From: Gerrit H. <ge...@us...> - 2001-10-25 17:23:54
|
The patch was relatively straight forward. There was a mod to sched.c in the IA64 patch that needed to be backed out to allow the MQ scheduler patch to be applied cleanly. I don't think there were any arch specific changes needed for IA64, although I don't remember for sure. If the port isn't obvious, let me know and I'll give it a try here. gerrit > On Wed, Oct 24, 2001 at 06:44:33PM +0200, Erich Focht wrote: > > > > did anybody addapt the MQ scheduler to IA64-Linux? I'd like to test its > > behavior on the NEC AzusA and am curious whether anybody did/does work on > > the IA64 version. > > > > I believe Gerrit Huizenga has ported the patch to an IA64 kernel > in the past. He may be able to provide some insight. The code > is mostly (if not entirely) architecture independent. > > -- > Mike > > _______________________________________________ > Lse-tech mailing list > Lse...@li... > https://lists.sourceforge.net/lists/listinfo/lse-tech > > |
From: Erich F. <fo...@es...> - 2001-10-25 17:53:28
|
Thanks for the hints, Natarjan is right, for IA64 the code should run out of the box. I had to replace "-1ULL" by "(cycles_t) -1" in kernel/sched.c, then the patch applied correctly. My first try with the fresh 2.4.13 kernel failed, it hangs in wait_init_idle... I'm just giving it another try with 2.4.12. Thanks for the help, Regards, Erich On Wed, 24 Oct 2001, Gerrit Huizenga wrote: > > The patch was relatively straight forward. There was a mod to > sched.c in the IA64 patch that needed to be backed out to allow > the MQ scheduler patch to be applied cleanly. I don't think there > were any arch specific changes needed for IA64, although I don't > remember for sure. If the port isn't obvious, let me know and I'll > give it a try here. > > gerrit > > > On Wed, Oct 24, 2001 at 06:44:33PM +0200, Erich Focht wrote: > > > > > > did anybody addapt the MQ scheduler to IA64-Linux? I'd like to test its > > > behavior on the NEC AzusA and am curious whether anybody did/does work on > > > the IA64 version. > > > > > > > I believe Gerrit Huizenga has ported the patch to an IA64 kernel > > in the past. He may be able to provide some insight. The code > > is mostly (if not entirely) architecture independent. > > > > -- > > Mike |
From: Erich F. <ef...@es...> - 2001-10-25 18:00:21
|
> My first try with the fresh 2.4.13 kernel failed, it hangs in > wait_init_idle... I'm just giving it another try with 2.4.12. Well, with 2.4.12 the kernel panics with "Attempted to kill idle task!" :-( I'll look into this tomorrow, but maybe someone has an idea? Thanks, Erich |
From: Martin J. B. <Mar...@us...> - 2001-10-25 18:40:06
|
>> My first try with the fresh 2.4.13 kernel failed, it hangs in >> wait_init_idle... I'm just giving it another try with 2.4.12. > > Well, with 2.4.12 the kernel panics with "Attempted to kill idle > task!" :-( I'll look into this tomorrow, but maybe someone has an idea? The wait_init_idle loop is there to protect you from a race condition between reschedule_idle and init_idle (the boot CPU can call reschedule_idle before all secondaries have called init_idle, which is bad). You might want to take a look at whether the MQ scheduler patch affects this area of the boot sequence. I'm assuming virgin 2.4.13 boots fine for you? M. |
From: Hubertus F. <fr...@wa...> - 2001-10-25 18:41:31
|
* Erich Focht <ef...@es...> [20011025 14;11]:" > > My first try with the fresh 2.4.13 kernel failed, it hangs in > > wait_init_idle... I'm just giving it another try with 2.4.12. > > Well, with 2.4.12 the kernel panics with "Attempted to kill idle > task!" :-( I'll look into this tomorrow, but maybe someone has an idea? > > Thanks, > Erich > Erich, ofcourse the patch will apply cleanly !!!!!!!!!!!!! Since we modified the 2 files I indicated in the plaform specific part you MUST recreate the similar changes for the IA64 platform as well. Otherwise you will end up with the bugs you get. These changes are literally minimal. Simply grep for i386 in the patch file and you see which file and which locations need modifications. If it runs, please feed the results back to Mike or me. -- Hubertus > > > _______________________________________________ > Lse-tech mailing list > Lse...@li... > https://lists.sourceforge.net/lists/listinfo/lse-tech |
From: Erich F. <fo...@es...> - 2001-10-26 15:12:44
Attachments:
2.4.13.mq-sched-ia64
|
Hello Hubertus, I took the comments in smpboot.c seriously and now it works :-) Only arch/ia64/kernel/smpboot.c had to be patched. Please find appended a patch for the 2.4.13 kernel + linux-2.4.13-ia64-011024.diff. There is also a cosmetic change in the kernel/sched.c patch which makes it apply to the IA64 kernel. Thanks for the help! Regards, Erich On Thu, 25 Oct 2001, Hubertus Franke wrote: > * Erich Focht <ef...@es...> [20011025 14;11]:" > > > My first try with the fresh 2.4.13 kernel failed, it hangs in > > > wait_init_idle... I'm just giving it another try with 2.4.12. > > > > Well, with 2.4.12 the kernel panics with "Attempted to kill idle > > task!" :-( I'll look into this tomorrow, but maybe someone has an idea? > > > > Thanks, > > Erich > > > > Erich, ofcourse the patch will apply cleanly !!!!!!!!!!!!! > > Since we modified the 2 files I indicated in the plaform specific part > you MUST recreate the similar changes for the IA64 platform as well. > Otherwise you will end up with the bugs you get. > These changes are literally minimal. > Simply grep for i386 in the patch file and you see which file and which > locations need modifications. If it runs, please feed the results > back to Mike or me. > > -- Hubertus |