linux/drivers
Kenneth Graunke e775278cd7 drm/xe: Invalidate L3 read-only cachelines for geometry streams too
Historically, the Vertex Fetcher unit has not been an L3 client.  That
meant that, when a buffer containing vertex data was written to, it was
necessary to issue a PIPE_CONTROL::VF Cache Invalidate to invalidate any
VF L2 cachelines associated with that buffer, so the new value would be
properly read from memory.

Since Tigerlake and later, VERTEX_BUFFER_STATE and 3DSTATE_INDEX_BUFFER
have included an "L3 Bypass Enable" bit which userspace drivers can set
to request that the vertex fetcher unit snoop L3.  However, unlike most
true L3 clients, the "VF Cache Invalidate" bit continues to only
invalidate the VF L2 cache - and not any associated L3 lines.

To handle that, PIPE_CONTROL has a new "L3 Read Only Cache Invalidation
Bit", which according to the docs, "controls the invalidation of the
Geometry streams cached in L3 cache at the top of the pipe."  In other
words, the vertex and index buffer data that gets cached in L3 when
"L3 Bypass Disable" is set.

Mesa always sets L3 Bypass Disable so that the VF unit snoops L3, and
whenever it issues a VF Cache Invalidate, it also issues a L3 Read Only
Cache Invalidate so that both L2 and L3 vertex data is invalidated.

xe is issuing VF cache invalidates too (which handles cases like CPU
writes to a buffer between GPU batches).  Because userspace may enable
L3 snooping, it needs to issue an L3 Read Only Cache Invalidate as well.

Fixes significant flickering in Firefox on Meteorlake, which was writing
to vertex buffers via the CPU between batches; the missing L3 Read Only
invalidates were causing the vertex fetcher to read stale data from L3.

Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/4460
Fixes: 6ef3bb6055 ("drm/xe: enable lite restore")
Cc: stable@vger.kernel.org # v6.13+
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://lore.kernel.org/r/20250330165923.56410-1-rodrigo.vivi@intel.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
(cherry picked from commit 61672806b5)
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
2025-04-07 13:16:07 -07:00
..
accel treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
accessibility treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
acpi libnvdimm additions for 6.15 2025-04-02 20:27:18 -07:00
amba
android
ata treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
atm treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
auxdisplay treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
base treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
bcma
block treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
bluetooth treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
bus treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
cache
cdrom
cdx
char treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
clk ARM and clkdev updates for 6.15-rc1 2025-04-03 12:21:44 -07:00
clocksource RISC-V Patches for the 6.15 Merge Window, Part 1 2025-04-04 09:49:17 -07:00
comedi treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
connector
counter Char/Misc fixes for 6.15-rc1 2025-04-02 18:03:34 -07:00
cpufreq treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
cpuidle pmdomain core: 2025-03-25 20:40:51 -07:00
crypto treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
cxl cxl for v6.15 2025-04-02 20:04:43 -07:00
dax
dca
devfreq
dio
dma treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
dma-buf treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
dpll
edac - Add infrastructure support to EDAC in order to be able to register memory 2025-03-25 14:00:26 -07:00
eisa
extcon
firewire treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
firmware treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
fpga
fsi
fwctl
gnss
gpio Networking changes for 6.15. 2025-03-26 21:48:21 -07:00
gpu drm/xe: Invalidate L3 read-only cachelines for geometry streams too 2025-04-07 13:16:07 -07:00
greybus treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
hid treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
hsi treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
hte treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
hv - The 6 patch series "Enable strict percpu address space checks" from 2025-04-01 09:29:18 -07:00
hwmon treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
hwspinlock
hwtracing Char/Misc/IIO driver updates for 6.15-rc1 2025-04-01 11:26:08 -07:00
i2c treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
i3c i3c: Add NULL pointer check in i3c_master_queue_ibi() 2025-03-31 11:44:00 +02:00
idle Power management updates for 6.15-rc1 2025-03-25 15:00:18 -07:00
iio treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
infiniband treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
input A set of final cleanups for the timer subsystem: 2025-04-06 08:35:37 -07:00
interconnect
iommu treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
ipack
irqchip irqdomain: Rename irq_get_default_host() to irq_get_default_domain() 2025-04-04 16:39:10 +02:00
isdn treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
leds treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
macintosh treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
mailbox treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
mcb
md treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
media treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
memory treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
memstick treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
message SCSI misc on 20250326 2025-03-26 19:57:34 -07:00
mfd * Maxim MAX77705: 2025-03-29 14:33:13 -07:00
misc treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
mmc treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
most treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
mtd treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
mux
net treewide: Convert new and leftover hrtimer_init() users 2025-04-05 10:30:17 +02:00
nfc treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
ntb Bug fixes for NTB Switchtec driver mw negative shift, Intel NTB link 2025-04-04 14:23:07 -07:00
nubus
nvdimm libnvdimm additions for 6.15 2025-04-02 20:27:18 -07:00
nvme treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
nvmem
of Devicetree for v6.15: 2025-03-29 11:23:16 -07:00
opp
parisc
parport treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
pci treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
pcmcia treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
peci
perf pci-v6.15-changes 2025-03-28 19:36:53 -07:00
phy phy-for-6.15 2025-04-01 12:47:11 -07:00
pinctrl Pin control changes for the v6.15 kernel cycle: 2025-03-29 16:59:16 -07:00
platform treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
pmdomain
pnp Staging driver updates for 6.15-rc1 2025-04-02 18:09:17 -07:00
power i2c-for-6.15-rc1 2025-04-01 14:21:02 -07:00
powercap Power management updates for 6.15-rc1 2025-03-25 15:00:18 -07:00
pps treewide: Convert new and leftover hrtimer_init() users 2025-04-05 10:30:17 +02:00
ps3
ptp treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
pwm Pin control changes for the v6.15 kernel cycle: 2025-03-29 16:59:16 -07:00
rapidio
ras
regulator These are objtool fixes and updates by Josh Poimboeuf, centered 2025-04-02 10:30:10 -07:00
remoteproc
reset remoteproc updates for v6.15 2025-03-29 17:18:50 -07:00
rpmsg
rtc treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
s390 treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
sbus
scsi treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
sh
siox
slimbus
soc soc: drivers for 6.15, part 2 2025-04-04 09:06:32 -07:00
soundwire soundwire updates for 6.15 2025-04-01 12:43:13 -07:00
spi spi: Fixes for v6.15 2025-04-04 09:09:34 -07:00
spmi
ssb
staging treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
target treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
tc
tee
thermal More thermal control updates for 6.15-rc1 2025-04-01 16:51:44 -07:00
thunderbolt USB/Thunderbolt update for 6.15-rc1 2025-04-02 18:23:31 -07:00
tty treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
ufs Revert "Merge tag 'irq-msi-2025-03-23' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip" 2025-03-28 11:22:54 -07:00
uio
usb treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
vdpa
vfio VFIO updates for v6.15-rc1 2025-04-01 19:35:19 -07:00
vhost virtio: features, fixes, cleanups 2025-04-01 18:52:54 -07:00
video treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
virt treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
virtio virtio: features, fixes, cleanups 2025-04-01 18:52:54 -07:00
w1
watchdog treewide: Switch/rename to timer_delete[_sync]() 2025-04-05 10:30:12 +02:00
xen - The 6 patch series "Enable strict percpu address space checks" from 2025-04-01 09:29:18 -07:00
zorro
Kconfig
Makefile