Menu
▾
▴
linux-abi-devel — Linux ABI development discussion
You can subscribe to this list here.
| 2001 |
Jan
(6) |
Feb
(6) |
Mar
(6) |
Apr
(12) |
May
(11) |
Jun
(43) |
Jul
(15) |
Aug
(16) |
Sep
(9) |
Oct
(42) |
Nov
(43) |
Dec
(15) |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2002 |
Jan
(20) |
Feb
(8) |
Mar
(12) |
Apr
(10) |
May
(11) |
Jun
(7) |
Jul
(1) |
Aug
|
Sep
(2) |
Oct
(2) |
Nov
|
Dec
|
|
From: snpe <sn...@sn...> - 2002-10-31 12:22:20
|
Can You set abi trace : echo 0xffff >/proc/sys/abi/trace and call program again What is your system : SMP or UP regards Haris Peco On Monday 28 October 2002 03:18 pm, Ricardo Javier Cardenes wrote: > Hi, > > I'm having some problems running Informix Dynamic Server OnLine 7.23 from > a Unixware 2.1 install on Linux ABI. I'm using the patch for 2.4.18, and > all binaries are working well... except the onmonitor, which segfaults. > > The only remarkable difference is that onmonitor is linked against > libthread, and in fact running the onmonitor with gdb: > > Program received signal SIGSEGV, Segmentation fault. > 0x4004dd76 in _lwp_self () from /usr/lib/libc.so.1 > > Does anyone knows somthing about this? > > Thanks in advance, > Ricardo. > > > ------------------------------------------------------- > This sf.net email is sponsored by:ThinkGeek > Welcome to geek heaven. > http://thinkgeek.com/sf > _______________________________________________ > Linux-abi-devel mailing list > Lin...@li... > https://lists.sourceforge.net/lists/listinfo/linux-abi-devel |
|
From: Ricardo J. C. <ri...@co...> - 2002-10-28 15:19:56
|
Hi, I'm having some problems running Informix Dynamic Server OnLine 7.23 from a Unixware 2.1 install on Linux ABI. I'm using the patch for 2.4.18, and all binaries are working well... except the onmonitor, which segfaults. The only remarkable difference is that onmonitor is linked against libthread, and in fact running the onmonitor with gdb: Program received signal SIGSEGV, Segmentation fault. 0x4004dd76 in _lwp_self () from /usr/lib/libc.so.1 Does anyone knows somthing about this? Thanks in advance, Ricardo. |
|
From: <jo...@hy...> - 2002-09-22 19:57:33
|
Hi,
while testing some old Xenix programs, I found two bugs in linux-abi. The
solution in both cases was in the iBCS code :-| .
case_1: these old programs tend to read a directory with read().
a) Xenix doesn't allow filenames longer than 14 characters. So longer
Linux entries have to be skipped (this was missing).
b) the long Linux inodes have to be folded to short inodes. This was
wrong, some type casts were missing.
This is in abi/svr4/misc.c
case_2: For old V3 programs (linux "file" says pre-SysV) Xenix allows to get
a read_lock for files opened with access mode WRONLY, Linux doesn't.
It looks to me as if this "feature" is part of the bug compattibility
of newer Xenix versions for those old V3 programs. I imagine that in
the dark ages of Unix there was a Xenix Version with a rather broken
implementation of locking. Read locks were handled as if they were
write locks and so on.
Parts of this "bug compattibility" are handled in user land: x286emul.
This one can't. So xnx_locking() has to set the access mode for the
file temporarily to RDWR while getting the lock. NOTE: this will
change the behaviour for V5 programs: they may get a lock in a
situation where Xenix doesn't allow the lock. To overcome this, we
probably have to invent an new personality.
This is in abi/cxenix/misc.c .
Please include this patch (or a better solution) in future linux-abi versions.
Christoph, are you still listening?
