From: Domenico A. <ca...@li...> - 2004-01-09 16:24:10
|
hello, i have the following error during the build of the uml kernel (linux 2.6.1 with 00-Combo-v8.bz2 applied). CC fs/proc/task_mmu.o fs/proc/task_mmu.c:86: error: initializer element is not constant fs/proc/task_mmu.c:86: error: (near initialization for `gate_vmarea.vm_start') fs/proc/task_mmu.c:88: error: initializer element is not constant fs/proc/task_mmu.c:88: error: (near initialization for `gate_vmarea.vm_end') make[2]: *** [fs/proc/task_mmu.o] Error 1 i had also needed to modify PTRACE_POKEUSR and PTRACE_POKEUSR in PTRACE_POKEUSER and PTRACE_POKEUSER in file arch/um/sys-i386/ptrace_user.c. cheers domenico -----[ Domenico Andreoli, aka cavok --[ http://filibusta.crema.unimi.it/~cavok/gpgkey.asc ---[ 3A0F 2F80 F79C 678A 8936 4FEE 0677 9033 A20E BC50 |
From: BlaisorBlade <bla...@ya...> - 2004-01-09 18:06:55
|
Alle 17:24, venerd=EC 9 gennaio 2004, Domenico Andreoli ha scritto: > hello, > > i have the following error during the build of the uml kernel (linux 2.6.1 > with 00-Combo-v8.bz2 applied). > > CC fs/proc/task_mmu.o > fs/proc/task_mmu.c:86: error: initializer element is not constant > fs/proc/task_mmu.c:86: error: (near initialization for > `gate_vmarea.vm_start') fs/proc/task_mmu.c:88: error: initializer element > is not constant > fs/proc/task_mmu.c:88: error: (near initialization for > `gate_vmarea.vm_end') make[2]: *** [fs/proc/task_mmu.o] Error 1 > i had also needed to modify PTRACE_POKEUSR and PTRACE_POKEUSR in > PTRACE_POKEUSER and PTRACE_POKEUSER in file arch/um/sys-i386/ptrace_user.= c. > > cheers > domenico =46irst, thanks for the reports. =46or the first bug (which depends on a 2.6.1 change) see this message: http://marc.theaimsgroup.com/?l=3Duser-mode-linux-devel&m=3D107365442315003= &w=3D2 Note I've not tested it. However, it seems quite non-intrusive and=20 straightforward. Watch that LKML thread to see if the author posts other=20 versions of the patch. Note (only if you know a bit of kernel programming): however, how the autho= r=20 notes, he should better call build_gate_map with __initcall. To see how, lo= ok=20 at the changes to mm/memory.c included in the Combo-v8. The second one happens depending on the host headers, I think; I'll add tha= t=20 change to the next version, when and if I release one (depending on my time= ). Bye =2D-=20 cat <<EOSIGN Paolo Giarrusso, aka Blaisorblade Linux Kernel 2.4.23/2.6.0 on an i686; Linux registered user n. 292729 EOSIGN |
From: Domenico A. <ca...@li...> - 2004-01-09 18:18:08
|
BlaisorBlade wrote: > Alle 17:24, venerd=EC 9 gennaio 2004, Domenico Andreoli ha scritto: ... >>i had also needed to modify PTRACE_POKEUSR and PTRACE_POKEUSR in >>PTRACE_POKEUSER and PTRACE_POKEUSER in file arch/um/sys-i386/ptrace_use= r.c. >> hmm... two of these were really PTRACE_PEEKUSR and PTRACE_PEEKUSER... > The second one happens depending on the host headers, I think; I'll add= that=20 > change to the next version, when and if I release one (depending on my = time). >=20 something like this (ore the reverse) should suffice: #ifndef PTRACE_POKEUSR # define PTRACE_POKEUSR PTRACE_POKEUSER #endif #ifndef PTRACE_PEEKUSR # define PTRACE_PEEKUSR PTRACE_PEEKUSER #endif regards domenico -----[ Domenico Andreoli, aka cavok --[ http://filibusta.crema.unimi.it/~cavok/gpgkey.asc ---[ 3A0F 2F80 F79C 678A 8936 4FEE 0677 9033 A20E BC50 |
From: BlaisorBlade <bla...@ya...> - 2004-01-09 19:39:03
|
