Re: [git pull] drm fixes + agp + one fb patch (bisected)

From: Alex Deucher
Date: Wed Jun 30 2010 - 12:04:43 EST

Next message: Christoph Egger: "[PATCH 02/33] Removing dead ARCH_DAVINCI_TNETV107X"
Previous message: Christoph Egger: "[PATCH 33/33] Removing dead MACH_U300_USE_I2S_AS_MASTER"
In reply to: Markus Trippelsdorf: "Re: [git pull] drm fixes + agp + one fb patch (bisected)"
Next in thread: Markus Trippelsdorf: "Re: [git pull] drm fixes + agp + one fb patch (bisected)"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

On Wed, Jun 30, 2010 at 11:50 AM, Markus Trippelsdorf
<markus@xxxxxxxxxxxxxxx> wrote:
> On Wed, Jun 30, 2010 at 11:34:53AM -0400, Alex Deucher wrote:
>> On Wed, Jun 30, 2010 at 11:14 AM, Markus Trippelsdorf
>> <markus@xxxxxxxxxxxxxxx> wrote:
>> > On Wed, Jun 30, 2010 at 11:03:33AM -0400, Alex Deucher wrote:
>> >> On Wed, Jun 30, 2010 at 10:58 AM, Markus Trippelsdorf
>> >> <markus@xxxxxxxxxxxxxxx> wrote:
>> >> > On Wed, Jun 30, 2010 at 10:49:41AM -0400, Alex Deucher wrote:
>> >> >> On Wed, Jun 30, 2010 at 3:31 AM, Markus Trippelsdorf
>> >> >> <markus@xxxxxxxxxxxxxxx> wrote:
>> >> >> > On Wed, Jun 30, 2010 at 08:54:40AM +0200, Markus Trippelsdorf wrote:
>> >> >> >> On Wed, Jun 30, 2010 at 02:03:04AM +0100, Dave Airlie wrote:
>> >> >> >> >
>> >> >> >> > one fb layer fix in a flag I introduced,
>> >> >> >> >
>> >> >> >> > the rest are drm fixes:
>> >> >> >> > radeon fixes: the larger ones in the command stream checker for older cards,
>> >> >> >> > which was causing a lot of userspace apps to fail. Also some powerpc server fixes.
>> >> >> >> > along with some updates to the evergreen command stream checker introduced in -rc1.
>> >> >> >> >
>> >> >> >> > agp: fix issue with warning on memory allocation + fallback to vmalloc.
>> >> >> >> > ttm: fix regression introduced in -rc1 in memory allocation paths.
>> >> >> >> >
>> >> >> >> > The following changes since commit 7e27d6e778cd87b6f2415515d7127eba53fe5d02:
>> >> >> >> >
>> >> >> >> > Linux 2.6.35-rc3 (2010-06-11 19:14:04 -0700)
>> >> >> >> >
>> >> >> >>
>> >> >> >> I've tested these patches and they break my setup (RS780). On reboot, the
>> >> >> >> monitor goes straight to powersaving mode and no framebuffer is shown.
>> >> >> >
>> >> >> > This is the result of the bisection:
>> >> >> >
>> >> >> > 07d4190327b02ab3aaad25a2d168f79d92e8f8c2 is the first bad commit
>> >> >> > commit 07d4190327b02ab3aaad25a2d168f79d92e8f8c2
>> >> >> > Author: Alex Deucher <alexdeucher@xxxxxxxxx>
>> >> >> > Date: Sat Jun 12 11:50:13 2010 -0400
>> >> >> >
>> >> >> > drm/radeon/kms: fix bandwidth calculation when sideport is present
>> >> >> >
>> >> >> > Fixes fdo bug 27529:
>> >> >> > https://bugs.freedesktop.org/show_bug.cgi?id=27529
>> >> >> >
>> >> >> > Reported-by: steckdenis@xxxxxxxx
>> >> >> > Signed-off-by: Alex Deucher <alexdeucher@xxxxxxxxx>
>> >> >> > Cc: stable <stable@xxxxxxxxxx>
>> >> >> > Signed-off-by: Dave Airlie <airlied@xxxxxxxxxx>
>> >> >>
>> >> >> Markus, can you send me a copy of your vbios? as root:
>> >> >> cd /sys/bus/pci/devices/<pci bus id of video card>
>> >> >> echo 1 > rom
>> >> >> cat rom ? /tmp/vbio.rom
>> >> >> echo 0 > rom
>> >> >
>> >> > Attached.
>> >> >
>> >> >>
>> >> >> Also, when you say "on reboot" do you mean the card comes up ok
>> >> >> initially with the new kernel and then fails after a reboot, or fails
>> >> >> directly after booting the new kernel? Also, is there an oops or
>> >> >> anything like that in dmesg?
>> >> >
>> >> > It fails directly after booting the new kernel. There is no oops or
>> >> > anything like it in dmesg, because the kernel fails so early that it
>> >> > could not log anything.
>> >>
>> >> Is the machine still pingable? Can you ssh in? or is it completely dead?
>> >
>> > I don't think so. There is no disk activivy after the hang, but maybe
>> > the kernel displays an oops, which I just can't see.
>>
>> The attached patch should fix the issue.
>
> Sorry, but it does not. I've still the same issue with your patch
> applied.