Joerg
--
------------------------------------------------------------------------------
Joerg Ahrens _/
Koenigsberger Strasse 32 _/_/
31226 Peine _/ _/
Tel.: 05171/57308 _/ _/_/_/_/
e-mail: jo...@hy... _/_/_/_/ _/
------------------------------------------------------------------------------
|
|
From: Olaf K. <ok...@ad...> - 2002-09-04 15:00:30
|
Hello, I have the need to run a very old Oracle 6 for Interactive Unix 3.2 under Linux. I compiled the 2.4.18 patches and am able to run all the Interactive Unix commands. But when i try to start the oracle db within "sqldba" with "startup", the command hangs. I already inserted some printk() and got some weirdness with semaphores: oracle creates a semaphore: Sep 4 15:49:36 n24 kernel: [oracle:439]: semsys(1, 0, 40, 0x3a0) Sep 4 15:49:36 n24 kernel: svr4_semsys: 0, 40, 928 Sep 4 15:49:36 n24 kernel: [oracle:439]: semget(0, 40, 1640) Sep 4 15:49:36 n24 kernel: [oracle:439]: semsys returns 163840 (edx:95) the pmon tries to get the semaphore: Sep 4 15:49:36 n24 kernel: [oracle:440]: semsys(2, 163840, -1073745302, 0x1) Sep 4 15:49:36 n24 kernel: svr4_semsys: 163840, -1073745302, 1 Sep 4 15:49:36 n24 kernel: svr1_semop - Args: 163840, (1, -1, 0), 1 Sep 4 15:49:36 n24 kernel: svr1_semop - semval: 0 Sep 4 15:49:36 n24 kernel: [oracle:440]: semop(1, -1, 00) after that the processes hang. The semop() manpage describes excatly this behavour - but I dont know why the process does this semop() call all the time - I even wrote a little program to set the semval, and after setting it to 10000, the oracle process called the semop() function until semval was 0 again. Any hints or tried someone this before me? Bye, Olaf |
|
From: <mal...@ya...> - 2002-07-28 14:31:41
|
Note: forwarded message attached. __________________________________________________ Do You Yahoo!? Everything you'll ever need on one web page from News and Sport to Email and Music Charts http://uk.my.yahoo.com |
|
From: Robert W M. <RW...@sy...> - 2002-06-17 13:06:14
|
Jason,
You are likely correct. It has been awhile since I have seen his name. And
my apology to Mike.
On Sunday 16 June 2002 21:14, you wrote:
> Is Mike Jagis or Mike Jagdis?
>
> I think it's Jagdis.
>
> Thanks,
>
> ##jason##
-Robert
--
Robert March
St Andrew's House
L'Orignal ON K0B 1K0
|
|
From: jason p. <jas...@sk...> - 2002-06-17 01:45:58
|
Is Mike Jagis or Mike Jagdis? I think it's Jagdis. Thanks, ##jason## On Sun, 16 Jun 2002 10:51:09 -0400 Robert W March <RW...@sy...> wrote: > On Sunday 16 June 2002 10:18, Malcolm Gambles wrote: > > I have redone the home-page, hope you like it > > > > Thanks, > > > > Malcolm [linux-abi web person] > > > Malcolm, > > Very nice indeed. Congrats on a job well done. The site is well designed > and easy to utilize. > > Mike Jagis (and undoubtedly others) made a major contribution to the > development of iBCS and iBCS2. In this regard, I feel that there should be > prominent mention of Mike Jagis' contribution both on the home page and in > the doc 'iBCS and the ABI'. What do you think? > > Regards, > Robert > -- > Robert March > St Andrew's House > L'Orignal ON K0B 1K0 > > _______________________________________________________________ > > Sponsored by: > ThinkGeek at http://www.ThinkGeek.com/ > _______________________________________________ > Linux-abi-devel mailing list > Lin...@li... > https://lists.sourceforge.net/lists/listinfo/linux-abi-devel -- ##jason## "In a world without fences, who needs Gates?" |
|
From: Robert W M. <RW...@sy...> - 2002-06-16 14:44:52
|
On Sunday 16 June 2002 10:18, Malcolm Gambles wrote:
> I have redone the home-page, hope you like it
>
> Thanks,
>
> Malcolm [linux-abi web person]
Malcolm,
Very nice indeed. Congrats on a job well done. The site is well designed
and easy to utilize.
Mike Jagis (and undoubtedly others) made a major contribution to the
development of iBCS and iBCS2. In this regard, I feel that there should be
prominent mention of Mike Jagis' contribution both on the home page and in
the doc 'iBCS and the ABI'. What do you think?
