 2 - This is exactly the error that I was getting before my last message. Oops that is bad. Sorry, had you quoted this before we could spare reinstallation. Cannot comment on it at all. I Cc to Gwenole, may be he can tell us what's going on. I am not using cooker myself. Returning to original problem - our last chance is to downgrade gcc. Is it possible to install gcc from 9.1 or they already become too incompatible? -andrey I haven't changed any settings other than the APM/GMT one, and this is a fresh install from scratch. Of course, these messages just indicate where the problem is, not what it is. I'd be happy to do some exploring, but I'm wondering if we're just at a moment in cooker where the kernel can't be compiled? Anyone care to give hope by saying they've successfully compiled in the last week or so? Anyway, there's that. Thanks again for all your time, Andrey. - John
Re: [Cooker] APM and time reset after suspend
Andrey Borzenkov wrote: On Thursday 12 June 2003 13:55, John Keller wrote: This is exactly the error that I was getting before my last message. Oops that is bad. Sorry, had you quoted this before we could spare reinstallation. Oh, that's all right. I would've mentioned it if I had really wanted to avoid a reinstall. - John
Re: [Cooker] APM and time reset after suspend
Andrey Borzenkov wrote: Good. If your system is using local time for CMOS - compile your kernel with CONFIG_APM_RTC_IS_GMT off. Please, do not do any other changes against default Mandrake config. If it helps I will do the patch. If your system keeps CMOS as GMT or turning the above off does not help - well, I do not know :)) We'll see. Well, I'm not sure but I think the latest version of gcc has messed things up. I install kernel-source, change the one line in .config to n, and started off with make oldconfig. By the time I hit make bzImage, it fails with an error (I'd specify, but I don't really want to pollute the list with too detail). I can wait a new kernel-source; on the other hand, it's entirely possible that I'm screwing something up. I've tried a few times with some variations, and looked around to see if I wasn't just doing things wrong. As idiot-proof as the compile process is, I don't think they expected me... :) BTW, I tried to cc you (Andrey) before, but it bounced back (your mail server didn't like one of the servers in the send chain. So, I'm posting this to the list as well just in case... - John
Re: [Cooker] APM and time reset after suspend
On Sunday 08 June 2003 21:45, you wrote: Andrey Borzenkov wrote: Good. If your system is using local time for CMOS - compile your kernel with CONFIG_APM_RTC_IS_GMT off. Please, do not do any other changes against default Mandrake config. If it helps I will do the patch. If your system keeps CMOS as GMT or turning the above off does not help - well, I do not know :)) We'll see. Well, I'm not sure but I think the latest version of gcc has messed things up. I install kernel-source, change the one line in .config to n, this should be commented out (usually). I am not sure what effect n has (it must work as well, but I never tried it). CONFIG_APM=y # CONFIG_APM_IGNORE_USER_SUSPEND is not set CONFIG_APM_DO_ENABLE=y CONFIG_APM_CPU_IDLE=y # CONFIG_APM_DISPLAY_BLANK is not set # CONFIG_APM_RTC_IS_GMT is not set # CONFIG_APM_ALLOW_INTS is not set # CONFIG_APM_REAL_MODE_POWER_OFF is not set (this is from 2.5 but you got an idea) and started off with make oldconfig. wrong. cp .config .config.old make mrproper cp .config.old .config edit Makefile and set EXTRAVERSION to something different ot just use make EXTRAVERSION=xxx ... on command line make oldconfig make dep make make modules make modules_install make install do not do it in /usr/src, you will screw up module compilation for vanilla mdk kernel. Copy sources somwehere in your home. Do not change order. By the time I hit make bzImage, it fails with an error (I'd specify, but I don't really want to pollute the list with too detail). sure; I hoped how to compile kernel on mdk be in FAQ somewhere. BTW, I tried to cc you (Andrey) before, but it bounced back (your mail server didn't like one of the servers in the send chain. Apparently this time I got it just fine. Anyway, you can keep list informed just the same :) -andrey
Re: [Cooker] APM and time reset after suspend
Michael Reinsch wrote: Hi! On Tue, 3 Jun 2003 10:51:29 +0200 John Keller [EMAIL PROTECTED] wrote: Michael (Reinsch), if you look at my bug report, you'll see that CLOCK_SYNC set to yes (as per the default install of the rpm) doesn't help. And I keep my config pretty vanilla -- other than the fact that it's running cooker... :) What is in your /etc/sysconfig/clock? UTC should not be set to true. In my opinion, if this is something that's broken at kernel config level, no band-aid fixes with scripts (whether it works or not) is the correct solution. Well, as this stuff can be handled in user space, why handle them in kernel space? The kernel apparently doesn't know enough to handle this correctly. That at least is my opinion. But it doesn't really matter for the user, so let's try and get this fixed the easiest way. The main question currently imho is: why does this script not set correct time. Maybe there is a bug in hwclock? Could you maybe try executing hwclock the same way it is executed by the script? Maybe you could also add a line logging the way hwclock is executed by the script. Hi, I have seen a similar problem on 9.1 where it is configured for BST. Each time the machine is rebooted an extra hour gets added to the system time.I am guessing here but if on shutdown the RTC is syched with system time and the extra hour is not subtracted this problem would occur. I have checked all the timeconfig files I can find and they all appear to be set to the correct values for my timezone. Just my twopennyworth Regards, C.H. Close
Re: [Cooker] APM and time reset after suspend
