In ice_flow_add_prof_sync(), struct ice_flow_prof_params has recently
grown in size hogging stack space when allocated there. Hogging stack
space should be avoided. Change allocation to be on the heap when needed.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Tested-by: Harikumar Bokkena <harikumarx.bokkena@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tremont has four L1 Topdown events, TOPDOWN_FE_BOUND.ALL,
TOPDOWN_BAD_SPECULATION.ALL, TOPDOWN_BE_BOUND.ALL and
TOPDOWN_RETIRING.ALL. They are available on GP counters.
Export them to sysfs and facilitate the perf stat tool.
$perf stat --topdown -- sleep 1
Performance counter stats for 'sleep 1':
retiring bad speculation frontend bound
backend bound
24.9% 16.8% 31.7%
26.6%
1.001224610 seconds time elapsed
0.001150000 seconds user
0.000000000 seconds sys
Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/1607457952-3519-1-git-send-email-kan.liang@linux.intel.com
In preparation to enable -Wimplicit-fallthrough for Clang, fix a warning
by explicitly adding a break statement instead of letting the code fall
through to the next case.
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://github.com/KSPP/linux/issues/115
In preparation to enable -Wimplicit-fallthrough for Clang, fix a warning
by explicitly adding a fallthrough pseudo-keyword as a replacement for
a /* fall through */ comment, instead of letting the code fall through
to the next case.
Notice that Clang doesn't recognize /* fall through */ comments as
implicit fall-through markings.
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://github.com/KSPP/linux/issues/115
In preparation to enable -Wimplicit-fallthrough for Clang, fix a warning
by explicitly adding a break statement instead of just letting the code
fall through to the next case.
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://github.com/KSPP/linux/issues/115
Fix to restore BTF if single-stepping causes a page fault and
it is cancelled.
Usually the BTF flag was restored when the single stepping is done
(in resume_execution()). However, if a page fault happens on the
single stepping instruction, the fault handler is invoked and
the single stepping is cancelled. Thus, the BTF flag is not
restored.
Fixes: 1ecc798c67 ("x86: debugctlmsr kprobes")
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/160389546985.106936.12727996109376240993.stgit@devnote2
Syzbot reported a lock inversion involving perf. The sore point being
perf holding exec_update_mutex() for a very long time, specifically
across a whole bunch of filesystem ops in pmu::event_init() (uprobes)
and anon_inode_getfile().
This then inverts against procfs code trying to take
exec_update_mutex.
Move the permission checks later, such that we need to hold the mutex
over less code.
Reported-by: syzbot+db9cdf3dd1f64252c6ef@syzkaller.appspotmail.com
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Sparc64 has non-pagetable aligned large page support; wire up the
pXX_leaf_size() functions to report the correct pagetable page size.
This enables PERF_SAMPLE_{DATA,CODE}_PAGE_SIZE to report accurate
pagetable leaf sizes.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20201126121121.301768209@infradead.org
Christophe Leroy wrote:
> I can help with powerpc 8xx. It is a 32 bits powerpc. The PGD has 1024
> entries, that means each entry maps 4M.
>
> Page sizes are 4k, 16k, 512k and 8M.
>
> For the 8M pages we use hugepd with a single entry. The two related PGD
> entries point to the same hugepd.
>
> For the other sizes, they are in standard page tables. 16k pages appear
> 4 times in the page table. 512k entries appear 128 times in the page
> table.
>
> When the PGD entry has _PMD_PAGE_8M bits, the PMD entry points to a
> hugepd with holds the single 8M entry.
>
> In the PTE, we have two bits: _PAGE_SPS and _PAGE_HUGE
>
> _PAGE_HUGE means it is a 512k page
> _PAGE_SPS means it is not a 4k page
>
> The kernel can by build either with 4k pages as standard page size, or
> 16k pages. It doesn't change the page table layout though.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20201126121121.364451610@infradead.org
When the seqcount_LOCKNAME_t group of data types were introduced, two
classes of seqlock.h sequence counter macros were added:
- An external public API which can either take a plain seqcount_t or
any of the seqcount_LOCKNAME_t variants.