Regards,
Robert
--
Robert March
St Andrew's House
L'Orignal ON K0B 1K0
|
|
From: <mal...@ya...> - 2002-06-16 14:18:10
|
I have redone the home-page, hope you like it Thanks, Malcolm [linux-abi web person] Linux A.B.I. homepage --------------------------------- Relive the FIFA World Cup goals with exclusive video highlights! http://fifaworldcup.yahoo.com/fc/en |
|
From: snpe <sn...@sn...> - 2002-06-13 22:06:39
|
echo 0xffff >/proc/sys/abi/trace On Thursday 13 June 2002 10:07 am, mal...@HB... wrote: > All, > > Using the Linux ABI how does one turn tracing on, > I seem to remember something like 'echo 1 > /proc/abi/tracing_thingi' > Could anyone help me out here? > This would be another useful addition to the docs methinks too! > > Thanks, > > Malcolm > > -----Forwarded Message----- > > From: Brad Pepers <br...@li...> > > To: Malcolm Gambles <mal...@nt...> > > Subject: Re: ABI Module > > Date: 12 Jun 2002 15:23:57 -0600 > > > > I just installed the kernel-source-2.4.18-3 RPM package which came with Red > > Hat 7.3 and it looks like the ABI version included in the kernel is the > > 2.4.17.0 release (from looking at the Documentation/abi/Changelog file). > So > it looks like I have the one you are suggesting will help IPC problems. > > > > Now where to go from here? The first thing I would like to do is get a > trace > of what calls the software is doing so I can try to nail it down a bit > better > as to just where the problem lies. There is one simple SCO program used at > > the start of installing UniData which is supposed to just check some > > permissions and such but its hanging so thats the one I want to start > tracing > and see what I get. > > > > So how do I turn on tracing for the install program? > > > > -- > > Brad Pepers > > br...@li... > > Malcolm > mal...@nt... > > > > > --------------------------------------------------------------------------- >--- HBOS plc, Registered in Scotland No. SC218813. Registered Office: The > Mound, Edinburgh EH1 1YZ. HBOS plc is a holding company, subsidiaries of > which are regulated by the Financial Services Authority. > =========================================================================== >=== |
|
From: <jo...@hy...> - 2002-06-08 17:43:40
|
Hi Tim,
>
> I upgraded a 2.4.7-10 kernel (RedHat 7.2) box to 2.4.18-3 (RedHat 7.3),
> using the RedHat upgrade. I had 286 binaries running under the x286emul,
> binfmt_xout, binfmt_aout, abi-ibcs on the 7.2 load. I have modules.conf
> autoloading as per the autoload.txt description of the personality/module
> relationships.
>
> I got the 2.4.18 kernel patch from the link to the bitkeeper site on the ABI
> home page and applied that to my source. After that the new kernel and abi
> modules will run the 286 binaries.
In abi for kernel 2.4.12 the handling of Xenix personality changed. According to
this, the personality entries in /etc/modules.conf changed too. I guess you have
a line
alias personality-7 abi-ibcs
this line has to be
alias personality-7 abi-sco
that's all.
Actually this change should be documented in Documentation/abi/autoload.txt.
> So the question is why does dfrun show different personalities?
The answer to this question is a bit more complicated. Let us have a closer
look at the three cases:
A) personality-7 is linked to abi-ibcs and no abi modules are loaded, the
only iregistered exec domain (cat /proc/execdomains) is "0-0 Linux [kernel]":
1. dfrun is loaded by binfmt_xout . binfmt_xout tries to set the
personality of the current process to PER_XENIX (which is 7). The kernel
loads the module linked to personality 7. abi-ibcs registers an execdomain
"1-2 iBCS2/iABI4 [abi-ibcs]" . As there is still no personality 7, the
kernel assigns the default execdomain to this process which is Linux. As the
personality didn't change, there is no message.
2. When the first lcall7 happens, the default_handler for execdomain Linux
detects, that the personality has been set incorrectly and sets it to 1
(PER_SVR4). So you get the message
> "Linux Kernel:[dfrun:#####]: set personality to 4100001".
(The personality is the last byte).
From now on, lcall7 for this process is handled by abi-ibcs.
B) personality-7 is linked to abi-ibcs and abi-sco has been loaded (by running some COFF
binary).
In step 1 the kernel finds personality 7 as ist has been registered by abi-sco and
you get the message
> [dfrun:#####]: set personality to 50000007.
and everything is fine.