Hi! On Tue, 3 Jun 2003 10:51:29 +0200 John Keller [EMAIL PROTECTED] wrote: Michael (Reinsch), if you look at my bug report, you'll see that CLOCK_SYNC set to yes (as per the default install of the rpm) doesn't help. And I keep my config pretty vanilla -- other than the fact that it's running cooker... :) What is in your /etc/sysconfig/clock? UTC should not be set to true. In my opinion, if this is something that's broken at kernel config level, no band-aid fixes with scripts (whether it works or not) is the correct solution. Well, as this stuff can be handled in user space, why handle them in kernel space? The kernel apparently doesn't know enough to handle this correctly. That at least is my opinion. But it doesn't really matter for the user, so let's try and get this fixed the easiest way. The main question currently imho is: why does this script not set correct time. Maybe there is a bug in hwclock? Could you maybe try executing hwclock the same way it is executed by the script? Maybe you could also add a line logging the way hwclock is executed by the script. -- Michael Reinsch [EMAIL PROTECTED] http://mr.uue.org pgp0.pgp Description: PGP signature
Re: [Cooker] APM and time reset after suspend
Andrey Borzenkov wrote: While working on unrelated issues I realized that kernel has the config option CONFIG_APM_RTC_IS_GMT that basically tells the system if CMOS time is stored in GMT or local time. Default on Mandrake seems to be GMT. Now the problem is that after returning from suspend and/or standby APM would set the time to CMOS clock erroneusly assuming it is GMT. That would result in exactly timezone shift observed by many people. The problem does exist and there were nunerous complaints at least in a.o.l.m. IMHO the design is braindamadged anyway and this option must be converted to runtime option (leaving config part as default value); this needs support from installer/drakx as well (i.e. it should correctly updated kernel boot parameter). May be it becomes non-issue with ACPI. The patch for APM is trivial enough; of course we need interseted party to test it :) THANK you, Andrey. I reported this some time ago, but the best I ever got was people asking me to confirm my setup. http://qa.mandrakesoft.com/show_bug.cgi?id=3125 I'm glad that someone who understands the kernel was able to discover the source of this problem. Michael (Reinsch), if you look at my bug report, you'll see that CLOCK_SYNC set to yes (as per the default install of the rpm) doesn't help. And I keep my config pretty vanilla -- other than the fact that it's running cooker... :) In my opinion, if this is something that's broken at kernel config level, no band-aid fixes with scripts (whether it works or not) is the correct solution. Patching the source of any problem seems more logical. Granted, I don't know the reason (if one) for the setting always being set to GMT, but considering the cited reports a fix would make several users happy. Oh, and I'd be more than happy to test any patches. - John
Re: [Cooker] APM and time reset after suspend
[Please, Cc to me on replies. I am not receiving cooker though I read archives. Using occasion to thank all oldtimers who are patient enough to do it :)] I reported this some time ago, but the best I ever got was people asking me to confirm my setup. http://qa.mandrakesoft.com/show_bug.cgi?id=3125 Yes looks very promising :) I'm glad that someone who understands the kernel was able to discover the source of this problem. Oh, you mean to say Chmouel does not understand kernel?!!! :)) Oh, and I'd be more than happy to test any patches. Good. If your system is using local time for CMOS - compile your kernel with CONFIG_APM_RTC_IS_GMT off. Please, do not do any other changes against default Mandrake config. If it helps I will do the patch. If your system keeps CMOS as GMT or turning the above off does not help - well, I do not know :)) We'll see. -andrey
Re: [Cooker] APM and time reset after suspend
Andrey Borzenkov wrote: [Please, Cc to me on replies. I am not receiving cooker though I read archives. Using occasion to thank all oldtimers who are patient enough to do it :)] Gotcha. In my excitement, I'd forgotten that you weren't on cooker. Oh, you mean to say Chmouel does not understand kernel?!!! :)) Not at all! I just meant that from among those who understand it, someone was able to find the cause of my problem. Also, it was a poke at me: put me in front of the kernel (code, config, you name it) and come back in 10 years. I'll still have a blank look on my face. Oh, and I'd be more than happy to test any patches. Good. If your system is using local time for CMOS - compile your kernel with CONFIG_APM_RTC_IS_GMT off. Please, do not do any other changes against default Mandrake config. If it helps I will do the patch. If your system keeps CMOS as GMT or turning the above off does not help - well, I do not know :)) We'll see. Sounds good. I on purpose configured my laptop as if I were a Windows user migrating over to Linux (besides, in the sense of using it for my desktop platform, I really am :). So, my CMOS + system are set to local time. I'll recompile sometime soon and report the results. Thanks for the tip! - John
Re: [Cooker] APM and time reset after suspend
Hi! On Sun, 1 Jun 2003 13:26:09 +0400 Andrey Borzenkov [EMAIL PROTECTED] wrote: IMHO the design is braindamadged anyway and this option must be converted to runtime option (leaving config part as default value); this needs support from installer/drakx as well (i.e. it should correctly updated kernel boot parameter). Have a look at /etc/sysconfig/suspend-scripts/suspend.d/clock which is run before and after suspend. If you set CLOCK_SYNC to yes (which is the default) in your /etc/sysconfig/suspend, the clock is set from the hardware clock upon resume, using the parameters from /etc/sysconfig/clock (UTC=true/false). So, if you observe a time drift, it seems to me that this script is either not executed or something is wrong with hwclock or your settings. -- Michael Reinsch [EMAIL PROTECTED] http://mr.uue.org pgp0.pgp Description: PGP signature
