* posix message queues
@ 2004-04-07 19:07 Andrew Morton
2004-04-07 19:15 ` Manfred Spraul
0 siblings, 1 reply; 11+ messages in thread
From: Andrew Morton @ 2004-04-07 19:07 UTC (permalink / raw)
To: Manfred Spraul; +Cc: linux-arch
Manfred, if/when this gets merged up the long-suffering arch maintainers
might appreciate a little test app to exercise the syscalls which must be
added. Do you have something suitable at hand?
Also, before we go too far the 64-bit guys may like to comment on the
syscall interface. For example, this:
struct mq_attr {
long mq_flags; /* message queue flags */
long mq_maxmsg; /* maximum number of messages */
long mq_msgsize; /* maximum message size */
long mq_curmsgs; /* number of messages currently queued */
long __reserved[4]; /* ignored for input, zeroed for output */
};
looks like it will require emulation for 32-bit apps. But if we were to
make these __u32 perhaps that could be avoided?
The patches are at
ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.5/2.6.5-mc2/broken-out/mq-01-codemove.patch
ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.5/2.6.5-mc2/broken-out/mq-02-syscalls.patch
ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.5/2.6.5-mc2/broken-out/mq-03-core-update.patch
ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.5/2.6.5-mc2/broken-out/mq-03-core.patch
ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.5/2.6.5-mc2/broken-out/mq-04-linuxext-poll.patch
ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.5/2.6.5-mc2/broken-out/mq-05-linuxext-mount.patch
ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.5/2.6.5-mc2/broken-out/mq-security-fix.patch
ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.5/2.6.5-mc2/broken-out/mq-update-01.patch
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: posix message queues
2004-04-07 19:07 posix message queues Andrew Morton
@ 2004-04-07 19:15 ` Manfred Spraul
2004-04-08 8:17 ` Arnd Bergmann
2004-04-09 23:45 ` David S. Miller
0 siblings, 2 replies; 11+ messages in thread
From: Manfred Spraul @ 2004-04-07 19:15 UTC (permalink / raw)
To: Andrew Morton; +Cc: linux-arch
Andrew Morton wrote:
>Manfred, if/when this gets merged up the long-suffering arch maintainers
>might appreciate a little test app to exercise the syscalls which must be
>added. Do you have something suitable at hand?
>
>
I have a few separate apps, I'll merge them into one app and post it to
lkml. Probably tomorrow or Friday.
>
>Also, before we go too far the 64-bit guys may like to comment on the
>syscall interface. For example, this:
>
>struct mq_attr {
> long mq_flags; /* message queue flags */
> long mq_maxmsg; /* maximum number of messages */
> long mq_msgsize; /* maximum message size */
> long mq_curmsgs; /* number of messages currently queued */
> long __reserved[4]; /* ignored for input, zeroed for output */
>};
>
>looks like it will require emulation for 32-bit apps. But if we were to
>make these __u32 perhaps that could be avoided?
>
>
No. The unix spec mandates 'long' for the final C-library api. Using u32
would mean that glibc would have to add an emulation layer in user space
for 64-bit archs. In the long run, 64-bit kernels running 64-bit apps
will be the common case. I don't want to add an emulation layer to the
common case.
--
Manfred
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: posix message queues
2004-04-07 19:15 ` Manfred Spraul
@ 2004-04-08 8:17 ` Arnd Bergmann
2004-04-08 8:49 ` Andrew Morton
` (2 more replies)
2004-04-09 23:45 ` David S. Miller
1 sibling, 3 replies; 11+ messages in thread
From: Arnd Bergmann @ 2004-04-08 8:17 UTC (permalink / raw)
To: Manfred Spraul; +Cc: linux-arch, Andrew Morton
On Wednesday 07 April 2004 21:15, you wrote:
> >looks like it will require emulation for 32-bit apps. But if we were to
> >make these __u32 perhaps that could be avoided?
> >
> >
> No. The unix spec mandates 'long' for the final C-library api. Using u32
> would mean that glibc would have to add an emulation layer in user space
> for 64-bit archs. In the long run, 64-bit kernels running 64-bit apps
> will be the common case. I don't want to add an emulation layer to the
> common case.
Coincidentally, I have just finished writing handlers for them.
I have tested the code with the open posix test suite and found the
same four failures for both 64-bit and compat mode, most tests
pass. The patch is against -mc1, but I guess it also applies to the
other trees around.
What worries me more than mq_attr compatibility is the conversion
of struct sigevent, which might turn out really hard when more
fields in there are used. AFAICS, the only other part in the kernel
ABI is sys_timer_create(), so maybe it's not too late to deprecate
the current structure and create a structure that can be used
properly for compat syscalls.