C) personality-7 is linked to abi-sco and no abi modules are loaded, the
only registered exec domain (cat /proc/execdomains) is "0-0 Linux [kernel]":
In step 1, abi-sco is loaded while binfmt_xout requests personality 7 .
This is confusing and I would like to have a guide "for those who want to know"
covering the internal mecanisms of linux-abi. I only know a small part of the whole
thing, so I am not able to write such a guide myself.
Joerg
--
------------------------------------------------------------------------------
Joerg Ahrens _/
Koenigsberger Strasse 32 _/_/
31226 Peine _/ _/
Tel.: 05171/57308 _/ _/_/_/_/
e-mail: jo...@hy... _/_/_/_/ _/
------------------------------------------------------------------------------
|
|
From: james <ja...@pr...> - 2002-05-31 21:33:16
|
Hi, My system is a Linux is Red Hat 7.2 and vanilla kernel 2.4.17 and linux = abi compiled buit-in kernel. My application is a SCO-Foxbase with 40 terminals. My problem is that sometimes i can=B4t open no one file with the = statement 'SET EXCLUSIVE OFF' but i can to operate all files with 'SET = EXCLUSIVE ON'. How my application is multi-user, when this happens, i am = forced to restart my system. What is happens? Thanks James Cleyton |
|
From: Tim J. <tim...@in...> - 2002-05-31 18:52:06
|
I'm hoping for some enlightenment about the relationship between personality and the various abi modules loaded. I upgraded a 2.4.7-10 kernel (RedHat 7.2) box to 2.4.18-3 (RedHat 7.3), using the RedHat upgrade. I had 286 binaries running under the x286emul, binfmt_xout, binfmt_aout, abi-ibcs on the 7.2 load. I have modules.conf autoloading as per the autoload.txt description of the personality/module relationships. I got the 2.4.18 kernel patch from the link to the bitkeeper site on the ABI home page and applied that to my source. After that the new kernel and abi modules will run the 286 binaries. However I have noticed something odd with the Dataflex database program (dfrun) which I wasn't true under the older kernel. When I run dfrun just letting autoloading occur I get errors "unix_lock:unknown error". Autoloading loads abi-util, lcall7, abi-srv4, abi-ibcs. The program runs but gives the errors. I have set logging to show all the kernel messages and get "Linux Kernel:[dfrun:#####]: set personality to 4100001". file on dfrun is "Microsoft a.out separate pure segmented word-swapped V2.3 V3.0 286 small model executable Large Text Large Data" If I run a 80386 COFF binary, autoloading loads abi-cxenix and abi-sco. Now "magically" the dfrun runs with out errors. The kernel message logging now shows [dfrun:#####]: set personality to 50000007. So the question is why does dfrun show different personalities? I suppose I could just start up everything and not worry but I am curious. Also given this how should modules.conf be configured to properly load the abi modules? Thanks Tim Jackson tja...@in... |
|
From: Alexander S. <as...@es...> - 2002-05-30 19:38:39
|
Hi! (sorry for imperfect English) I am running MSM-UNIX 4.3.0 for SCO under iBCS2. It is work fine mainly. However I have investigate that sigsuspend implementation in iBCS is _incorrect_. Although sigsuspend returns EINTR when signal arrives, signal handlers are not invoked. Probably similar problem affects sigpause emulation too. As far as I can see linux-abi contains almost the same abi_sigsuspend code (http://www.kernel.org/pub/linux/kernel/people/hch/linux-abi/v2.4/linux-abi-2.4.18.0.patch.gz) So I believe my idea & code below will be useful for linux-abi developers. What do you think ? ------------------------------------------------------------------------------ [signal.c] #ifndef __sparc__ int abi_sigsuspend(struct pt_regs * regs) { unsigned long * set; unsigned long oldset; old_sigset_t newset; int error; #if defined(EMU_BSD) && defined(PER_BSD) if (personality(PER_BSD)) { oldset = get_syscall_parameter (regs, 0); } else #endif { set = (unsigned long *)get_syscall_parameter (regs, 0); error = get_user(oldset, set); if (error) return error; } newset = map_bitvec(oldset, current->exec_domain->signal_map); #ifdef IBCS_TRACE if ((ibcs_trace & TRACE_SIGNAL) || ibcs_func_p->trace) printk("iBCS: sigsuspend oldset, newset = %lx %lx\n", oldset, newset); #endif { #if 0 extern do_sigpause(unsigned int, struct pt_regs *); return do_sigpause(newset, regs); #endif } // modby assur { volatile struct pt_regs *sregs; #define alloca(size) __builtin_alloca(size) // allocate sregs on our stack sregs = (struct pt_regs*)alloca(sizeof(struct pt_regs)); *sregs = *regs; // initialize it sregs->ebx = 0; // modify according linux-syscall conventions sregs->ecx = oldset; sregs->edx = newset; // (mask) error = SYS(sigsuspend)(); if (regs->esp != sregs->esp) // frame created on user stack { char *sigframe = (char*)(sregs->esp); // see setup_frame() // sorry, offset_of(struct sigframe, sc) hardcoded as 8 struct sigcontext *sc = (struct sigcontext*)(sigframe + 8); sc->ebx = regs->ebx; // restore altered registers sc->ecx = regs->ecx; sc->edx = regs->edx; *regs = *sregs; // invoke signal-handler } } return error; } #endif /* __sparc__ */ ------------------------------------------------------------------------------ WBR Alexander |
|
From: Pier A. <pa...@li...> - 2002-05-25 07:48:07
|
I have installed Red Hat 7.3 and with Red Hat 7.2 there wasn't problem with MF Cobol with modprobe binfmt_coff and modprobe abi-ibcs. (file = 80386 COFF executabe not stripped). Now with Red Hat 7.3 and kernel 2.4.18-3 with the command ACCEPT FROM DATE the date is always 700104, but the system date is ok. If I change the system date it's accept always 700104. What is the problem ? Thank pac |
|
From: Marc-Christian P. <m....@gm...> - 2002-05-24 19:14:08
|
Hi there,
with 2.4.18 ABI Patch applied, i get the following error with make xconfi=
g:
can't read "CONSTANT_M": no such variable
while executing
"list $CONSTANT_M"
(procedure "writeconfig" line 428)
invoked from within
"writeconfig .config include/linux/autoconf.h"
invoked from within
".f0.right.save invoke"
("uplevel" body line 1)
invoked from within
"uplevel #0 [list $w invoke]"
(procedure "tkButtonUp" line 7)
invoked from within
"tkButtonUp .f0.right.save
"
(command bound to event)
Without the patch it works ok. Any hints?
--=20
Kind regards
Marc-Christian Petersen
|
|
From: Stephan S. <sp...@bf...> - 2002-05-22 06:42:37
|
Hi Haris,
a few months ago we had a similar problem. We also run a RDBMS (not
Oracle, though) with linux-abi and about twice a week all SCO binaries
would die with segfault at the same time. We traced the problem to an
error in the standard kernel (even without abi). When programs modify
their local descriptor table (LDT, a terrible 386ism), the kernel will
not correctly change the LDT on a particular CPU under certain
circumstances. The problem arises on SMP kernels when you have (emulated
abi) programs that really need their LDT and also run programs that modif=
y
their LDT.
A kernel modified with the appended patch has been running here on a
dual Athlon box for about two months without any problems. The patch is=20
against 2.4.19-pre4 but should work on 2.4.18 as well.
Hope this helps.
