340.004: read: 1; total=84
> 341.004: read: 1; total=85
> ^C
> ==
>
> The after bringing the program back into the foreground, I would have
> expected to get an overrun count of 334 or thereabouts, but it looks as
> though I'm only getting the least signifi
Hi,
poll_wait() callback may modify the waitqueue without holding the
context private lock.
Signed-off-by: Davi E. M. Arnaut <[EMAIL PROTECTED]>
diff --git a/fs/eventfd.c b/fs/eventfd.c
index 480e2b3..9c672be 100644
--- a/fs/eventfd.c
+++ b/fs/eventfd.c
@@ -50,7 +50,7 @@ int eventfd_signal(struc
Hi,
Gathering signals in bulk enables server applications to drain a signal
queue (almost full of realtime signals) more efficiently by reducing the
syscall and file look-up overhead.
Very similar to the sigtimedwait4() call described by Niels Provos,
Chuck Lever, and Stephen Tweedie in a paper e
Hi,
Shrink task_struct by replacing the notifier callback with a
notifier list. The block_all_signals() function (and the signal
notifier mechanism) has only one user at the moment, which is drm.
Pahole output for task_struct:
i386 before:
/* size: 2640, cachelines: 42 */
/* sum members:
Christoph Hellwig wrote:
> On Sun, May 20, 2007 at 12:40:11AM -0300, Davi Arnaut wrote:
>> Hi,
>>
>> Shrink task_struct by replacing the notifier callback with a
>> notifier list. The block_all_signals() function (and the signal
>> notifier mechanism) has only one
Davide Libenzi wrote:
> On Mon, 21 May 2007, Oleg Nesterov wrote:
>
>>> + schedule();
>>> + locked = signalfd_lock(ctx, &lk);
>>> + if (unlikely(!locked)) {
>>> + /*
>>> +* Let the caller read zero byte, ala socket
>>> +
Hi,
If multiple threads are parked on epoll_wait (on a single epoll fd) and
events become available, epoll performs a wake up of all threads of the
poll wait list, causing a thundering herd of processes trying to grab
the eventpoll lock.
This patch addresses this by using exclusive waiters (wake
Davide Libenzi wrote:
> On Fri, 4 May 2007, Davi Arnaut wrote:
>
>> Hi,
>>
>> If multiple threads are parked on epoll_wait (on a single epoll fd) and
>> events become available, epoll performs a wake up of all threads of the
>> poll wait list, causing a thunder
livery machinery is
not necessary. And it makes me wonder why I hadn't followed its "watch"
approach for futexes:
futex_init(); // Davide's anon fd
futex_add_watch(int fd, void *addr, int val, uint32_t mask);
futex_rm_watch(int fd, uint32_t wd);
Anyway, this unifying event machin
Ulrich Drepper wrote:
> On 5/5/07, Davi Arnaut <[EMAIL PROTECTED]> wrote:
>> A google search turns up a few users. It also addresses some complaints
>> from Drepper.
>
> There is a huge problem with this approach and we're back at the
> inadequate interface
Ulrich Drepper wrote:
> On 5/7/07, Davi Arnaut <[EMAIL PROTECTED]> wrote:
>> See Linus's message on this same thread.
>
> No. I'm talking about the userlevel side, not kernel side.
So you probably knew the answer before asking the question.
> If a thread is ca
Ulrich Drepper wrote:
> On 5/7/07, Davi Arnaut <[EMAIL PROTECTED]> wrote:
>> Anyway, we could extend epoll to be mmapable...
>
> Welcome to kevent, well, except with a lot more ballast and awkward
> interfaces.
So an mmapable epoll is equivalent to kevent.. great! Well
Add pollfs_fs.h header which contains the kernel-side declarations
and auxiliary macros for type safety checks. Those macros can be
simplified later.
Signed-off-by: Davi E. M. Arnaut <[EMAIL PROTECTED]>
---
include/linux/pollfs_fs.h | 57 ++
1 file c
upling the core filesystem from the "subsystems" (mere push and
pop operations).
Currently implemented waitable "objects" are: signals, futexes, ai/o blocks and
timers.
More details at each patch.
http://haxent.com/~davi/pollfs/
Comments are welcome.
--
Davi Arnaut
-
To
Make the plaio syscall available to user-space on x86_64.
Signed-off-by: Davi E. M. Arnaut <[EMAIL PROTECTED]>
---
arch/x86_64/ia32/ia32entry.S |1 +
include/asm-x86_64/unistd.h |4 +++-
2 files changed, 4 insertions(+), 1 deletion(-)
Index: linux-2.6/arch/x86_64/ia32/ia32entry.S
=
Export the new plaio syscall prototype. While there, make it "conditional".
Signed-off-by: Davi E. M. Arnaut <[EMAIL PROTECTED]>
---
include/linux/syscalls.h |2 ++
kernel/sys_ni.c |1 +
2 files changed, 3 insertions(+)
Index: linux-2.6/include/linux/syscalls.h
Make the pltimer syscall available to user-space on x86_64.