Alle 19:17, venerd=EC 9 gennaio 2004, Domenico Andreoli ha scritto: > BlaisorBlade wrote: > > Alle 17:24, venerd=EC 9 gennaio 2004, Domenico Andreoli ha scritto: > >>i had also needed to modify PTRACE_POKEUSR and PTRACE_POKEUSR in > >>PTRACE_POKEUSER and PTRACE_POKEUSER in file > >> arch/um/sys-i386/ptrace_user.c. > > hmm... two of these were really PTRACE_PEEKUSR and PTRACE_PEEKUSER... > > > The second one happens depending on the host headers, I think; I'll add > > that change to the next version, when and if I release one (depending on > > my time). > > something like this (ore the reverse) should suffice: > > #ifndef PTRACE_POKEUSR > # define PTRACE_POKEUSR PTRACE_POKEUSER > #endif > #ifndef PTRACE_PEEKUSR > # define PTRACE_PEEKUSR PTRACE_PEEKUSER > #endif > > regards > domenico I'll check what to do, since 2.4.23-1um(or a near release) fixed this in th= e=20 simpler way, i.e. just by changing USR into USER, IIRC. For the first=20 patch(the one I pointed to), I've just seen that it is included into=20 2.6.1-mm1. Bye =2D-=20 cat <<EOSIGN Paolo Giarrusso, aka Blaisorblade Linux Kernel 2.4.23/2.6.0 on an i686; Linux registered user n. 292729 EOSIGN |
From: Ingo M. <mi...@el...> - 2004-01-11 12:53:54
|
* BlaisorBlade <bla...@ya...> wrote: > For the first bug (which depends on a 2.6.1 change) see this message: > http://marc.theaimsgroup.com/?l=user-mode-linux-devel&m=107365442315003&w=2 > > Note I've not tested it. However, it seems quite non-intrusive and > straightforward. Watch that LKML thread to see if the author posts > other versions of the patch. FYI, 2.6.1 broke the UML arch somewhere. Applying the v8 patch (and fixing up the trivial breakages) produces an UML kernel that boots and gets to a shell prompt, but the first 'ls' either causes an infinite loop in skas mode, or a segfault (and kernel hang) in tt mode. (note that this happens with my combo patch variant too, so it's a 2.6.1-induced breakage.) host-skas3 works fine in 2.6.1, so i could use an older UML binary on 2.6.1+v8. Ingo |
From: Domenico A. <ca...@li...> - 2004-01-11 15:37:43
|
Ingo Molnar wrote: > FYI, 2.6.1 broke the UML arch somewhere. Applying the v8 patch (and > fixing up the trivial breakages) produces an UML kernel that boots and > gets to a shell prompt, but the first 'ls' either causes an infinite > loop in skas mode, or a segfault (and kernel hang) in tt mode. > > (note that this happens with my combo patch variant too, so it's a > 2.6.1-induced breakage.) > i have this same behavior also with uml 2.6.0 + v8 on host 2.6.1 + v8. > host-skas3 works fine in 2.6.1, so i could use an older UML binary on > 2.6.1+v8. > if i use a 2.4.23-1um on the same host i get the bash prompt but i'm not even able to write 'ls'. uml clearly doesn't like me :(( i've never been able to build a working uml kernel :((( -----[ Domenico Andreoli, aka cavok --[ http://filibusta.crema.unimi.it/~cavok/gpgkey.asc ---[ 3A0F 2F80 F79C 678A 8936 4FEE 0677 9033 A20E BC50 |
From: Nuno S. <nun...@vg...> - 2004-01-12 01:41:32
|
Domenico Andreoli wrote: >> > if i use a 2.4.23-1um on the same host i get the bash prompt but i'm not > even able to write 'ls'. > I had a similar problem with 2.4.23-um1. If the host machine has a clock bigger than 2GHZ you'll need a little modification. In arch/um/kernel/time.c modify the get_host_hz(void) function: static unsigned long long get_host_hz(void) { char mhzline[16], *end; int ret, mult, rest, len; unsigned long long mhz; ret = cpu_feature("cpu MHz", mhzline, sizeof(mhzline) / sizeof(mhzline[0])); ..etc... The only thing different is mhz declared as unsigned long long. This patch was posted by someone else and it fixes the problem for me. Hope this helps, Nuno Silva |