Stephan
diff -ur linux-2.4.19-pre4-ppro.orig/arch/i386/kernel/ldt.c linux-2.4.19-=
pre4-ppro/arch/i386/kernel/ldt.c
--- linux-2.4.19-pre4-ppro.orig/arch/i386/kernel/ldt.c Wed Oct 17 23:46:2=
9 2001
+++ linux-2.4.19-pre4-ppro/arch/i386/kernel/ldt.c Fri Mar 22 07:23:21 200=
2
@@ -101,7 +101,6 @@
memset(segments, 0, LDT_ENTRIES*LDT_ENTRY_SIZE);
wmb();
mm->context.segments =3D segments;
- mm->context.cpuvalid =3D 1UL << smp_processor_id();
load_LDT(mm);
}
=20
diff -ur linux-2.4.19-pre4-ppro.orig/arch/i386/kernel/process.c linux-2.4=
.19-pre4-ppro/arch/i386/kernel/process.c
--- linux-2.4.19-pre4-ppro.orig/arch/i386/kernel/process.c Mon Feb 25 20:=
37:53 2002
+++ linux-2.4.19-pre4-ppro/arch/i386/kernel/process.c Fri Mar 22 07:23:08=
2002
@@ -569,7 +569,6 @@
memcpy(ldt, old_ldt, LDT_ENTRIES*LDT_ENTRY_SIZE);
}
new_mm->context.segments =3D ldt;
- new_mm->context.cpuvalid =3D ~0UL; /* valid on all CPU's - they can't h=
ave stale data */
}
=20
/*
diff -ur linux-2.4.19-pre4-ppro.orig/include/asm-i386/mmu.h linux-2.4.19-=
pre4-ppro/include/asm-i386/mmu.h
--- linux-2.4.19-pre4-ppro.orig/include/asm-i386/mmu.h Thu Jul 26 03:03:0=
4 2001
+++ linux-2.4.19-pre4-ppro/include/asm-i386/mmu.h Fri Mar 22 07:22:52 200=
2
@@ -7,7 +7,6 @@
*/
typedef struct {=20
void *segments;
- unsigned long cpuvalid;
} mm_context_t;
=20
#endif
diff -ur linux-2.4.19-pre4-ppro.orig/include/asm-i386/mmu_context.h linux=
-2.4.19-pre4-ppro/include/asm-i386/mmu_context.h
--- linux-2.4.19-pre4-ppro.orig/include/asm-i386/mmu_context.h Fri Mar 22=
07:16:02 2002
+++ linux-2.4.19-pre4-ppro/include/asm-i386/mmu_context.h Fri Mar 22 08:1=
6:50 2002
@@ -27,20 +27,31 @@
=20
static inline void switch_mm(struct mm_struct *prev, struct mm_struct *n=
ext, struct task_struct *tsk, unsigned cpu)
{
+ void *new_ldt;
+ void *old_ldt;
+
+ /*
+ * Re-load LDT if necessary
+ */
+ new_ldt =3D default_ldt;
+ if (next->context.segments)
+ new_ldt =3D next->context.segments;
+ asm ("movb 7(%1),%%ah\n"
+ "movb 4(%1),%%al\n"
+ "shl $16,%%eax\n"
+ "movw 2(%1),%%ax" : "=3D&a"(old_ldt) :
+ "r"(gdt_table + __LDT(cpu)));
+ if (old_ldt !=3D new_ldt)
+ load_LDT(next);
+
if (prev !=3D next) {
/* stop flush ipis for the previous mm */
clear_bit(cpu, &prev->cpu_vm_mask);
- /*
- * Re-load LDT if necessary
- */
- if (prev->context.segments !=3D next->context.segments)
- load_LDT(next);
#ifdef CONFIG_SMP
cpu_tlbstate[cpu].state =3D TLBSTATE_OK;
cpu_tlbstate[cpu].active_mm =3D next;
#endif
set_bit(cpu, &next->cpu_vm_mask);
- set_bit(cpu, &next->context.cpuvalid);
/* Re-load page tables */
asm volatile("movl %0,%%cr3": :"r" (__pa(next->pgd)));
}
@@ -55,8 +66,6 @@
*/
local_flush_tlb();
}
- if (!test_and_set_bit(cpu, &next->context.cpuvalid))
- load_LDT(next);
}
#endif
}
--
Stephan Springl BFW Werner V=F6lk GmbH
sp...@bf... B=FCro f=FCr W=E4rmeme=DFger=E4=
te
+49 89 82917-452 Drosselgasse 5
82166 Gr=E4felfing/M=FCnchen
|
|
From: snpe <sn...@sn...> - 2002-05-18 11:44:27
|
Hello, I am used linux-abi 2.4.17 (vanilla kenrel and linux-abi patch). All is ok with uniprocesor kernel, but I am problem with SMP Hardware is : Compaq Proliant 1600 2 X 500 MHz Pentium II (Deschutes) 512 MB RAM 2X 9GB + 1X 18 GB disk TLAN network card Linux is Red Hat 7.2 and vanilla kernel 2.4.17 and linux abi (I am tried linux 2.4.18 and linux-abi 2.4.18 and result is same) Aplications are Oracle forms (runform30,runmenu30,sqlforms30 etc) After any times aplications don't work, it finish with 'segmentation fault') Sometimes, aplicattions finish after load_object (in binfmt_coff.c) withou any system call, next time I have few system call and then aplications finish Linux-abi is modul and I am used abi-util,lcall7,abi-svr4, abi-cxenix,abi-sco,binfmt_coff. It is ok with kernel 2.2 and ibcs2 (one or two processors ) I set 0xffff in /proc/sys/abi/trace and output in atach regards haris peco |