Arnd <><
http://www.arndb.de/patches/linux/2.6.5-mc1/compat_mq.diff
include/linux/compat.h | 16 +++
include/linux/mqueue.h | 4
include/linux/posix_types.h | 1
include/linux/syscalls.h | 1
include/linux/types.h | 1
ipc/Makefile | 3
ipc/compat_mq.c | 185 ++++++++++++++++++++++++++++++++++++++++++++
kernel/sys.c | 5 +
8 files changed, 212 insertions(+), 4 deletions(-)
diff -Nru a/include/linux/compat.h b/include/linux/compat.h
--- a/include/linux/compat.h Thu Apr 8 01:50:22 2004
+++ b/include/linux/compat.h Thu Apr 8 01:50:22 2004
@@ -90,6 +90,22 @@
compat_uptr_t sival_ptr;
} compat_sigval_t;
+typedef struct compat_sigevent {
+ compat_sigval_t sigev_value;
+ compat_int_t sigev_signo;
+ compat_int_t sigev_notify;
+ union {
+ compat_int_t _pad[SIGEV_PAD_SIZE];
+ compat_int_t _tid;
+
+ struct {
+ compat_uptr_t _function;
+ compat_uptr_t _attribute;
+ } _sigev_thread;
+ } _sigev_un;
+} compat_sigevent_t;
+
+
long compat_sys_semctl(int first, int second, int third, void __user *uptr);
long compat_sys_msgsnd(int first, int second, int third, void __user *uptr);
long compat_sys_msgrcv(int first, int second, int msgtyp, int third,
diff -Nru a/include/linux/mqueue.h b/include/linux/mqueue.h
--- a/include/linux/mqueue.h Thu Apr 8 01:50:22 2004
+++ b/include/linux/mqueue.h Thu Apr 8 01:50:22 2004
@@ -18,9 +18,9 @@
#ifndef _LINUX_MQUEUE_H
#define _LINUX_MQUEUE_H
-#define MQ_PRIO_MAX 32768
+#include <linux/types.h>
-typedef int mqd_t;
+#define MQ_PRIO_MAX 32768
struct mq_attr {
long mq_flags; /* message queue flags */
diff -Nru a/include/linux/posix_types.h b/include/linux/posix_types.h
--- a/include/linux/posix_types.h Thu Apr 8 01:50:22 2004
+++ b/include/linux/posix_types.h Thu Apr 8 01:50:22 2004
@@ -42,6 +42,7 @@
/* Type of a SYSV IPC key. */
typedef int __kernel_key_t;
+typedef int __kernel_mqd_t;
#include <asm/posix_types.h>
diff -Nru a/include/linux/syscalls.h b/include/linux/syscalls.h
--- a/include/linux/syscalls.h Thu Apr 8 01:50:22 2004
+++ b/include/linux/syscalls.h Thu Apr 8 01:50:22 2004
@@ -48,7 +48,6 @@
struct timezone;
struct tms;
struct utimbuf;
-typedef int mqd_t;
struct mq_attr;
#include <linux/config.h>
diff -Nru a/include/linux/types.h b/include/linux/types.h
--- a/include/linux/types.h Thu Apr 8 01:50:22 2004
+++ b/include/linux/types.h Thu Apr 8 01:50:22 2004
@@ -31,6 +31,7 @@
typedef __kernel_suseconds_t suseconds_t;
typedef __kernel_timer_t timer_t;
typedef __kernel_clockid_t clockid_t;
+typedef __kernel_mqd_t mqd_t;
#ifdef __KERNEL__
typedef __kernel_uid32_t uid_t;
diff -Nru a/ipc/Makefile b/ipc/Makefile
--- a/ipc/Makefile Thu Apr 8 01:50:22 2004
+++ b/ipc/Makefile Thu Apr 8 01:50:22 2004
@@ -4,5 +4,6 @@
obj-$(CONFIG_SYSVIPC_COMPAT) += compat.o
obj-$(CONFIG_SYSVIPC) += util.o msgutil.o msg.o sem.o shm.o
-obj-$(CONFIG_POSIX_MQUEUE) += mqueue.o msgutil.o
+obj_mq-$(CONFIG_COMPAT) += compat_mq.o
+obj-$(CONFIG_POSIX_MQUEUE) += mqueue.o msgutil.o $(obj_mq-y)
diff -Nru a/ipc/compat_mq.c b/ipc/compat_mq.c
--- /dev/null Wed Dec 31 16:00:00 1969
+++ b/ipc/compat_mq.c Thu Apr 8 01:50:22 2004
@@ -0,0 +1,185 @@
+/*
+ * ipc/compat_mq.c
+ * 32 bit emulation for POSIX message queue system calls
+ *
+ * Copyright (C) 2004 IBM Deutschland Entwicklung GmbH, IBM Corporation
+ * Author: Arnd Bergmann <arnd@arndb.de>
+ */
+
+#include <linux/compat.h>
+#include <linux/fs.h>
+#include <linux/kernel.h>
+#include <linux/mqueue.h>
+#include <linux/syscalls.h>
+
+#include <asm/uaccess.h>
+
+struct compat_mq_attr {
+ compat_long_t mq_flags; /* message queue flags */
+ compat_long_t mq_maxmsg; /* maximum number of messages */
+ compat_long_t mq_msgsize; /* maximum message size */
+ compat_long_t mq_curmsgs; /* number of messages currently queued */
+ compat_long_t __reserved[4]; /* ignored for input, zeroed for output */
+};
+
+static inline int get_compat_mq_attr(struct mq_attr *attr,
+ const struct compat_mq_attr __user *uattr)
+{
+ if (verify_area(VERIFY_READ, uattr, sizeof *uattr))
+ return -EFAULT;
+
+ return __get_user(attr->mq_flags, &uattr->mq_flags)
+ | __get_user(attr->mq_maxmsg, &uattr->mq_maxmsg)
+ | __get_user(attr->mq_msgsize, &uattr->mq_msgsize)
+ | __get_user(attr->mq_curmsgs, &uattr->mq_curmsgs);
+}
+
+static inline int put_compat_mq_attr(const struct mq_attr *attr,
+ struct compat_mq_attr __user *uattr)
+{
+ if (clear_user(uattr, sizeof *uattr))
+ return -EFAULT;
+
+ return __put_user(attr->mq_flags, &uattr->mq_flags)
+ | __put_user(attr->mq_maxmsg, &uattr->mq_maxmsg)
+ | __put_user(attr->mq_msgsize, &uattr->mq_msgsize)
+ | __put_user(attr->mq_curmsgs, &uattr->mq_curmsgs);
+}
+
+asmlinkage long compat_sys_mq_open(const char __user *u_name,
+ int oflag, compat_mode_t mode,
+ struct compat_mq_attr __user *u_attr)
+{
+ struct mq_attr attr;
+ mm_segment_t oldfs;
+ char *name;
+ long ret;
+
+ if (!u_attr)
+ return sys_mq_open(u_name, oflag, mode, 0);
+
+ if (get_compat_mq_attr(&attr, u_attr))
+ return -EFAULT;
+
+ name = getname(u_name);
+ if (IS_ERR(name))
+ return PTR_ERR(name);
+
+ oldfs = get_fs();
+ set_fs(KERNEL_DS);
+ ret = sys_mq_open(name, oflag, mode, &attr);
+ set_fs(oldfs);
+
+ putname(name);
+ return ret;
+}
+