Signed-off-by: Davi E. M. Arnaut <[EMAIL PROTECTED]>
---
arch/x86_64/ia32/ia32entry.S |1 +
include/asm-x86_64/unistd.h |4 +++-
2 files changed, 4 insertions(+), 1 deletion(-)
Index: linux-2.6/arch/x86_64/ia32/ia32entry.S
===
Make the pltimer syscall available to user-space on x86.
Signed-off-by: Davi E. M. Arnaut <[EMAIL PROTECTED]>
---
arch/i386/kernel/syscall_table.S |1 +
include/asm-i386/unistd.h|3 ++-
2 files changed, 3 insertions(+), 1 deletion(-)
Index: linux-2.6/include/asm-i386/unistd.h
==
Asynchronously wait for FUTEX_WAKE operation on a futex if it still contains
a given value. There can be only one futex wait per file descriptor. However,
it can be rearmed (possibly at a different address) anytime.
The pollable futex approach is far superior (send and receive events from
userspac
Export the new plfutex syscall prototype. While there, make it "conditional".
Signed-off-by: Davi E. M. Arnaut <[EMAIL PROTECTED]>
---
include/linux/syscalls.h |2 ++
kernel/sys_ni.c |1 +
2 files changed, 3 insertions(+)
Index: linux-2.6/include/linux/syscalls.h
==
Make the plfutex syscall available to user-space on x86.
Signed-off-by: Davi E. M. Arnaut <[EMAIL PROTECTED]>
---
arch/i386/kernel/syscall_table.S |1 +
include/asm-i386/unistd.h|3 ++-
2 files changed, 3 insertions(+), 1 deletion(-)
Index: linux-2.6/include/asm-i386/unistd.h
==
The aio_ring_empty() function returns true if the AIO event ring has no
elements, false otherwise.
Signed-off-by: Davi E. M. Arnaut <[EMAIL PROTECTED]>
---
fs/aio.c| 17 +
include/linux/aio.h |1 +
2 files changed, 18 insertions(+)
Index: linux-2.6/fs/aio.c
===
Export the new pltimer syscall prototype. While there, make it "conditional".
Signed-off-by: Davi E. M. Arnaut <[EMAIL PROTECTED]>
---
include/linux/syscalls.h |2 ++
kernel/sys_ni.c |1 +
2 files changed, 3 insertions(+)
Index: linux-2.6/include/linux/syscalls.h
==
Make the plsignal syscall available to user-space on x86.
Signed-off-by: Davi E. M. Arnaut <[EMAIL PROTECTED]>
---
arch/i386/kernel/syscall_table.S |1 +
include/asm-i386/unistd.h|3 ++-
2 files changed, 3 insertions(+), 1 deletion(-)
Index: linux-2.6/include/asm-i386/unistd.h
=
Submit, retrieve, or poll aio requests for completion through a
file descriptor. User supplies a aio_context_t that is used to
fetch a reference to the kioctx. Once the file descriptor is
closed, the reference is decremented.
Signed-off-by: Davi E. M. Arnaut <[EMAIL PROTECTED]>
---
fs/pollfs/Mak
Add a wait queue to the task_struct in order to be able to
associate (wait for) a signal with other resources.
Signed-off-by: Davi E. M. Arnaut <[EMAIL PROTECTED]>
---
include/linux/init_task.h |1 +
include/linux/sched.h |1 +
kernel/fork.c |1 +
kernel/signal.c
Make the plaio syscall available to user-space on x86.
Signed-off-by: Davi E. M. Arnaut <[EMAIL PROTECTED]>
---
arch/i386/kernel/syscall_table.S |1 +
include/asm-i386/unistd.h|3 ++-
2 files changed, 3 insertions(+), 1 deletion(-)
Index: linux-2.6/include/asm-i386/unistd.h
Per file descriptor high-resolution timers. A classic unix file interface for
the POSIX timer_(create|settime|gettime|delete) family of functions.
Signed-off-by: Davi E. M. Arnaut <[EMAIL PROTECTED]>
---
fs/pollfs/Makefile |1
fs/pollfs/timer.c | 198 ++
Make the plsignal syscall available to user-space on x86_64.
Signed-off-by: Davi E. M. Arnaut <[EMAIL PROTECTED]>
---
arch/x86_64/ia32/ia32entry.S |3 ++-
include/asm-x86_64/unistd.h |4 +++-
2 files changed, 5 insertions(+), 2 deletions(-)
Index: linux-2.6/include/asm-x86_64/unistd.h
Make the plfutex syscall available to user-space on x86_64.
Signed-off-by: Davi E. M. Arnaut <[EMAIL PROTECTED]>
---
arch/x86_64/ia32/ia32entry.S |1 +
include/asm-x86_64/unistd.h |4 +++-
2 files changed, 4 insertions(+), 1 deletion(-)
Index: linux-2.6/arch/x86_64/ia32/ia32entry.S
===
Break apart and export the futex_wait function in order to be able to
associate (wait for) a futex with other resources.