|
From: jason p. <jas...@sk...> - 2002-05-18 07:12:01
|
hello, i have successfully converted my Unix application to Linux RH7.1 with updated kernel 2.4.17 on an IBM xSeries dual processor system. so far it runs well except that everytime i backup my data via iOmega disk using Advansys SCSI controller it hangs unless clients/terminals/users logged off first. i have it running for more than three weeks and so far this is the most alarming problem that remains unsolved. hope anybody out there can be of help regarding this issue. TIA, -- ##jason## "In a world without fences, who needs Gates?" |
|
From: snpe <sn...@sn...> - 2002-05-13 14:27:18
|
I have problem with this lists. I use localhost for mail delivery and server refuse mail with dialup connections (this is before cuple days) I send this mail with ISP server This problem is not my, only regards haris peco |
|
From: Christoph H. <hc...@in...> - 2002-05-12 11:13:35
|
On Sat, May 11, 2002 at 10:06:27PM +0200, Marc-Christian Petersen wrote: > Hi there, > > have a look: > > This appens if i select abi statically in kernel config: It looks to me as if you use some vendor kernel with Al's getattr patches applied. Linux-ABI uses the same code copied into the abi-utils module to not change more core kernel code then nessecary. You could try to remove the code in the abi/util/ directory. |
|
From: Marc-Christian P. <m....@gm...> - 2002-05-11 20:06:54
|
Hi there,
have a look:
This appens if i select abi statically in kernel config:
ld -m elf_i386 -T /usr/src/linux-2.4.18/arch/i386/vmlinux.lds -e stext=20
arch/i386/kernel/head.o arch/i386/kernel/init_task.o init/main.o=20
init/version.o --start-group arch/i386/kernel/kernel.o arch/i386/mm/mm.o=20
kernel/kernel.o mm/mm.o fs/fs.o ipc/ipc.o drivers/acpi/acpi.o=20
drivers/char/char.o drivers/block/block.o drivers/misc/misc.o=20
drivers/net/net.o drivers/media/media.o drivers/char/drm/drm.o=20
drivers/ide/idedriver.o drivers/cdrom/driver.o drivers/pci/driver.o=20
drivers/pnp/pnp.o drivers/video/video.o drivers/md/mddev.o=20
drivers/sensors/sensor.o net/network.o crypto/crypto.o abi/abi.o=20
/usr/src/linux-2.4.18/arch/i386/lib/lib.a /usr/src/linux-2.4.18/lib/lib.a=
=20
/usr/src/linux-2.4.18/arch/i386/lib/lib.a --end-group -o vmlinux
abi/abi.o: In function `getattr_minix':
abi/abi.o(.text+0x80): multiple definition of `getattr_minix'
fs/fs.o(.text+0x9f70): first defined here
abi/abi.o: In function `vfs_fstat':
abi/abi.o(.text+0x254): multiple definition of `vfs_fstat'
fs/fs.o(.text+0xa154): first defined here
ld: Warning: size of symbol `vfs_fstat' changed from 137 to 124 in abi/ab=
i.o
abi/abi.o: In function `vfs_lstat':
abi/abi.o(.text+0x1c4): multiple definition of `vfs_lstat'
fs/fs.o(.text+0xa028): first defined here
ld: Warning: size of symbol `vfs_lstat' changed from 146 to 144 in abi/ab=
i.o
abi/abi.o: In function `getattr_full':
abi/abi.o(.text+0x0): multiple definition of `getattr_full'
fs/fs.o(.text+0x9ef0): first defined here
abi/abi.o: In function `vfs_stat':
abi/abi.o(.text+0x138): multiple definition of `vfs_stat'
fs/fs.o(.text+0xa0bc): first defined here
ld: Warning: size of symbol `vfs_stat' changed from 151 to 140 in abi/abi=
=2Eo
make[1]: *** [kallsyms] Error 1
make[1]: Leaving directory `/usr/src/linux-2.4.18'
make: *** [vmlinux] Error 2
What can this be? Any help is really appreciated! :-)
Many thanks!!