+static struct timespec __user *compat_prepare_timeout(
+ const struct compat_timespec __user *u_abs_timeout)
+{
+ struct timespec ts, __user *u_ts;
+
+ if (!u_abs_timeout)
+ return 0;
+
+ u_ts = compat_alloc_user_space(sizeof(*u_ts));
+ if (get_compat_timespec(&ts, u_abs_timeout)
+ || copy_to_user(u_ts, &ts, sizeof(*u_ts)))
+ return ERR_PTR(-EFAULT);
+
+ return u_ts;
+}
+
+asmlinkage long compat_sys_mq_timedsend(mqd_t mqdes,
+ const char __user *u_msg_ptr,
+ size_t msg_len, unsigned int msg_prio,
+ const struct compat_timespec __user *u_abs_timeout)
+{
+ struct timespec __user *u_ts;
+
+ u_ts = compat_prepare_timeout(u_abs_timeout);
+ if (IS_ERR(u_ts))
+ return -EFAULT;
+
+ return sys_mq_timedsend(mqdes, u_msg_ptr, msg_len,
+ msg_prio, u_ts);
+}
+
+asmlinkage ssize_t compat_sys_mq_timedreceive(mqd_t mqdes,
+ char __user *u_msg_ptr,
+ size_t msg_len, unsigned int __user *u_msg_prio,
+ const struct compat_timespec __user *u_abs_timeout)
+{
+ struct timespec *u_ts;
+
+ u_ts = compat_prepare_timeout(u_abs_timeout);
+ if (IS_ERR(u_ts))
+ return -EFAULT;
+
+ return sys_mq_timedreceive(mqdes, u_msg_ptr, msg_len,
+ u_msg_prio, u_ts);
+}
+
+static int get_compat_sigevent(struct sigevent *event,
+ const struct compat_sigevent __user *u_event)
+{
+ if (verify_area(VERIFY_READ, u_event, sizeof(*u_event)))
+ return -EFAULT;
+
+ return __get_user(event->sigev_value.sival_int,
+ &u_event->sigev_value.sival_int)
+ | __get_user(event->sigev_signo, &u_event->sigev_signo)
+ | __get_user(event->sigev_notify, &u_event->sigev_notify)
+ | __get_user(event->sigev_notify_thread_id,
+ &u_event->sigev_notify_thread_id);
+}
+
+asmlinkage long compat_sys_mq_notify(mqd_t mqdes,
+ const struct compat_sigevent __user *u_notification)
+{
+ mm_segment_t oldfs;
+ struct sigevent notification;
+ long ret;
+
+ if (!u_notification)
+ return sys_mq_notify(mqdes, 0);
+
+ if (get_compat_sigevent(¬ification, u_notification))
+ return -EFAULT;
+
+ oldfs = get_fs();
+ set_fs(KERNEL_DS);
+ ret = sys_mq_notify(mqdes, ¬ification);
+ set_fs(oldfs);
+
+ return ret;
+}
+
+asmlinkage long compat_sys_mq_getsetattr(mqd_t mqdes,
+ const struct compat_mq_attr __user *u_mqstat,
+ struct compat_mq_attr __user *u_omqstat)
+{
+ struct mq_attr mqstat, omqstat;
+ struct mq_attr *p_mqstat = 0, *p_omqstat = 0;
+ mm_segment_t oldfs;
+ long ret;
+
+ if (u_mqstat) {
+ p_mqstat = &mqstat;
+ if (get_compat_mq_attr(p_mqstat, u_mqstat))
+ return -EFAULT;
+ }
+
+ if (u_omqstat)
+ p_omqstat = &omqstat;
+
+ oldfs = get_fs();
+ set_fs(KERNEL_DS);
+ ret = sys_mq_getsetattr(mqdes, p_mqstat, p_omqstat);
+ set_fs(oldfs);
+
+ if (ret)
+ return ret;
+
+ return (u_omqstat) ? put_compat_mq_attr(&omqstat, u_omqstat) : 0;
+}
diff -Nru a/kernel/sys.c b/kernel/sys.c
--- a/kernel/sys.c Thu Apr 8 01:50:22 2004
+++ b/kernel/sys.c Thu Apr 8 01:50:22 2004
@@ -266,6 +266,11 @@
cond_syscall(sys_mq_timedreceive)
cond_syscall(sys_mq_notify)
cond_syscall(sys_mq_getsetattr)
+cond_syscall(compat_sys_mq_open)
+cond_syscall(compat_sys_mq_timedsend)
+cond_syscall(compat_sys_mq_timedreceive)
+cond_syscall(compat_sys_mq_notify)
+cond_syscall(compat_sys_mq_getsetattr)
/* arch-specific weak syscall entries */
cond_syscall(sys_pciconfig_read)
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: posix message queues
2004-04-08 8:17 ` Arnd Bergmann
@ 2004-04-08 8:49 ` Andrew Morton
2004-04-08 14:08 ` Manfred Spraul
2004-04-08 20:24 ` Andrew Morton
2 siblings, 0 replies; 11+ messages in thread
From: Andrew Morton @ 2004-04-08 8:49 UTC (permalink / raw)
To: Arnd Bergmann; +Cc: manfred, linux-arch
Arnd Bergmann <arnd@arndb.de> wrote:
>
> > No. The unix spec mandates 'long' for the final C-library api. Using u32
> > would mean that glibc would have to add an emulation layer in user space
> > for 64-bit archs. In the long run, 64-bit kernels running 64-bit apps
> > will be the common case. I don't want to add an emulation layer to the
> > common case.
>
> Coincidentally, I have just finished writing handlers for them.
In file included from kernel/sys.c:8:
include/linux/compat.h:98: error: `SIGEV_PAD_SIZE' undeclared here (not in a function)
---
25-akpm/include/linux/compat.h | 1 +
1 files changed, 1 insertion(+)
diff -puN include/linux/compat.h~compat_mq-fix include/linux/compat.h
--- 25/include/linux/compat.h~compat_mq-fix 2004-04-08 01:49:19.539180328 -0700
+++ 25-akpm/include/linux/compat.h 2004-04-08 01:49:19.541180024 -0700
@@ -13,6 +13,7 @@
#include <linux/sem.h>
#include <asm/compat.h>
+#include <asm/siginfo.h>
#define compat_jiffies_to_clock_t(x) \
(((unsigned long)(x) * COMPAT_USER_HZ) / HZ)
_
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: posix message queues
2004-04-08 8:17 ` Arnd Bergmann
2004-04-08 8:49 ` Andrew Morton
@ 2004-04-08 14:08 ` Manfred Spraul
2004-04-08 20:24 ` Andrew Morton
2 siblings, 0 replies; 11+ messages in thread
From: Manfred Spraul @ 2004-04-08 14:08 UTC (permalink / raw)
To: Arnd Bergmann; +Cc: linux-arch, Andrew Morton
Arnd Bergmann wrote:
>What worries me more than mq_attr compatibility is the conversion
>of struct sigevent, which might turn out really hard when more
>fields in there are used.
