radeon: Fix system hang issue when using KMS with older cards
authorAdis Hamzić <adis@hamzadis.com>
Sun, 2 Jun 2013 14:47:54 +0000 (16:47 +0200)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 13 Jun 2013 16:45:00 +0000 (09:45 -0700)
commitf6321eec602f4eed6a78b0cbb91514d7c4cb74b7
tree4187e4559dae676e1aed8fc99adcd1734b9649ed
parented753efb898ba2929d19f9cc050c432be54aeae7
radeon: Fix system hang issue when using KMS with older cards

commit e49f3959a96dc279860af7e86e6dbcfda50580a5 upstream.

The current radeon driver initialization routines, when using KMS, are written
so that the IRQ installation routine is called before initializing the WB buffer
and the CP rings. With some ASICs, though, the IRQ routine tries to access the
GFX_INDEX ring causing a call to RREG32 with the value of -1 in
radeon_fence_read. This, in turn causes the system to completely hang with some
cards, requiring a hard reset.

A call stack that can cause such a hang looks like this (using rv515 ASIC for the
example here):
 * rv515_init (rv515.c)
 * radeon_irq_kms_init (radeon_irq_kms.c)
 * drm_irq_install (drm_irq.c)
 * radeon_driver_irq_preinstall_kms (radeon_irq_kms.c)
 * rs600_irq_process (rs600.c)
 * radeon_fence_process - due to SW interrupt (radeon_fence.c)
 * radeon_fence_read (radeon_fence.c)
 * hang due to RREG32(-1)

The patch moves the IRQ installation to the card startup routine, after the ring
has been initialized, but before the IRQ has been set. This fixes the issue, but
requires a check to see if the IRQ is already installed, as is the case in the
system resume codepath.
I have tested the patch on three machines using the rv515, the rv770 and the
evergreen ASIC. They worked without issues.

This seems to be a known issue and has been reported on several bug tracking
sites by various distributions (see links below). Most of reports recommend
booting the system with KMS disabled and then enabling KMS by reloading the
radeon module. For some reason, this was indeed a usable workaround, however,
UMS is now deprecated and disabled by default.

Bug reports:
https://bugzilla.redhat.com/show_bug.cgi?id=845745
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/561789
https://bbs.archlinux.org/viewtopic.php?id=156964

Signed-off-by: Adis Hamzić <adis@hamzadis.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
13 files changed:
drivers/gpu/drm/radeon/evergreen.c
drivers/gpu/drm/radeon/ni.c
drivers/gpu/drm/radeon/r100.c
drivers/gpu/drm/radeon/r300.c
drivers/gpu/drm/radeon/r420.c
drivers/gpu/drm/radeon/r520.c
drivers/gpu/drm/radeon/r600.c
drivers/gpu/drm/radeon/rs400.c
drivers/gpu/drm/radeon/rs600.c
drivers/gpu/drm/radeon/rs690.c
drivers/gpu/drm/radeon/rv515.c
drivers/gpu/drm/radeon/rv770.c
drivers/gpu/drm/radeon/si.c