--=20
Kind regards
Marc-Christian Petersen
http://sourceforge.net/projects/wolk
PGP/GnuPG Key: 1024D/569DE2E3DB441A16
Fingerprint: 3469 0CF8 CA7E 0042 7824 080A 569D E2E3 DB44 1A16
Key available at wwwkeys.pgp.net. Encrypted e-mail preferred.
|
|
From: K.S.Chen <ks...@bi...> - 2002-04-29 01:03:16
|
Hi, In linux-2.4.8-abi/abi/solaris/lfs.c, fl_ibcs_to_linux[] is used but never defined elsewhere. Please comment if following patch is correct. thanks K.S.Chen diff -Nru linux/abi/solaris/lfs.c linux-2.4.18/abi/solaris/lfs.c --- linux/abi/solaris/lfs.c Mon Apr 29 09:01:36 2002 +++ linux-2.4.18/abi/solaris/lfs.c Sun Apr 28 16:53:24 2002 @@ -21,8 +21,7 @@ #include <abi/util/map.h> -extern unsigned short fl_ibcs_to_linux[]; -extern unsigned short fl_ibcs_to_linux[]; +extern unsigned short fl_svr4_to_linux[]; int @@ -35,7 +34,7 @@ char *p; struct sockaddr_un addr; - fd = sys_open(fname, map_flags(flag, fl_ibcs_to_linux) | O_LARGEFILE, mode); + fd = sys_open(fname, map_flags(flag, fl_svr4_to_linux) | O_LARGEFILE, mode); if (fd < 0) return fd; |
|
From: <ks...@bi...> - 2002-04-28 11:23:31
|
Hi, Recently, I tried to upgrade my notebook from linux-2.4.9-abi to linux-2.4.18-abi. And found the sco development tool 'ld' is broken in linux-2.4.18-abi. After digging in the problem, I found the mmap function implementation is different in these two patches. In linux-2.4.9-abi/abi/sco/sysent.c, svr4_mmap is used; in linux-2.4.18-abi/abi/sco/sysent.c, sco_mmap is used. After modifying linux-2.4.18-abi/abi/sco/mmap.c from mapaddr = do_mmap(file, addr, len, prot, flags | MAP_FIXED, off); to mapaddr = do_mmap(file, addr, len, prot, flags, off); the tool 'ld' works again. Is this modification safe ? K.S.Chen |
|
From: Robert W M. <RW...@sy...> - 2002-04-09 13:32:42
|
On Tuesday 09 April 2002 05H:42, you wrote: > On Mon, Apr 08, 2002 at 09:17:05PM -0400, Robert W March wrote: > > I patched the 2.4.18 kernel with the new 2.4.18 patch file from > > kernel.org. The patch rolled in without error. I configured the abi > > items as modules. During 'make modules' I got the following failure: > > Please try this patch: > > --- abi/svr4/sysi86.c~ Tue Apr 9 14:34:36 2002 > +++ abi/svr4/sysi86.c Tue Apr 9 14:33:55 2002 > @@ -7,6 +7,7 @@ > > #include <linux/sched.h> > #include <linux/kernel.h> > +#include <linux/mm.h> > #include <linux/ptrace.h> > #include <linux/syscall.h> > #include <asm/uaccess.h> > @@ -129,7 +130,7 @@ > * least. For now let's lie... > * (actually SCO Unix 3.4 gives me -1...) > */ > - return put_user(FP_387, arg1); > + return put_user(FP_387, (unsigned long *)arg1); > case STIME: > /* > * Set the system time. The argument is a long, Christoph, That works. make modules completes without error and my particular application, Progress RDMS, appears to run OK with the following modules manually installed. Module Size Used by Not tainted abi-sco 12536 2 abi-cxenix 6556 0 [abi-sco] abi-svr4 74032 0 [abi-sco abi-cxenix] lcall7 1584 0 [abi-sco abi-cxenix] abi-util 1808 0 [abi-sco abi-svr4 lcall7] Thank you. -Robert |
17007 messages has been excluded from this view by a project administrator.