>
What is the problem for the wrappers with sigevent?
>+
>+static int get_compat_sigevent(struct sigevent *event,
>+ const struct compat_sigevent __user *u_event)
>+{
>+ if (verify_area(VERIFY_READ, u_event, sizeof(*u_event)))
>+ return -EFAULT;
>+
>+ return __get_user(event->sigev_value.sival_int,
>+ &u_event->sigev_value.sival_int)
>+ | __get_user(event->sigev_signo, &u_event->sigev_signo)
>+ | __get_user(event->sigev_notify, &u_event->sigev_notify)
>+ | __get_user(event->sigev_notify_thread_id,
>+ &u_event->sigev_notify_thread_id);
>+}
>+
>
The dangerous case for the wrappers is mq_notify for SIGEV_THREAD: The
kernel will read (with copy_from_user()) from the pointer
u_event->sigev_value.sival_ptr. Fixed 32-byte buffer. The data is opaque
- the kernel copies it back to the same process without any
interpretation. Would that work on all archs?
--
Manfred
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: posix message queues
2004-04-08 8:17 ` Arnd Bergmann
2004-04-08 8:49 ` Andrew Morton
2004-04-08 14:08 ` Manfred Spraul
@ 2004-04-08 20:24 ` Andrew Morton
2 siblings, 0 replies; 11+ messages in thread
From: Andrew Morton @ 2004-04-08 20:24 UTC (permalink / raw)
To: Arnd Bergmann; +Cc: manfred, linux-arch
Arnd Bergmann <arnd@arndb.de> wrote:
>
> > No. The unix spec mandates 'long' for the final C-library api. Using u32
> > would mean that glibc would have to add an emulation layer in user space
> > for 64-bit archs. In the long run, 64-bit kernels running 64-bit apps
> > will be the common case. I don't want to add an emulation layer to the
> > common case.
>
> Coincidentally, I have just finished writing handlers for them.
It broke ppc64. Please check other architectures for the same pattern.
In file included from arch/ppc64/kernel/signal.c:29:
include/asm/ppc32.h:144: redefinition of `struct compat_sigevent'
include/asm/ppc32.h:156: redefinition of `compat_sigevent_t'
include/linux/compat.h:107: `compat_sigevent_t' previously declared here
---
25-power4-akpm/include/asm-ppc64/ppc32.h | 14 --------------
1 files changed, 14 deletions(-)
diff -puN include/asm-ppc64/ppc32.h~compat_mq-ppc-fix include/asm-ppc64/ppc32.h
--- 25-power4/include/asm-ppc64/ppc32.h~compat_mq-ppc-fix 2004-04-08 13:21:48.262814248 -0700
+++ 25-power4-akpm/include/asm-ppc64/ppc32.h 2004-04-08 13:22:35.262669184 -0700
@@ -141,20 +141,6 @@ struct ucontext32 {
struct mcontext32 uc_mcontext;
};
-typedef struct compat_sigevent {
- compat_sigval_t sigev_value;
- int sigev_signo;
- int sigev_notify;
- union {
- int _pad[SIGEV_PAD_SIZE];
- int _tid;
- struct {
- compat_uptr_t _function;
- compat_uptr_t _attribute;
- } _sigev_thread;
- } _sigev_un;
-} compat_sigevent_t;
-
struct ipc_kludge_32 {
unsigned int msgp;
int msgtyp;
_
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: posix message queues
@ 2004-04-08 22:22 Arnd Bergmann
0 siblings, 0 replies; 11+ messages in thread
From: Arnd Bergmann @ 2004-04-08 22:22 UTC (permalink / raw)
To: Andrew Morton; +Cc: Arnd Bergmann, manfred, linux-arch
Andrew Morton <akpm@osdl.org> schrieb am 08.04.2004, 22:24:28:
>
> It broke ppc64. Please check other architectures for the same pattern.
>
> In file included from arch/ppc64/kernel/signal.c:29:
> include/asm/ppc32.h:144: redefinition of `struct compat_sigevent'
> include/asm/ppc32.h:156: redefinition of `compat_sigevent_t'
> include/linux/compat.h:107: `compat_sigevent_t' previously declared here
D'oh! I remember thinking that would break but forgot to fix it. On the
other architectures, the existing structure is called sigevent_t32 or
similar, so it does not conflict but should still be changed to use
the common definition.
Arnd <><
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: posix message queues
2004-04-07 19:15 ` Manfred Spraul
2004-04-08 8:17 ` Arnd Bergmann
@ 2004-04-09 23:45 ` David S. Miller
2004-04-10 11:19 ` Manfred Spraul
1 sibling, 1 reply; 11+ messages in thread
From: David S. Miller @ 2004-04-09 23:45 UTC (permalink / raw)
To: Manfred Spraul; +Cc: akpm, linux-arch
On Wed, 07 Apr 2004 21:15:44 +0200
Manfred Spraul <manfred@colorfullife.com> wrote:
> No. The unix spec mandates 'long' for the final C-library api. Using u32
> would mean that glibc would have to add an emulation layer in user space
> for 64-bit archs. In the long run, 64-bit kernels running 64-bit apps
> will be the common case. I don't want to add an emulation layer to the
> common case.
We need to support this stuff when running 32-bit apps, you cannot
avoid this Manfred. So we'll need kernel compat layer code to
translate these structures containing 'long'. Please code up the
necessary compat.c code for this posix message queues feature.
Thanks.