- An internal API which takes only a plain seqcount_t.
To distinguish between the two groups, the "*_seqcount_t_*" pattern was
used for the latter. This confused a number of mm/ call-site developers,
and Linus also commented that it was not a standard practice for marking
seqlock.h internal APIs.
Distinguish the latter group of macros by prefixing a "do_".
Signed-off-by: Ahmed S. Darwish <a.darwish@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/CAHk-=wikhGExmprXgaW+MVXG1zsGpztBbVwOb23vetk41EtTBQ@mail.gmail.com
Reader optimistic spinning is helpful when the reader critical section
is short and there aren't that many readers around. It also improves
the chance that a reader can get the lock as writer optimistic spinning
disproportionally favors writers much more than readers.
Since commit d3681e269f ("locking/rwsem: Wake up almost all readers
in wait queue"), all the waiting readers are woken up so that they can
all get the read lock and run in parallel. When the number of contending
readers is large, allowing reader optimistic spinning will likely cause
reader fragmentation where multiple smaller groups of readers can get
the read lock in a sequential manner separated by writers. That reduces
reader parallelism.
One possible way to address that drawback is to limit the number of
readers (preferably one) that can do optimistic spinning. These readers
act as representatives of all the waiting readers in the wait queue as
they will wake up all those waiting readers once they get the lock.
Alternatively, as reader optimistic lock stealing has already enhanced
fairness to readers, it may be easier to just remove reader optimistic
spinning and simplifying the optimistic spinning code as a result.
Performance measurements (locking throughput kops/s) using a locking
microbenchmark with 50/50 reader/writer distribution and turbo-boost
disabled was done on a 2-socket Cascade Lake system (48-core 96-thread)
to see the impacts of these changes:
1) Vanilla - 5.10-rc3 kernel
2) Before - 5.10-rc3 kernel with previous patches in this series
2) limit-rspin - 5.10-rc3 kernel with limited reader spinning patch
3) no-rspin - 5.10-rc3 kernel with reader spinning disabled
# of threads CS Load Vanilla Before limit-rspin no-rspin
------------ ------- ------- ------ ----------- --------
2 1 5,185 5,662 5,214 5,077
4 1 5,107 4,983 5,188 4,760
8 1 4,782 4,564 4,720 4,628
16 1 4,680 4,053 4,567 3,402
32 1 4,299 1,115 1,118 1,098
64 1 3,218 983 1,001 957
96 1 1,938 944 957 930
2 20 2,008 2,128 2,264 1,665
4 20 1,390 1,033 1,046 1,101
8 20 1,472 1,155 1,098 1,213
16 20 1,332 1,077 1,089 1,122
32 20 967 914 917 980
64 20 787 874 891 858
96 20 730 836 847 844
2 100 372 356 360 355
4 100 492 425 434 392
8 100 533 537 529 538
16 100 548 572 568 598
32 100 499 520 527 537
64 100 466 517 526 512
96 100 406 497 506 509
The column "CS Load" represents the number of pause instructions issued
in the locking critical section. A CS load of 1 is extremely short and
is not likey in real situations. A load of 20 (moderate) and 100 (long)
are more realistic.
It can be seen that the previous patches in this series have reduced
performance in general except in highly contended cases with moderate
or long critical sections that performance improves a bit. This change
is mostly caused by the "Prevent potential lock starvation" patch that
reduce reader optimistic spinning and hence reduce reader fragmentation.
The patch that further limit reader optimistic spinning doesn't seem to
have too much impact on overall performance as shown in the benchmark
data.
The patch that disables reader optimistic spinning shows reduced
performance at lightly loaded cases, but comparable or slightly better
performance on with heavier contention.
This patch just removes reader optimistic spinning for now. As readers
are not going to do optimistic spinning anymore, we don't need to
consider if the OSQ is empty or not when doing lock stealing.
