http://bugs.freedesktop.org/show_bug.cgi?id=10561
ajax at nwnk dot net a...@nwnk.net changed:
What|Removed |Added
Version|CVS |git
--- Comment
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #31 from Michel Dänzer mic...@daenzer.net 2009-06-08 03:16:24
PST ---
(In reply to comment #29)
Can you reproduce it with current master or radeon-rewrite branch of mesa?
Mesa doesn't matter for this... Looking at
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #29 from Maciej Cencora m.cenc...@gmail.com 2009-06-07 16:01:44
PST ---
Can you reproduce it with current master or radeon-rewrite branch of mesa?
--
Configure bugmail: http://bugs.freedesktop.org/userprefs.cgi?tab=email
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #30 from Bernie Innocenti ber...@codewiz.org 2009-06-07 16:59:30
PST ---
(In reply to comment #29)
Can you reproduce it with current master or radeon-rewrite branch of mesa?
Sorry, I don't have access to a machine equipped with
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #28 from [EMAIL PROTECTED] 2007-06-14 01:25 PST ---
Looks like compat_radeon_cp_texture() is simply missing the code to update the
32 bit ioctl parameters before returning. So probably this will happen with any
32 bit app
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #25 from [EMAIL PROTECTED] 2007-04-09 23:45 PST ---
I really think we need some solid way to debug lockups; they are really
annoying and hard to debug right now. I think Nicolai Haehnle has done some
patches for this. I'm
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #26 from [EMAIL PROTECTED] 2007-04-10 02:03 PST ---
The following patch fixed SecondLife for me by giving
up after a few retries.
The client uploads this 1024x1024 texture several times
per second (but not every frame, I
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #27 from [EMAIL PROTECTED] 2007-04-10 02:54 PST ---
(In reply to comment #26)
Maybe the retry thing should go in for robustness, but the real bug is
obviously somewhere else.
Right, and even retries should only be
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #7 from [EMAIL PROTECTED] 2007-04-09 01:03 PST ---
(In reply to comment #6)
Since gdb is unworkable, I chose another route and analyzed strace output.
That's rarely very useful for this kind of problem unfortunately.
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #8 from [EMAIL PROTECTED] 2007-04-09 01:35 PST ---
(In reply to comment #7)
Can you still kill the secondlife process and get X back working?
Yes, I need to use kill -9. The process does not even react to
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #9 from [EMAIL PROTECTED] 2007-04-09 01:47 PST ---
It means there are currently no indirect buffers available. This can happen
intermittently during normal operation, but when it persists that probably
means the card
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #10 from [EMAIL PROTECTED] 2007-04-09 01:58 PST ---
First of all, does this also happen with a 32 bit kernel?
(In reply to comment #8)
Can you still kill the secondlife process and get X back working?
Yes, I need
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #11 from [EMAIL PROTECTED] 2007-04-09 01:59 PST ---
Created an attachment (id=9520)
-- (http://bugs.freedesktop.org/attachment.cgi?id=9520action=view)
Don't sleep between texture upload ioctls, drop hardware lock instead
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #12 from [EMAIL PROTECTED] 2007-04-09 02:06 PST ---
(In reply to comment #9)
There's no UNLOCK! Is it implicit in the swap operation?
The locking fastpaths work without ioctls; the ioctls are only needed when a
client
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #13 from [EMAIL PROTECTED] 2007-04-09 05:24 PST ---
I'm not sure it's relevant: there are no other GL clients
running at the same time.
But the SecondLife client spawns multiple threads.
Maybe it's hitting a
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #14 from [EMAIL PROTECTED] 2007-04-09 05:47 PST ---
(In reply to comment #11)
Created an attachment (id=9520)
-- (http://bugs.freedesktop.org/attachment.cgi?id=9520action=view) [details]
Don't sleep between texture
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #15 from [EMAIL PROTECTED] 2007-04-09 05:51 PST ---
(In reply to comment #13)
But the SecondLife client spawns multiple threads.
Maybe it's hitting a multithreading bug in libGL.
And this is the strace output with
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #16 from [EMAIL PROTECTED] 2007-04-09 06:08 PST ---
(In reply to comment #15)
As you can see, two threads are fiddling with GL at the same time.
Actually, I only see one thread dealing with the DRM file descriptor. Am I
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #17 from [EMAIL PROTECTED] 2007-04-09 06:18 PST ---
I enabled RADEON_DEBUG=allmsg and this are the last lines of
the output before hanging:
r300Enable( GL_TEXTURE_2D = GL_TRUE )
r300Enable( GL_COLOR_MATERIAL = GL_TRUE )
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #18 from [EMAIL PROTECTED] 2007-04-09 06:28 PST ---
(In reply to comment #16)
(In reply to comment #15)
As you can see, two threads are fiddling with GL at the same time.
Actually, I only see one thread dealing with
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #19 from [EMAIL PROTECTED] 2007-04-09 06:31 PST ---
By the way, can anybody try running the SecondLife client on
their systems to see if the bug can be reproduced?
It's here:
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #20 from [EMAIL PROTECTED] 2007-04-09 06:40 PST ---
(In reply to comment #18)
I added a retry count in the while loop, and now the client exits with an
error.
Not what I asked for...
BTW, isn't 1024x1024 too big for a
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #21 from [EMAIL PROTECTED] 2007-04-09 06:41 PST ---
(In reply to comment #19)
By the way, can anybody try running the SecondLife client on
their systems to see if the bug can be reproduced?
I can't reproduce it on my
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #22 from [EMAIL PROTECTED] 2007-04-09 07:09 PST ---
(In reply to comment #21)
I can't reproduce it on my PowerBook. Can you try a 32 bit kernel?
Sorry, I can't. The userland is 64bit.
Any other difference? Do you have
http://bugs.freedesktop.org/show_bug.cgi?id=10561
[EMAIL PROTECTED] changed:
What|Removed |Added
CC||[EMAIL PROTECTED]
--- Comment
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #24 from [EMAIL PROTECTED] 2007-04-09 22:58 PST ---
(In reply to comment #23)
I think the no-usleep patch should be committed as it helps to get gdb
working.
Unfortunately, I realized in the meantime it doesn't work in
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #4 from [EMAIL PROTECTED] 2007-04-08 03:22 PST ---
(In reply to comment #3)
I tried attaching gdb to the process, but this manage to
hang the system completely (no ping, etc.).
Are you trying this via ssh? You can't do
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #5 from [EMAIL PROTECTED] 2007-04-08 10:30 PST ---
(In reply to comment #4)
Are you trying this via ssh? You can't do it from the same X display because
gdb may interrupt the process while it's holding the DRM hardware
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #6 from [EMAIL PROTECTED] 2007-04-08 12:19 PST ---
Since gdb is unworkable, I chose another route
and analyzed strace output.
It's pretty clear what's going on:
futex(0x9b29a38, FUTEX_WAKE, 2147483647) = 0
ioctl(13,
http://bugs.freedesktop.org/show_bug.cgi?id=10561
--- Comment #1 from [EMAIL PROTECTED] 2007-04-07 04:07 PST ---
Ah, and other GL applications generally work, including Quake3.
--
Configure bugmail: http://bugs.freedesktop.org/userprefs.cgi?tab=email
--- You are receiving this
http://bugs.freedesktop.org/show_bug.cgi?id=10561
[EMAIL PROTECTED] changed:
What|Removed |Added
Status|NEW |NEEDINFO
--- Comment #2 from
http://bugs.freedesktop.org/show_bug.cgi?id=10561
[EMAIL PROTECTED] changed:
What|Removed |Added
Status|NEEDINFO|NEW
--- Comment #3 from
32 matches
Mail list logo