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: posix message queues
2004-04-09 23:45 ` David S. Miller
@ 2004-04-10 11:19 ` Manfred Spraul
2004-04-10 11:53 ` Manfred Spraul
0 siblings, 1 reply; 11+ messages in thread
From: Manfred Spraul @ 2004-04-10 11:19 UTC (permalink / raw)
To: David S. Miller; +Cc: akpm, linux-arch
[-- Attachment #1: Type: text/plain, Size: 1169 bytes --]
David S. Miller wrote:
>On Wed, 07 Apr 2004 21:15:44 +0200
>Manfred Spraul <manfred@colorfullife.com> wrote:
>
>
>
>>No. The unix spec mandates 'long' for the final C-library api. Using u32
>>would mean that glibc would have to add an emulation layer in user space
>>for 64-bit archs. In the long run, 64-bit kernels running 64-bit apps
>>will be the common case. I don't want to add an emulation layer to the
>>common case.
>>
>>
>
>We need to support this stuff when running 32-bit apps, you cannot
>avoid this Manfred.
>
Of course. But the compat stuff is only used if a 32-bit app runs on a
64-bit kernel. If I'd use u32 for the kernel interface, then a compat
wrappers would be necessary if a 64-bit app runs on a 64-bit kernel, and
that would be the wrong thing.
> So we'll need kernel compat layer code to
>translate these structures containing 'long'. Please code up the
>necessary compat.c code for this posix message queues feature.
>
>
Arndt did 90% of the coding, it's already in 2.6.5-mm3. The only missing
part was the SIGEV_THREAD handling, I've attached a patch that adds
this, together a small user space test app.
--
Manfred
[-- Attachment #2: mq_test.c --]
[-- Type: text/x-csrc, Size: 9849 bytes --]
/*
* Test app for Linux Kernel posix message queues.
*
* Copyright (C) 2004 Manfred Spraul
*
* Redistribution of this file is permitted under the terms of the GNU
* Public License (GPL)
* $Header: /home/manfred/cvs-tree/mq/mq_test.c,v 1.3 2004/04/10 11:17:31 manfred Exp $
*/
#define _GNU_SOURCE
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <errno.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/socket.h>
#include <fcntl.h>
#include <string.h>
#include <limits.h>
/*
* Kernel-interface based on the Mqueue library:
*
* Copyright (C) 2003 Krzysztof Benedyczak & Michal Wronski
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
It is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with this software; if not, write to the Free
Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
02111-1307 USA.
*/
/****************** implementation *****************/
#define __NR_mq_open 277
#define __NR_mq_unlink (__NR_mq_open+1)
#define __NR_mq_timedsend (__NR_mq_open+2)
#define __NR_mq_timedreceive (__NR_mq_open+3)
#define __NR_mq_notify (__NR_mq_open+4)
#define __NR_mq_getsetattr (__NR_mq_open+5)
#define MQ_PRIO_MAX 32768
typedef int mqd_t;
struct mq_attr {
long mq_flags; /* message queue flags */
long mq_maxmsg; /* maximum number of messages */
long mq_msgsize; /* maximum message size */
long mq_curmsgs; /* number of messages currently queued */
long __reserved[4]; /* unused */
};
#include <signal.h>
#include <fcntl.h>
#include <time.h>
#include <stdarg.h>
mqd_t mq_open(const char *name, int oflag, /* mode_t mode, struct mq_attr *attr */ ...);
int mq_close(mqd_t mqdes);
int mq_unlink(const char *name);
int mq_send(mqd_t mqdes, const char *msg_ptr, size_t msg_len, unsigned int msg_prio);
int mq_timedsend(mqd_t mqdes, const char *msg_ptr, size_t msg_len, unsigned int msg_prio, const struct timespec *abs_timeout);
ssize_t mq_receive(mqd_t mqdes, char *msg_ptr, size_t msg_len, unsigned int *msg_prio);
ssize_t mq_timedreceive(mqd_t mqdes, char *__restrict msg_ptr, size_t msg_len, unsigned int *__restrict msg_prio, const struct timespec *__restrict abs_timeout);
int mq_notify(mqd_t mqdes, const struct sigevent *notification);
int mq_getattr(mqd_t mqdes, struct mq_attr *mqstat);
int mq_setattr(mqd_t mqdes, const struct mq_attr *__restrict mqstat, struct mq_attr *__restrict omqstat);
/*
* kernel interfaces. We use glibc's syscall(3) instead of the macros
* _syscall1, _syscall2, etc, as the macros generate compilation errors
* when mqueue.c is built as a dynamic shared library.
*/
static inline mqd_t __mq_open(const char *name,
int oflag, mode_t mode, struct mq_attr* attr)
{
return syscall(__NR_mq_open, name, oflag, mode, attr);
}
static inline int __mq_unlink(const char *name)
{
return syscall(__NR_mq_unlink, name);
}
static inline int __mq_notify(mqd_t mqdes, const struct sigevent *notification)
{
return syscall(__NR_mq_notify, mqdes, notification);
}
static inline int __mq_getsetattr(mqd_t mqdes, const struct mq_attr *mqstat, struct mq_attr *omqstat)
{
return syscall(__NR_mq_getsetattr, mqdes, mqstat, omqstat);
}
static inline int __mq_timedsend(mqd_t mqdes, const char *msg_ptr,
size_t msg_len, unsigned int msg_prio,
const struct timespec *abs_timeout)
{
return syscall(__NR_mq_timedsend, mqdes, msg_ptr, msg_len,
msg_prio, abs_timeout);
}
static inline ssize_t __mq_timedreceive(mqd_t mqdes, char *msg_ptr,
size_t msg_len, unsigned int *msg_prio,
const struct timespec *abs_timeout)
{
return syscall(__NR_mq_timedreceive, mqdes, msg_ptr, msg_len,
msg_prio, abs_timeout);
}
static inline int is_valid_path(const char *name)
{
if (name[0] != '/') {
errno = EINVAL;
return -1;
}
return 0;
}
/*
* application-visible wrappers around the kernel interfaces
*/
mqd_t mq_open(const char *name, int oflag, ...)