New patch attached. The rs690_pm_info function wasn't getting called
at all on rs780/rs880 and it needs to be.

Alex
From b79c8299fec572953ef6d870990293c0ee4ca678 Mon Sep 17 00:00:00 2001
From: Alex Deucher <alexdeucher@xxxxxxxxx>
Date: Wed, 30 Jun 2010 12:02:03 -0400
Subject: [PATCH] drm/radeon/kms/igp: fix possible divide by 0 in bandwidth code (v2)

Some IGP systems specify the system memory clock in the Firmware
table rather than the IGP info table. Check both and make sure
we have a value system memory clock value.

v2: make sure rs690_pm_info is called on rs780/rs880 as well.

Signed-off-by: Alex Deucher <alexdeucher@xxxxxxxxx>
---
drivers/gpu/drm/radeon/r600.c | 4 +++-
drivers/gpu/drm/radeon/radeon.h | 1 +
drivers/gpu/drm/radeon/rs690.c | 35 +++++++++++++++++++----------------
3 files changed, 23 insertions(+), 17 deletions(-)

diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c
index 7f48593..1d5057b 100644
--- a/drivers/gpu/drm/radeon/r600.c
+++ b/drivers/gpu/drm/radeon/r600.c
@@ -1224,8 +1224,10 @@ int r600_mc_init(struct radeon_device *rdev)
rdev->mc.visible_vram_size = rdev->mc.aper_size;
r600_vram_gtt_location(rdev, &rdev->mc);

- if (rdev->flags & RADEON_IS_IGP)
+ if (rdev->flags & RADEON_IS_IGP) {
+ rs690_pm_info(rdev);
rdev->mc.igp_sideport_enabled = radeon_atombios_sideport_present(rdev);
+ }
radeon_update_bandwidth_info(rdev);
return 0;
}
diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h
index 35f0800..9fcdba7 100644
--- a/drivers/gpu/drm/radeon/radeon.h
+++ b/drivers/gpu/drm/radeon/radeon.h
@@ -177,6 +177,7 @@ void radeon_pm_resume(struct radeon_device *rdev);
void radeon_combios_get_power_modes(struct radeon_device *rdev);
void radeon_atombios_get_power_modes(struct radeon_device *rdev);
void radeon_atom_set_voltage(struct radeon_device *rdev, u16 level);
+void rs690_pm_info(struct radeon_device *rdev);