From: BlaisorBlade <bla...@ya...> - 2004-01-11 16:56:44
|
Alle 13:54, domenica 11 gennaio 2004, Ingo Molnar ha scritto: > > Note I've not tested it. However, it seems quite non-intrusive and > > straightforward. Watch that LKML thread to see if the author posts > > other versions of the patch. > > FYI, 2.6.1 broke the UML arch somewhere. Applying the v8 patch (and > fixing up the trivial breakages) produces an UML kernel that boots and > gets to a shell prompt, but the first 'ls' either causes an infinite > loop in skas mode, or a segfault (and kernel hang) in tt mode. Yes, I was trying some other changes while getting a similar bug: INIT: PANIC: segmentation violation at 0xa016db58! sleeping for 30 seconds. INIT: PANIC: segmentation violation at 0xa016db58! sleeping for 30 seconds. I was confused by the fact that Andreoli seemed to have something working with the fixmap change, and tried to search something in my patches. Thanx a lot! So I won't get crazy searching a bug in my changes! > (note that this happens with my combo patch variant too, so it's a > 2.6.1-induced breakage.) Something worth noting is that while in arch/i386/kernel/irq.c(and in all arch's irq.c) some fixes were done, they weren't applied to arch/um/kernel/irq.c. Open patch-2.6.1 and compare the changes to those two files: there is this piece which is not applied to the UML arch: int show_interrupts(struct seq_file *p, void *v) { - int i, j; + int i = *(loff_t *) v, j; and what follows in that hunk. This missing change probably is not *this* bug(all could be but we are not doing cat /proc/interrupts), but it's a bug anyway. -- cat <<EOSIGN Paolo Giarrusso, aka Blaisorblade Linux Kernel 2.4.23/2.6.0 on an i686; Linux registered user n. 292729 EOSIGN |
From: Domenico A. <ca...@li...> - 2004-01-12 01:45:06
|
BlaisorBlade wrote: > Alle 13:54, domenica 11 gennaio 2004, Ingo Molnar ha scritto: >>FYI, 2.6.1 broke the UML arch somewhere. Applying the v8 patch (and >>fixing up the trivial breakages) produces an UML kernel that boots and >>gets to a shell prompt, but the first 'ls' either causes an infinite >>loop in skas mode, or a segfault (and kernel hang) in tt mode. > > > Yes, I was trying some other changes while getting a similar bug: > > INIT: PANIC: segmentation violation at 0xa016db58! sleeping for 30 seconds. > INIT: PANIC: segmentation violation at 0xa016db58! sleeping for 30 seconds. > > I was confused by the fact that Andreoli seemed to have something working with > the fixmap change, and tried to search something in my patches. > "seemed" is quite the right word, it was the most working uml kernel i have ever had, but it was far from being somewhat usable. sorry to have confused you :( ciao domenico ps: paolo, mi fa un certo effetto dover comunicare in inglese ad un altro italiano :)) -----[ Domenico Andreoli, aka cavok --[ http://filibusta.crema.unimi.it/~cavok/gpgkey.asc ---[ 3A0F 2F80 F79C 678A 8936 4FEE 0677 9033 A20E BC50 |
From: Jeff D. <jd...@ad...> - 2004-01-09 23:25:54
|
ca...@li... said: > i had also needed to modify PTRACE_POKEUSR and PTRACE_POKEUSR in > PTRACE_POKEUSER and PTRACE_POKEUSER in file arch/um/sys-i386/ > ptrace_user.c. This is fixed in the upcoming 2.6 UML patch. The non-constant vma I'll fix in the 2.6.1 patch if it's not fixed in mainline first. In the meantime, Nitika Danilov has posted a fix for that. Jeff |