{
unsigned long mode;
struct mq_attr *attr;
va_list ap;
va_start(ap, oflag);
mode = va_arg(ap, unsigned long);
attr = va_arg(ap, struct mq_attr *);
va_end(ap);
if (is_valid_path(name) < 0)
return (mqd_t)-1;
/* omit leading slash */
return __mq_open(name + 1, oflag, mode, attr);
}
int mq_close(mqd_t mqdes)
{
return close(mqdes);
}
int mq_unlink(const char *name)
{
if (is_valid_path(name) < 0)
return -1;
return __mq_unlink(name+1);
}
int mq_send(mqd_t mqdes, const char *msg_ptr, size_t msg_len,
unsigned int msg_prio)
{
return __mq_timedsend(mqdes, msg_ptr, msg_len, msg_prio, NULL);
}
ssize_t mq_receive(mqd_t mqdes, char *msg_ptr, size_t msg_len,
unsigned int *msg_prio)
{
return __mq_timedreceive(mqdes, msg_ptr, msg_len, msg_prio, NULL);
}
int mq_timedsend(mqd_t mqdes, const char *msg_ptr, size_t msg_len,
unsigned int msg_prio,
const struct timespec *abs_timeout)
{
return __mq_timedsend(mqdes, msg_ptr, msg_len, msg_prio, abs_timeout);
}
ssize_t mq_timedreceive(mqd_t mqdes, char *msg_ptr, size_t msg_len,
unsigned int *msg_prio,
const struct timespec *abs_timeout)
{
return __mq_timedreceive(mqdes, msg_ptr, msg_len, msg_prio,
abs_timeout);
}
int mq_getattr(mqd_t mqdes, struct mq_attr *mqstat)
{
return __mq_getsetattr(mqdes, NULL, mqstat);
}
int mq_setattr( mqd_t mqdes, const struct mq_attr *mqstat,
struct mq_attr *omqstat)
{
return __mq_getsetattr(mqdes, mqstat, omqstat);
}
int mq_notify(mqd_t mqdes, const struct sigevent *notification)
{
return __mq_notify(mqdes, notification);
}
int got_signal = 0;
void usr1_handler(int unused)
{
printf("usr1 signal caught.\n");
got_signal = 1;
}
char buffer[131072];
int main(int argc, char **argv)
{
static const char qname[] = "/test_123";
int err;
mqd_t m, m2;
struct mq_attr ma;
printf("Test 1: mq_open()\n");
m = mq_open(qname, O_CREAT|O_RDWR, S_IRUSR | S_IWUSR, NULL);
if (m == -1) {
printf("mq_open() failed, errno %d.\n", errno);
return 1;
}
printf("Test 2: check exclusive open()\n");
m2 = mq_open(qname, O_EXCL|O_CREAT|O_RDWR, S_IRUSR | S_IWUSR, NULL);
if (m2 != -1 || errno != EEXIST) {
printf("mq_open() succeeded or wrong errno (%d).\n", errno);
return 1;
}
printf("Test 3: close handle\n");
err = mq_close(m);
if (err) {
printf("mq_close() failed.\n");
}
printf("Test 4: mq_unlink\n");
err = mq_unlink(qname);
if (err) {
printf("mq_unlink() failed, errno %d.\n", errno);
return 1;
}
printf("Test 4: mq_unlink on nonexisting file\n");
err = mq_unlink(qname);
if (err != -1 || errno != ENOENT) {
printf("mq_unlink() failed, errno %d.\n", errno);
return 1;
}
printf("Test 5: mq_open()\n");
m = mq_open(qname, O_CREAT|O_EXCL|O_RDWR, S_IRUSR | S_IWUSR, NULL);
if (m == -1) {
printf("mq_open() failed, errno %d.\n", errno);
return 1;
}
printf("Test 6: mq_getattr()\n");
err = mq_getattr(m, &ma);
if (err || ma.mq_curmsgs != 0) {
printf("mq_getattr() failed, errno %d, curmsg %ld.\n", errno, ma.mq_curmsgs);
return 1;
}
printf("Test 7: setup signal based notify().\n");
signal (SIGUSR1,usr1_handler);
struct sigevent event;
event.sigev_notify = SIGEV_SIGNAL;
event.sigev_signo = SIGUSR1;
err = mq_notify(m, &event);
if (err) {
printf("mq_notify() failed, errno %d.\n", errno);
return 1;
}
printf("Test 8: mq_send()\n");
err = mq_send(m, qname, strlen(qname), 1);
if (err) {
printf("mq_send() failed, errno %d.\n", errno);
return 1;
}
if (!got_signal) {
printf("No signal received.\n");
return 1;
}
printf("Test 9: mq_getattr() with one pending message\n");
err = mq_getattr(m, &ma);
if (err || ma.mq_curmsgs != 1) {
printf("mq_getattr() failed, errno %d, curmsg %ld.\n", errno, ma.mq_curmsgs);
return 1;
}
printf("Test 10: mq_receive()\n");
err = mq_receive(m, buffer, sizeof(buffer), NULL);
if (err < 0) {
printf("mq_receive() failed, errno %d.\n", errno);
return 1;
}
printf("Test 11: data comparison\n");
if (strcmp(qname, buffer)) {
printf("Data mismatch!!!.\n");
return 1;
}
printf("Test 12: request notification cookie.\n");
{
int fd, i;
char inc[32], outc[64];
fd = socket(PF_NETLINK, SOCK_RAW, 0);
if (fd == -1) {
printf("opening netlink socket failed, errno %d.\n", errno);
return 1;
}
for (i=0;i<32;i++)
inc[i] = 64+2*i;
event.sigev_notify = SIGEV_THREAD;
event.sigev_signo = fd;
event.sigev_value.sival_ptr = inc;
err = mq_notify(m, &event);
if (err < 0) {
printf("mq_notify() failed, errno %d.\n", errno);
return 1;
}
err = mq_send(m, qname, strlen(qname), 1);
if (err) {
printf("mq_send() failed, errno %d.\n", errno);
return 1;
}