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Davidlohr Bueso <dbueso@suse.de>
Link: https://lkml.kernel.org/r/20201121041416.12285-6-longman@redhat.com
If the optimistic spinning queue is empty and the rwsem does not have
the handoff or write-lock bits set, it is actually not necessary to
call rwsem_optimistic_spin() to spin on it. Instead, it can steal the
lock directly as its reader bias is in the count already. If it is
the first reader in this state, it will try to wake up other readers
in the wait queue.
With this patch applied, the following were the lock event counts
after rebooting a 2-socket system and a "make -j96" kernel rebuild.
rwsem_opt_rlock=4437
rwsem_rlock=29
rwsem_rlock_steal=19
So lock stealing represents about 0.4% of all the read locks acquired
in the slow path.
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Davidlohr Bueso <dbueso@suse.de>
Link: https://lkml.kernel.org/r/20201121041416.12285-4-longman@redhat.com
The lock handoff bit is added in commit 4f23dbc1e6 ("locking/rwsem:
Implement lock handoff to prevent lock starvation") to avoid lock
starvation. However, allowing readers to do optimistic spinning does
introduce an unlikely scenario where lock starvation can happen.
The lock handoff bit may only be set when a waiter is being woken up.
In the case of reader unlock, wakeup happens only when the reader count
reaches 0. If there is a continuous stream of incoming readers acquiring
read lock via optimistic spinning, it is possible that the reader count
may never reach 0 and so the handoff bit will never be asserted.
One way to prevent this scenario from happening is to disallow optimistic
spinning if the rwsem is currently owned by readers. If the previous
or current owner is a writer, optimistic spinning will be allowed.
If the previous owner is a reader but the reader count has reached 0
before, a wakeup should have been issued. So the handoff mechanism
will be kicked in to prevent lock starvation. As a result, it should
be OK to do optimistic spinning in this case.
This patch may have some impact on reader performance as it reduces
reader optimistic spinning especially if the lock critical sections
are short the number of contending readers are small.
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Davidlohr Bueso <dbueso@suse.de>
Link: https://lkml.kernel.org/r/20201121041416.12285-3-longman@redhat.com
In preparation for converting exec_update_mutex to a rwsem so that
multiple readers can execute in parallel and not deadlock, add
down_read_interruptible. This is needed for perf_event_open to be
converted (with no semantic changes) from working on a mutex to
wroking on a rwsem.
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/87k0tybqfy.fsf@x220.int.ebiederm.org
In preparation for converting exec_update_mutex to a rwsem so that
multiple readers can execute in parallel and not deadlock, add
down_read_killable_nested. This is needed so that kcmp_lock
can be converted from working on a mutexes to working on rw_semaphores.
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/87o8jabqh3.fsf@x220.int.ebiederm.org
Samsung mach/soc changes for v5.11
1. Do not use of_machine_is_compatible() in early CPU hotplug core. Full
device tree walk causes "suspicious RCU usage" warnings.
2. Clear prefetch bits in default l2c_aux_val of L310 L2C - they are not
needed.
3. Extend cpuidle support to P4 Note boards (Exynos4412).
* tag 'samsung-soc-5.11' of git://git.kernel.org/pub/scm/linux/kernel/git/krzk/linux:
ARM: exynos: extend cpuidle support to P4 Note boards
ARM: exynos: clear prefetch bits in default l2c_aux_val
ARM: exynos: Simplify code in Exynos3250 CPU core restart path
Link: https://lore.kernel.org/r/20201201204404.22675-4-krzk@kernel.org
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Weqaar Janjua says:
====================
This patch set adds AF_XDP selftests based on veth to selftests/bpf.
# Topology:
# ---------
# -----------
# _ | Process | _
# / ----------- \
# / | \
# / | \
# ----------- | -----------
# | Thread1 | | | Thread2 |
# ----------- | -----------
# | | |
# ----------- | -----------
# | xskX | | | xskY |
# ----------- | -----------
# | | |
# ----------- | ----------
# | vethX | --------- | vethY |
# ----------- peer ----------
# | | |
# namespaceX | namespaceY
These selftests test AF_XDP SKB and Native/DRV modes using veth Virtual
Ethernet interfaces.