/*
* Fences.
diff --git a/drivers/gpu/drm/radeon/rs690.c b/drivers/gpu/drm/radeon/rs690.c
index 64b94a8..f4f0a61 100644
--- a/drivers/gpu/drm/radeon/rs690.c
+++ b/drivers/gpu/drm/radeon/rs690.c
@@ -79,7 +79,13 @@ void rs690_pm_info(struct radeon_device *rdev)
tmp.full = dfixed_const(100);
rdev->pm.igp_sideport_mclk.full = dfixed_const(info->info.ulBootUpMemoryClock);
rdev->pm.igp_sideport_mclk.full = dfixed_div(rdev->pm.igp_sideport_mclk, tmp);
- rdev->pm.igp_system_mclk.full = dfixed_const(le16_to_cpu(info->info.usK8MemoryClock));
+ if (info->info.usK8MemoryClock)
+ rdev->pm.igp_system_mclk.full = dfixed_const(le16_to_cpu(info->info.usK8MemoryClock));
+ else if (rdev->clock.default_mclk) {
+ rdev->pm.igp_system_mclk.full = dfixed_const(rdev->clock.default_mclk);
+ rdev->pm.igp_system_mclk.full = dfixed_div(rdev->pm.igp_system_mclk, tmp);
+ } else
+ rdev->pm.igp_system_mclk.full = dfixed_const(400);
rdev->pm.igp_ht_link_clk.full = dfixed_const(le16_to_cpu(info->info.usFSBClock));
rdev->pm.igp_ht_link_width.full = dfixed_const(info->info.ucHTLinkWidth);
break;
@@ -87,34 +93,31 @@ void rs690_pm_info(struct radeon_device *rdev)
tmp.full = dfixed_const(100);
rdev->pm.igp_sideport_mclk.full = dfixed_const(info->info_v2.ulBootUpSidePortClock);
rdev->pm.igp_sideport_mclk.full = dfixed_div(rdev->pm.igp_sideport_mclk, tmp);
- rdev->pm.igp_system_mclk.full = dfixed_const(info->info_v2.ulBootUpUMAClock);
+ if (info->info_v2.ulBootUpUMAClock)
+ rdev->pm.igp_system_mclk.full = dfixed_const(info->info_v2.ulBootUpUMAClock);
+ else if (rdev->clock.default_mclk)
+ rdev->pm.igp_system_mclk.full = dfixed_const(rdev->clock.default_mclk);
+ else
+ rdev->pm.igp_system_mclk.full = dfixed_const(66700);
rdev->pm.igp_system_mclk.full = dfixed_div(rdev->pm.igp_system_mclk, tmp);
rdev->pm.igp_ht_link_clk.full = dfixed_const(info->info_v2.ulHTLinkFreq);
rdev->pm.igp_ht_link_clk.full = dfixed_div(rdev->pm.igp_ht_link_clk, tmp);
rdev->pm.igp_ht_link_width.full = dfixed_const(le16_to_cpu(info->info_v2.usMinHTLinkWidth));
break;
default:
- tmp.full = dfixed_const(100);
/* We assume the slower possible clock ie worst case */
- /* DDR 333Mhz */
- rdev->pm.igp_sideport_mclk.full = dfixed_const(333);
- /* FIXME: system clock ? */
- rdev->pm.igp_system_mclk.full = dfixed_const(100);
- rdev->pm.igp_system_mclk.full = dfixed_div(rdev->pm.igp_system_mclk, tmp);
- rdev->pm.igp_ht_link_clk.full = dfixed_const(200);
+ rdev->pm.igp_sideport_mclk.full = dfixed_const(200);
+ rdev->pm.igp_system_mclk.full = dfixed_const(200);
+ rdev->pm.igp_ht_link_clk.full = dfixed_const(1000);
rdev->pm.igp_ht_link_width.full = dfixed_const(8);
DRM_ERROR("No integrated system info for your GPU, using safe default\n");
break;
}
} else {
- tmp.full = dfixed_const(100);
/* We assume the slower possible clock ie worst case */
- /* DDR 333Mhz */
- rdev->pm.igp_sideport_mclk.full = dfixed_const(333);
- /* FIXME: system clock ? */
- rdev->pm.igp_system_mclk.full = dfixed_const(100);
- rdev->pm.igp_system_mclk.full = dfixed_div(rdev->pm.igp_system_mclk, tmp);
- rdev->pm.igp_ht_link_clk.full = dfixed_const(200);
+ rdev->pm.igp_sideport_mclk.full = dfixed_const(200);
+ rdev->pm.igp_system_mclk.full = dfixed_const(200);
+ rdev->pm.igp_ht_link_clk.full = dfixed_const(1000);
rdev->pm.igp_ht_link_width.full = dfixed_const(8);
DRM_ERROR("No integrated system info for your GPU, using safe default\n");
}
--
1.7.0.1

Next message: Christoph Egger: "[PATCH 02/33] Removing dead ARCH_DAVINCI_TNETV107X"
Previous message: Christoph Egger: "[PATCH 33/33] Removing dead MACH_U300_USE_I2S_AS_MASTER"
In reply to: Markus Trippelsdorf: "Re: [git pull] drm fixes + agp + one fb patch (bisected)"
Next in thread: Markus Trippelsdorf: "Re: [git pull] drm fixes + agp + one fb patch (bisected)"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]