err = recv(fd, outc, sizeof(outc), MSG_NOSIGNAL|MSG_DONTWAIT);
if (err != 32) {
printf("recv unexpected result %d, errno %d.\n", err, errno);
return 1;
}
for (i=0;i<31;i++) {
if (outc[i] != 64+2*i) {
printf("Data mismatch in cookie at offset %d: %d.\n",
i, outc[i]);
return 1;
}
}
if (outc[31] != 1) {
printf("state mismatch in cookie: got %d.\n", outc[31]);
}
}
printf("Test 13: another mq_unlink\n");
err = mq_unlink(qname);
if (err) {
printf("mq_unlink() failed, errno %d.\n", errno);
return 1;
}
printf("Test 14: mq_close()\n");
err = mq_close(m);
if (err) {
printf("mq_close() failed, errno %d.\n", errno);
return 1;
}
return 0;
}
[-- Attachment #3: patch-mqueue-compat --]
[-- Type: text/plain, Size: 793 bytes --]
// $Header$
// Kernel Version:
// VERSION = 2
// PATCHLEVEL = 6
// SUBLEVEL = 5
// EXTRAVERSION = -mm3
--- 2.6/ipc/compat_mq.c 2004-04-10 09:59:40.000000000 +0200
+++ build-2.6/ipc/compat_mq.c 2004-04-10 13:15:40.414231437 +0200
@@ -139,6 +139,7 @@
{
mm_segment_t oldfs;
struct sigevent notification;
+ char cookie[NOTIFY_COOKIE_LEN];
long ret;
if (!u_notification)
@@ -147,6 +148,14 @@
if (get_compat_sigevent(¬ification, u_notification))
return -EFAULT;
+ if (notification.sigev_notify == SIGEV_THREAD) {
+ if (copy_from_user(cookie, u_notification.sigev_value.sival_ptr,
+ NOTIFY_COOKIE_LEN)) {
+ return -EFAULT;
+ }
+ notification.sigev_value.sival_ptr = cookie;
+ }
+
oldfs = get_fs();
set_fs(KERNEL_DS);
ret = sys_mq_notify(mqdes, ¬ification);
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: posix message queues
2004-04-10 11:19 ` Manfred Spraul
@ 2004-04-10 11:53 ` Manfred Spraul
2004-04-10 20:43 ` Arnd Bergmann
0 siblings, 1 reply; 11+ messages in thread
From: Manfred Spraul @ 2004-04-10 11:53 UTC (permalink / raw)
To: linux-arch; +Cc: David S. Miller, akpm
[-- Attachment #1: Type: text/plain, Size: 329 bytes --]
Manfred Spraul wrote:
>
>+if (notification.sigev_notify == SIGEV_THREAD) {
>+ if (copy_from_user(cookie, u_notification.sigev_value.sival_ptr,
>
notification, not u_notification: sival_ptr is a union on sival_int, and
sival_int was copied to kernel space by get_compat_sigevent.
Updated patch attached, sorry.
--
Manfred
[-- Attachment #2: patch-mqueue-compat --]
[-- Type: text/plain, Size: 791 bytes --]
// $Header$
// Kernel Version:
// VERSION = 2
// PATCHLEVEL = 6
// SUBLEVEL = 5
// EXTRAVERSION = -mm3
--- 2.6/ipc/compat_mq.c 2004-04-10 09:59:40.000000000 +0200
+++ build-2.6/ipc/compat_mq.c 2004-04-10 13:46:26.713887203 +0200
@@ -139,6 +139,7 @@
{
mm_segment_t oldfs;
struct sigevent notification;
+ char cookie[NOTIFY_COOKIE_LEN];
long ret;
if (!u_notification)
@@ -147,6 +148,14 @@
if (get_compat_sigevent(¬ification, u_notification))
return -EFAULT;
+ if (notification.sigev_notify == SIGEV_THREAD) {
+ if (copy_from_user(cookie, notification.sigev_value.sival_ptr,
+ NOTIFY_COOKIE_LEN)) {
+ return -EFAULT;
+ }
+ notification.sigev_value.sival_ptr = cookie;
+ }
+
oldfs = get_fs();
set_fs(KERNEL_DS);
ret = sys_mq_notify(mqdes, ¬ification);
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: posix message queues
2004-04-10 11:53 ` Manfred Spraul
@ 2004-04-10 20:43 ` Arnd Bergmann
0 siblings, 0 replies; 11+ messages in thread
From: Arnd Bergmann @ 2004-04-10 20:43 UTC (permalink / raw)
To: Manfred Spraul
Cc: linux-arch, David S. Miller, akpm, Arnd Bergmann, Jakub Jelinek
On Saturday 10 April 2004 13:53, Manfred Spraul wrote:
> Manfred Spraul wrote:
>
> >
> >+if (notification.sigev_notify == SIGEV_THREAD) {
> >+ if (copy_from_user(cookie, u_notification.sigev_value.sival_ptr,
> >
> notification, not u_notification: sival_ptr is a union on sival_int, and
> sival_int was copied to kernel space by get_compat_sigevent.
> Updated patch attached, sorry.
I'm afraid that is still incorrect, at least on big-endian machines:
> @@ -147,6 +148,14 @@
> if (get_compat_sigevent(¬ification, u_notification))
> return -EFAULT;
>
> + if (notification.sigev_notify == SIGEV_THREAD) {
> + if (copy_from_user(cookie, notification.sigev_value.sival_ptr,
> + NOTIFY_COOKIE_LEN)) {
> + return -EFAULT;
> + }
> + notification.sigev_value.sival_ptr = cookie;
> + }
sival_ptr is not valid when passed through get_compat_sigevent, because
the 32 bit pointer might be copied to the upper half of the kernel pointer.
Even worse, the infamous s390 31 bit pointer conversion is missing.
This patch fixes that problem and two others reported by Jakub.
> > Arndt did 90% of the coding, it's already in 2.6.5-mm3.