The test program contains two threads, each thread is single socket with
a unique UMEM. It validates in-order packet delivery and packet content
by sending packets to each other.
Prerequisites setup by script test_xsk.sh:
Set up veth interfaces as per the topology shown ^^:
* setup two veth interfaces and one namespace
** veth<xxxx> in root namespace
** veth<yyyy> in af_xdp<xxxx> namespace
** namespace af_xdp<xxxx>
* create a spec file veth.spec that includes this run-time configuration
*** xxxx and yyyy are randomly generated 4 digit numbers used to avoid
conflict with any existing interface
Adds xsk framework test to validate veth xdp DRV and SKB modes.
The following tests are provided:
1. AF_XDP SKB mode
Generic mode XDP is driver independent, used when the driver does
not have support for XDP. Works on any netdevice using sockets and
generic XDP path. XDP hook from netif_receive_skb().
a. nopoll - soft-irq processing
b. poll - using poll() syscall
c. Socket Teardown
Create a Tx and a Rx socket, Tx from one socket, Rx on another.
Destroy both sockets, then repeat multiple times. Only nopoll mode
is used
d. Bi-directional Sockets
Configure sockets as bi-directional tx/rx sockets, sets up fill
and completion rings on each socket, tx/rx in both directions.
Only nopoll mode is used
2. AF_XDP DRV/Native mode
Works on any netdevice with XDP_REDIRECT support, driver dependent.
Processes packets before SKB allocation. Provides better performance
than SKB. Driver hook available just after DMA of buffer descriptor.
a. nopoll
b. poll
c. Socket Teardown
d. Bi-directional Sockets
* Only copy mode is supported because veth does not currently support
zero-copy mode
Total tests: 8
Flow:
* Single process spawns two threads: Tx and Rx
* Each of these two threads attach to a veth interface within their
assigned namespaces
* Each thread creates one AF_XDP socket connected to a unique umem
for each veth interface
* Tx thread transmits 10k packets from veth<xxxx> to veth<yyyy>
* Rx thread verifies if all 10k packets were received and delivered
in-order, and have the right content
v2 changes:
* Move selftests/xsk to selftests/bpf
* Remove Makefiles under selftests/xsk, and utilize selftests/bpf/Makefile
v3 changes:
* merge all test scripts test_xsk_*.sh into test_xsk.sh
v4 changes:
* merge xsk_env.sh into xsk_prereqs.sh
* test_xsk.sh add cliarg -c for color-coded output
* test_xsk.sh PREREQUISITES disables IPv6 on veth interfaces
* test_xsk.sh PREREQUISITES adds xsk framework test
* test_xsk.sh is independently executable
* xdpxceiver.c Tx/Rx validates only IPv4 packets with TOS 0x9, ignores
others
====================
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Adds following tests:
1. AF_XDP SKB mode
d. Bi-directional Sockets
Configure sockets as bi-directional tx/rx sockets, sets up fill
and completion rings on each socket, tx/rx in both directions.
Only nopoll mode is used
2. AF_XDP DRV/Native mode
d. Bi-directional Sockets
* Only copy mode is supported because veth does not currently support
zero-copy mode
Signed-off-by: Weqaar Janjua <weqaar.a.janjua@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Tested-by: Yonghong Song <yhs@fb.com>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Link: https://lore.kernel.org/bpf/20201207215333.11586-6-weqaar.a.janjua@intel.com
Adds following tests:
2. AF_XDP DRV/Native mode
Works on any netdevice with XDP_REDIRECT support, driver dependent.
Processes packets before SKB allocation. Provides better performance
than SKB. Driver hook available just after DMA of buffer descriptor.
a. nopoll
b. poll
* Only copy mode is supported because veth does not currently support
zero-copy mode
Signed-off-by: Weqaar Janjua <weqaar.a.janjua@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Tested-by: Yonghong Song <yhs@fb.com>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Link: https://lore.kernel.org/bpf/20201207215333.11586-4-weqaar.a.janjua@intel.com
This patch adds AF_XDP selftests framework under selftests/bpf.
Topology:
---------
----------- -----------
| xskX | --------- | xskY |
----------- | -----------
| | |
----------- | ----------
| vethX | --------- | vethY |
----------- peer ----------
| | |
namespaceX | namespaceY
Prerequisites setup by script test_xsk.sh:
Set up veth interfaces as per the topology shown ^^:
* setup two veth interfaces and one namespace
** veth<xxxx> in root namespace
** veth<yyyy> in af_xdp<xxxx> namespace
** namespace af_xdp<xxxx>
* create a spec file veth.spec that includes this run-time configuration
*** xxxx and yyyy are randomly generated 4 digit numbers used to avoid
conflict with any existing interface
* tests the veth and xsk layers of the topology
Signed-off-by: Weqaar Janjua <weqaar.a.janjua@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Tested-by: Yonghong Song <yhs@fb.com>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Link: https://lore.kernel.org/bpf/20201207215333.11586-2-weqaar.a.janjua@intel.com
Toke Høiland-Jørgensen says:
====================
This series restores the test_offload.py selftest to working order. It seems a
number of subtle behavioural changes have crept into various subsystems which
broke test_offload.py in a number of ways. Most of these are fairly benign
changes where small adjustments to the test script seems to be the best fix,
but one is an actual kernel bug that I've observed in the wild caused by a bad
interaction between xdp_attachment_flags_ok() and the rework of XDP program
handling in the core netdev code.
Patch 1 fixes the bug by removing xdp_attachment_flags_ok(), and the reminder of
the patches are adjustments to test_offload.py, including a new feature for
netdevsim to force a BPF verification fail. Please see the individual patches
for details.
Changelog:
v4:
- Accidentally truncated the Fixes: hashes in patches 3/4 to 11 chars
v3:
- Add Fixes: tags
v2:
- Replace xdp_attachment_flags_ok() with a check in dev_xdp_attach()
- Better packing of struct nsim_dev
====================
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
A few of the tests in test_offload.py expects to see a certain number of
maps created, and checks this by counting the number of maps returned by
bpftool. There is already a filter that will remove any maps already there
at the beginning of the test, but bpftool now creates a map for the PID
iterator rodata on each invocation, which makes the map count wrong. Fix
this by also filtering the pid_iter.rodata map by name when counting.
Fixes: d53dee3fe0 ("tools/bpftool: Show info for processes holding BPF map/prog/link/btf FDs")
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Link: https://lore.kernel.org/bpf/160752226387.110217.9887866138149423444.stgit@toke.dk
When setting the ethtool feature flag fails (as expected for the test), the
kernel now tracks that the feature was requested to be 'off' and refuses to
subsequently disable it again. So reset it back to 'on' so a subsequent
disable (that's not supposed to fail) can succeed.
Fixes: 417ec26477 ("selftests/bpf: add offload test based on netdevsim")
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Link: https://lore.kernel.org/bpf/160752226280.110217.10696241563705667871.stgit@toke.dk
Since commit 6f8a57ccf8 ("bpf: Make verifier log more relevant by
default"), the verifier discards log messages for successfully-verified
programs. This broke test_offload.py which is looking for a verification
message from the driver callback. Change test_offload.py to use the toggle
in netdevsim to make the verification fail before looking for the
verification message.
Fixes: 6f8a57ccf8 ("bpf: Make verifier log more relevant by default")
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Link: https://lore.kernel.org/bpf/160752226069.110217.12370824996153348073.stgit@toke.dk
This adds a new debugfs toggle ('bpf_bind_verifier_accept') that can be
used to make netdevsim reject BPF programs from being accepted by the
verifier. If this toggle (which defaults to true) is set to false,
nsim_bpf_verify_insn() will return EOPNOTSUPP on the last
instruction (after outputting the 'Hello from netdevsim' verifier message).
This makes it possible to check the verification callback in the driver
from test_offload.py in selftests, since the verifier now clears the
verifier log on a successful load, hiding the message from the driver.
Fixes: 6f8a57ccf8 ("bpf: Make verifier log more relevant by default")
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Link: https://lore.kernel.org/bpf/160752225964.110217.12584017165318065332.stgit@toke.dk
Since commit 7f0a838254 ("bpf, xdp: Maintain info on attached XDP BPF
programs in net_device"), the XDP program attachment info is now maintained
in the core code. This interacts badly with the xdp_attachment_flags_ok()
check that prevents unloading an XDP program with different load flags than
it was loaded with. In practice, two kinds of failures are seen:
- An XDP program loaded without specifying a mode (and which then ends up
in driver mode) cannot be unloaded if the program mode is specified on
unload.
- The dev_xdp_uninstall() hook always calls the driver callback with the
mode set to the type of the program but an empty flags argument, which
means the flags_ok() check prevents the program from being removed,
leading to bpf prog reference leaks.
The original reason this check was added was to avoid ambiguity when
multiple programs were loaded. With the way the checks are done in the core
now, this is quite simple to enforce in the core code, so let's add a check
there and get rid of the xdp_attachment_flags_ok() callback entirely.
Fixes: 7f0a838254 ("bpf, xdp: Maintain info on attached XDP BPF programs in net_device")
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Link: https://lore.kernel.org/bpf/160752225751.110217.10267659521308669050.stgit@toke.dk
For SM8250, we need to write the BDF to SID mapping in PCIe controller
register space for proper working. This is accomplished by extracting
the BDF and SID values from "iommu-map" property in DT and writing those
in the register address calculated from the hash value of BDF. In case
of collisions, the index of the next entry will also be written.
For the sake of it, let's introduce a "config_sid" callback and do it
conditionally for SM8250.
Link: https://lore.kernel.org/r/20201208121402.178011-4-mani@kernel.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Bjorn Andersson <bjorn.andersson@linaro.org>
For BOs imported from outside of amdgpu, setting of amdgpu_gem_object_funcs
was missing in amdgpu_dma_buf_create_obj. Fix by refactoring BO creation
and amdgpu_gem_object_funcs setting into single function called
from both code paths.
Fixes: d693def4fd ("drm: Remove obsolete GEM and PRIME callbacks from struct drm_driver")
v2: Use use amdgpu_gem_object_create() directly
v3: fix warning
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Release dmabuf reference before returning from kfd_ioctl_import_dmabuf.
amdgpu_amdkfd_gpuvm_import_dmabuf takes a reference to the underlying
GEM BO and doesn't keep the reference to the dmabuf wrapper.
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
Without additional HostVM Latency, Renoir takes 2us longer to exit
self-refresh. This causes underflow in certain cases.
[How]
Add table for Renoir with updated sr exit latencies for WM set A.
Signed-off-by: Sung Lee <sung.lee@amd.com>
Reviewed-by: Yongqiang Sun <yongqiang.sun@amd.com>
Reviewed-by: Roman Li <Roman.Li@amd.com>
Acked-by: Eryk Brol <eryk.brol@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
At very high pixel clock, bandwidth calculation exceeds 32 bit size
and overflow value. This causes the resulting selection of link rate
to be inaccurate.
[How]
Change order of operation and use fixed point to deal with integer
accuracy. Also address bug found when forcing link rate.
Signed-off-by: Chris Park <Chris.Park@amd.com>
Reviewed-by: Wenjing Liu <Wenjing.Liu@amd.com>
Acked-by: Eryk Brol <eryk.brol@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
There is still a warning when CONFIG_DEBUG_FS is disabled:
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c:1145:13: error: 'amdgpu_ras_debugfs_create_ctrl_node' defined but not used [-Werror=unused-function]
1145 | static void amdgpu_ras_debugfs_create_ctrl_node(struct amdgpu_device *adev)
Change the code again to make the compiler actually drop
this code but not warn about it.
Fixes: ae2bf61ff3 ("drm/amdgpu: guard ras debugfs creation/removal based on CONFIG_DEBUG_FS")
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