Signed-off-by: Davi E. M. Arnaut <[EMAIL PROTECTED]>
---
include/linux/futex.h | 80 ++
kernel/futex.c| 130 ++--
Export the new plsignal syscall prototype. While there, make it "conditional".
Signed-off-by: Davi E. M. Arnaut <[EMAIL PROTECTED]>
---
include/linux/syscalls.h |2 ++
kernel/sys_ni.c |1 +
2 files changed, 3 insertions(+)
Index: linux-2.6/include/linux/syscalls.h
=
Retrieve multiple per-process signals through a file descriptor. The mask
of signals can be changed at any time. Also, the compat code can be kept
very simple.
Signed-off-by: Davi E. M. Arnaut <[EMAIL PROTECTED]>
---
fs/pollfs/Makefile |2
fs/pollfs/signal.c | 144 +
Compat handlers for the pollable signal operations. Later the0 compat operations
can operate on a per call basis.
Signed-off-by: Davi E. M. Arnaut <[EMAIL PROTECTED]>
---
fs/pollfs/signal.c | 85 +
1 file changed, 85 insertions(+)
Index: lin
The key feature of the pollfs file operations is to internally handle
pollable (waitable) resources as files without exporting complex and
bug-prone underlying (VFS) implementation details.
All resource handlers are required to implement the read, write, poll,
release operations and must not block
Eric Dumazet wrote:
> Davi Arnaut a écrit :
>> Asynchronously wait for FUTEX_WAKE operation on a futex if it still contains
>> a given value. There can be only one futex wait per file descriptor. However,
>> it can be rearmed (possibly at a different address) anytime.
>
Eric Dumazet wrote:
> Davi Arnaut a écrit :
>> Eric Dumazet wrote:
>>> Davi Arnaut a écrit :
>>>> Asynchronously wait for FUTEX_WAKE operation on a futex if it still
>>>> contains
>>>> a given value. There can be only one futex wait per
Eric Dumazet wrote:
> Davi Arnaut a écrit :
>> Eric Dumazet wrote:
>>> Davi Arnaut a écrit :
>>>> Asynchronously wait for FUTEX_WAKE operation on a futex if it still
>>>> contains
>>>> a given value. There can be only one futex wait per
Ulrich Drepper wrote:
> On 5/1/07, Davi Arnaut <[EMAIL PROTECTED]> wrote:
>> The pollable futex approach is far superior (send and receive events from
>> userspace or kernel) to eventfd and fixes (supercedes) FUTEX_FD at the same
>> time.
>> [...]
>
>
Ulrich Drepper wrote:
> On 5/1/07, Davi Arnaut <[EMAIL PROTECTED]> wrote:
>> The pollable futex approach is far superior (send and receive events from
>> userspace or kernel) to eventfd and fixes (supercedes) FUTEX_FD at the same
>> time.
>> [...]
>
>
ner and safer
to do it right instead of piling on more and more workarounds for
special situations.
It simple as is, there is no need to overdesign.
--
Davi Arnaut
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
Mo
Ulrich Drepper wrote:
On 5/2/07, Davi Arnaut <[EMAIL PROTECTED]> wrote:
It's quite easy to implement this scheme by write()ing the futexes all
at once but that would break the one futex per fd association. For
atomicity: if one of the futexes can't be queued, we would rollba
Ulrich Drepper wrote:
On 5/2/07, Davi Arnaut <[EMAIL PROTECTED]> wrote:
thread A:
int fd = plfutex(addr, 0);
do
poll(fdset+fd);
process network events
queue obj to thread B
if fd:
job processed
thread B:
wait_job();
proce
vent_ (address/value).
--
Davi Arnaut
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Ulrich Drepper wrote:
On 5/2/07, Davi Arnaut <[EMAIL PROTECTED]> wrote:
NO! Every single waiter of the _file descriptor_ is waked, not of the
futex.
And how is this better? In this world of yours a program must have
one file descriptor for each single futex which is used like thi
Davide Libenzi wrote:
On Wed, 2 May 2007, Davi Arnaut wrote:
Davide Libenzi wrote:
On Tue, 1 May 2007, Andrew Morton wrote:
David, could you provide some feedback please? The patches are stunningly
free of comments, but you used to do that to me pretty often so my
Davide Libenzi wrote:
On Wed, 2 May 2007, Davi Arnaut wrote:
So in this case I may borrow some signalfd code :-) I really like the
signalfd approach, but IMHO the code is quite ugly and duplicates
a lot of hairy code.
Ugly, really? Please ...
+ while (!mutex_trylock(&
Thomas Gleixner wrote:
> On Wed, 2007-05-02 at 02:22 -0300, Davi Arnaut wrote:
>> plain text document attachment (pollfs-timer.patch)
>> Per file descriptor high-resolution timers. A classic unix file interface for
>> the POSIX timer_(create|settime|gettime|delete
48 matches
Mail list logo