BTW, my name is 'Arnd', not 'Arndt'. I hope we can stop this now before
this becomes some 'Russel' vs. 'Russell' problem ;-)
Arnd <><
--
[PATCH] More fixups for compat_mq
- handle SIGEV_THREAD
- don't try to convert u_attr in sys_mq_open if !O_CREAT
- handle __SI_MESGQ in copy_siginfo_to_user32 (still missing for
ppc64 and parisc)
===== ipc/compat_mq.c 1.1 vs edited =====
--- 1.1/ipc/compat_mq.c Sat Apr 10 16:46:12 2004
+++ edited/ipc/compat_mq.c Sat Apr 10 20:48:02 2004
@@ -55,7 +55,7 @@
char *name;
long ret;
- if (!u_attr)
+ if ((oflag & O_CREAT) == 0 || !u_attr)
return sys_mq_open(u_name, oflag, mode, 0);
if (get_compat_mq_attr(&attr, u_attr))
@@ -139,6 +139,8 @@
{
mm_segment_t oldfs;
struct sigevent notification;
+ char cookie[NOTIFY_COOKIE_LEN];
+ compat_uptr_t u_cookie;
long ret;
if (!u_notification)
@@ -146,6 +148,15 @@
if (get_compat_sigevent(¬ification, u_notification))
return -EFAULT;
+
+ if (notification.sigev_notify == SIGEV_THREAD) {
+ u_cookie = (compat_uptr_t)notification.sigev_value.sival_int;
+ if (copy_from_user(cookie, compat_ptr(u_cookie),
+ NOTIFY_COOKIE_LEN)) {
+ return -EFAULT;
+ }
+ notification.sigev_value.sival_ptr = cookie;
+ }
oldfs = get_fs();
set_fs(KERNEL_DS);
===== arch/ia64/ia32/ia32_signal.c 1.24 vs edited =====
--- 1.24/arch/ia64/ia32/ia32_signal.c Wed Feb 25 11:31:13 2004
+++ edited/arch/ia64/ia32/ia32_signal.c Sat Apr 10 21:06:07 2004
@@ -114,7 +114,12 @@
err |= __get_user(to->si_band, &from->si_band);
err |= __get_user(to->si_fd, &from->si_fd);
break;
- /* case __SI_RT: This is not generated by the kernel as of now. */
+ case __SI_RT: /* This is not generated by the kernel as of now. */
+ case __SI_MESGQ:
+ err |= __get_user(to->si_pid, &from->si_pid);
+ err |= __get_user(to->si_uid, &from->si_uid);
+ err |= __get_user(to->si_int, &from->si_int);
+ break;
}
}
return err;
===== arch/mips/kernel/signal32.c 1.13 vs edited =====
--- 1.13/arch/mips/kernel/signal32.c Wed Feb 25 11:31:13 2004
+++ edited/arch/mips/kernel/signal32.c Sat Apr 10 21:08:53 2004
@@ -358,7 +358,12 @@
err |= __put_user(from->si_band, &to->si_band);
err |= __put_user(from->si_fd, &to->si_fd);
break;
- /* case __SI_RT: This is not generated by the kernel as of now. */
+ case __SI_RT: /* This is not generated by the kernel as of now. */
+ case __SI_MESGQ:
+ err |= __put_user(from->si_pid, &to->si_pid);
+ err |= __put_user(from->si_uid, &to->si_uid);
+ err |= __put_user(from->si_int, &to->si_int);
+ break;
}
}
return err;
===== arch/s390/kernel/compat_signal.c 1.7 vs edited =====
--- 1.7/arch/s390/kernel/compat_signal.c Sat Mar 27 12:40:46 2004
+++ edited/arch/s390/kernel/compat_signal.c Sat Apr 10 21:05:01 2004
@@ -74,6 +74,10 @@
err |= __copy_to_user(&to->_sifields._pad, &from->_sifields._pad, SI_PAD_SIZE);
else {
switch (from->si_code >> 16) {
+ case __SI_RT: /* This is not generated by the kernel as of now. */
+ case __SI_MESGQ:
+ err |= __put_user(from->si_int, &to->si_int);
+ /* fallthrough */
case __SI_KILL >> 16:
err |= __put_user(from->si_pid, &to->si_pid);
err |= __put_user(from->si_uid, &to->si_uid);
@@ -96,7 +100,6 @@
break;
default:
break;
- /* case __SI_RT: This is not generated by the kernel as of now. */
}
}
return err;
===== arch/sparc64/kernel/signal32.c 1.32 vs edited =====
--- 1.32/arch/sparc64/kernel/signal32.c Fri Mar 26 23:16:00 2004
+++ edited/arch/sparc64/kernel/signal32.c Sat Apr 10 21:13:56 2004
@@ -129,7 +129,12 @@
err |= __put_user(from->si_trapno, &to->si_trapno);
err |= __put_user((long)from->si_addr, &to->si_addr);
break;
- /* case __SI_RT: This is not generated by the kernel as of now. */
+ case __SI_RT: /* This is not generated by the kernel as of now. */
+ case __SI_MESGQ:
+ err |= __put_user(from->si_pid, &to->si_pid);
+ err |= __put_user(from->si_uid, &to->si_uid);
+ err |= __put_user(from->si_int, &to->si_int);
+ break;
}
}
return err;
===== arch/x86_64/ia32/ia32_signal.c 1.19 vs edited =====
--- 1.19/arch/x86_64/ia32/ia32_signal.c Mon Mar 8 15:23:47 2004
+++ edited/arch/x86_64/ia32/ia32_signal.c Sat Apr 10 21:03:13 2004
@@ -85,7 +85,11 @@
err |= __put_user(from->si_overrun, &to->si_overrun);
err |= __put_user((u32)(u64)from->si_ptr, &to->si_ptr);
break;
- /* case __SI_RT: This is not generated by the kernel as of now. */
+ case __SI_RT: /* This is not generated by the kernel as of now. */
+ case __SI_MESGQ:
+ err |= __put_user(from->si_uid, &to->si_uid);
+ err |= __put_user(from->si_int, &to->si_int);
+ break;
}
}
return err;
^ permalink raw reply [flat|nested] 11+ messages in thread
end of thread, other threads:[~2004-04-10 20:34 UTC | newest]
Thread overview: 11+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2004-04-07 19:07 posix message queues Andrew Morton
2004-04-07 19:15 ` Manfred Spraul
2004-04-08 8:17 ` Arnd Bergmann
2004-04-08 8:49 ` Andrew Morton
2004-04-08 14:08 ` Manfred Spraul
2004-04-08 20:24 ` Andrew Morton
2004-04-09 23:45 ` David S. Miller
2004-04-10 11:19 ` Manfred Spraul
2004-04-10 11:53 ` Manfred Spraul
2004-04-10 20:43 ` Arnd Bergmann
-- strict thread matches above, loose matches on Subject: below --
2004-04-08 22:22 Arnd Bergmann
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox