When the GPU powergate is controlled by a generic power domain provider,
the reset will automatically be asserted and deasserted as part of the
power-ungating procedure.
On some Jetson TX2 boards, doing an additional assert and deassert of
the GPU outside of the power-ungate procedure can cause the GPU to go
into a bad state where the memory interface can no longer access system
memory.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
When Nouveau is instantiated on top of a platform device, the dev->pdev
field will be NULL and calling pci_disable_device() will crash. Move the
PCI disabling code to the PCI specific driver removal code.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
There is no BAR2 on GP10B and there is no need to map through BAR2
because all memory is shared between the GPU and the CPU. Add a custom
implementation of the fault sub-device that uses nvkm_memory_addr()
instead of nvkm_memory_bar2() to return the address of a pinned fault
buffer.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
The sizeof is currently on args.src and args.dst and should be on
*args.src and *args.dst. Fortunately these sizes just so happen
to be the same size so it worked, however, this should be fixed
and it also cleans up static analysis warnings
Addresses-Coverity: ("sizeof not portable")
Fixes: f268307ec7 ("nouveau: simplify nouveau_dmem_migrate_vma")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
This commit is seperate from the previous one to make it easier to
revert in the future. Basically, while working on making MSTOs per-head
as opposed to per-head-per-connector I discovered these lovely issues:
https://gitlab.freedesktop.org/xorg/xserver/merge_requests/277https://gitlab.gnome.org/GNOME/mutter/issues/759
Note as well that Intel already has a temporary workaround for this in
their kernel driver. So, unfortunately we need to follow suit to avoid
causing a regression in userspace. Once these issues get fixed, this
commit should be reverted.
Signed-off-by: Lyude Paul <lyude@redhat.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Currently, for every single MST capable DRM connector we create a set of
fake encoders, one for each possible head. Unfortunately this ends up
being a huge waste of encoders. While this currently isn't causing us
any problems, it's extremely close to doing so.
The ThinkPad P71 is a good example of this. Originally when trying to
figure out why nouveau was failing to load on this laptop, I discovered
it was because nouveau was creating too many encoders. This ended up
being because we were mistakenly creating MST encoders for the eDP port,
however we are still extremely close to hitting the encoder limit on
this machine as it exposes 1 eDP port and 5 DP ports, resulting in 31
encoders.
So while this fix didn't end up being necessary to fix the P71, we still
need to implement this so that we avoid hitting the encoder limit for
valid display configurations in the event that some machine with more
connectors then this becomes available. Plus, we don't want to let good
code go to waste :)
So, use less encoders by only creating one MSTO per head. Then, attach
each new MSTC to each MSTO which corresponds to a head that it's parent
DP port is capable of using. This brings the number of encoders we
register on the ThinkPad P71 from 31, down to just 15. Yay!
Signed-off-by: Lyude Paul <lyude@redhat.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
When drm_connector_helper_funcs->atomic_best_encoder is defined,
->best_encoder is ignored by the atomic modesetting helpers. That being
said, this hook is completely broken anyway - it always returns the
first msto for a given mstc, despite the fact it might already be in
use.
So, just get rid of it. We'll need this in a moment anyway, when we make
mstos per-head as opposed to per-connector.
Changes since v1:
* Fix typo in documentation - imirkin
Signed-off-by: Lyude Paul <lyude@redhat.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
The hardware supports either size. Also add checks to ensure that only
these two sizes may be used for supplying a LUT.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Commit 5fde30a2684041f9820aa9dc4fbd0009a45076a9 in envytools modified
some of the Falcon V5 encodings, regenerate the relevant FW with this.
Also modify build rules to include SPDX header in generated files.
Tested on GM107, with no issues noted.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Some of cpuidle drivers for ARMv7 can be compile tested on this
architecture because they do not depend on mach-specific bits. Enable
compile testing for big.LITTLE, Kirkwood, Zynq, AT91, Exynos and mvebu
cpuidle drivers.
Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org>
Acked-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
SAM9X60's PMC uses different offset for master clock register.
Add a member of type struct pmc_reg_config in struct reg_config,
fill it correspondingly for SAMA5D2 and SAM9X60 and use it in
poweroff() function.
Signed-off-by: Claudiu Beznea <claudiu.beznea@microchip.com>
Acked-by: Nicolas Ferre <nicolas.ferre@microchip.com>
Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.com>
This driver uses AT91_PMC_MCKR in poweroff() function. But the
SAM9X60's PMC versions maps AT91_PMC_MCKR functionality at different
offset compared to the SAMA5D2's one. This patch prepares the field
so that different AT91_PMC_MCKR's offsets to be introduced in
struct reg_config so that proper offset to be used for AT91_PMC_MCKR
based on compatible string.
Signed-off-by: Claudiu Beznea <claudiu.beznea@microchip.com>
Acked-by: Nicolas Ferre <nicolas.ferre@microchip.com>
Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.com>
Add bq25892, bq25895 and bq25896 to list of supported device IDs for
DeviceTree and I2C.
Signed-off-by: Yauhen Kharuzhy <jekhor@gmail.com>
Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.com>
Add bq25892, bq25895 and bq25896 to list of supported device IDs in the
bq25890 DT binding document.
Signed-off-by: Yauhen Kharuzhy <jekhor@gmail.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.com>
Support BQ25892 and BQ25896 chips by this driver. They shared one chip
ID 0, so distinquish them by device revisions (2 for 25896 and 1 for
25892).
Signed-off-by: Yauhen Kharuzhy <jekhor@gmail.com>
Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.com>
With CONFIG_PROVE_RCU_LIST, I had many suspicious RCU warnings
when I ran ftracetest trigger testcases.
-----
# dmesg -c > /dev/null
# ./ftracetest test.d/trigger
...
# dmesg | grep "RCU-list traversed" | cut -f 2 -d ] | cut -f 2 -d " "
kernel/trace/trace_events_hist.c:6070
kernel/trace/trace_events_hist.c:1760
kernel/trace/trace_events_hist.c:5911
kernel/trace/trace_events_trigger.c:504
kernel/trace/trace_events_hist.c:1810
kernel/trace/trace_events_hist.c:3158
kernel/trace/trace_events_hist.c:3105
kernel/trace/trace_events_hist.c:5518
kernel/trace/trace_events_hist.c:5998
kernel/trace/trace_events_hist.c:6019
kernel/trace/trace_events_hist.c:6044
kernel/trace/trace_events_trigger.c:1500
kernel/trace/trace_events_trigger.c:1540
kernel/trace/trace_events_trigger.c:539
kernel/trace/trace_events_trigger.c:584
-----
I investigated those warnings and found that the RCU-list
traversals in event trigger and hist didn't need to use
RCU version because those were called only under event_mutex.
I also checked other RCU-list traversals related to event
trigger list, and found that most of them were called from
event_hist_trigger_func() or hist_unregister_trigger() or
register/unregister functions except for a few cases.
Replace these unneeded RCU-list traversals with normal list
traversal macro and lockdep_assert_held() to check the
event_mutex is held.
Link: http://lkml.kernel.org/r/157680910305.11685.15110237954275915782.stgit@devnote2
Cc: stable@vger.kernel.org
Fixes: 30350d65ac ("tracing: Add variable support to hist triggers")
Reviewed-by: Tom Zanussi <zanussi@kernel.org>
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Pull NFS client bugfixes from Anna Schumaker:
"Three NFS over RDMA fixes for bugs Chuck found that can be hit during
device removal:
- Fix create_qp crash on device unload
- Fix completion wait during device removal
- Fix oops in receive handler after device removal"
* tag 'nfs-for-5.5-2' of git://git.linux-nfs.org/projects/anna/linux-nfs:
xprtrdma: Fix oops in Receive handler after device removal
xprtrdma: Fix completion wait during device removal
xprtrdma: Fix create_qp crash on device unload
When initializing an fs-verity hash algorithm, also initialize a mempool
that contains a single preallocated hash request object. Then replace
the direct calls to ahash_request_alloc() and ahash_request_free() with
allocating and freeing from this mempool.
This eliminates the possibility of the allocation failing, which is
desirable for the I/O path.
This doesn't cause deadlocks because there's no case where multiple hash
requests are needed at a time to make forward progress.
Link: https://lore.kernel.org/r/20191231175545.20709-1-ebiggers@kernel.org
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Eric Biggers <ebiggers@google.com>
When fs-verity verifies data pages, currently it reads each Merkle tree
page synchronously using read_mapping_page().
Therefore, when the Merkle tree pages aren't already cached, fs-verity
causes an extra 4 KiB I/O request for every 512 KiB of data (assuming
that the Merkle tree uses SHA-256 and 4 KiB blocks). This results in
more I/O requests and performance loss than is strictly necessary.
Therefore, implement readahead of the Merkle tree pages.
For simplicity, we take advantage of the fact that the kernel already
does readahead of the file's *data*, just like it does for any other
file. Due to this, we don't really need a separate readahead state
(struct file_ra_state) just for the Merkle tree, but rather we just need
to piggy-back on the existing data readahead requests.
We also only really need to bother with the first level of the Merkle
tree, since the usual fan-out factor is 128, so normally over 99% of
Merkle tree I/O requests are for the first level.
Therefore, make fsverity_verify_bio() enable readahead of the first
Merkle tree level, for up to 1/4 the number of pages in the bio, when it
sees that the REQ_RAHEAD flag is set on the bio. The readahead size is
then passed down to ->read_merkle_tree_page() for the filesystem to
(optionally) implement if it sees that the requested page is uncached.
While we're at it, also make build_merkle_tree_level() set the Merkle
tree readahead size, since it's easy to do there.
However, for now don't set the readahead size in fsverity_verify_page(),
since currently it's only used to verify holes on ext4 and f2fs, and it
would need parameters added to know how much to read ahead.
This patch significantly improves fs-verity sequential read performance.
Some quick benchmarks with 'cat'-ing a 250MB file after dropping caches:
On an ARM64 phone (using sha256-ce):
Before: 217 MB/s
After: 263 MB/s
(compare to sha256sum of non-verity file: 357 MB/s)
In an x86_64 VM (using sha256-avx2):
Before: 173 MB/s
After: 215 MB/s
(compare to sha256sum of non-verity file: 223 MB/s)
Link: https://lore.kernel.org/r/20200106205533.137005-1-ebiggers@kernel.org
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Eric Biggers <ebiggers@google.com>
When it builds the first level of the Merkle tree, FS_IOC_ENABLE_VERITY
sequentially reads each page of the file using read_mapping_page().
This works fine if the file's data is already in pagecache, which should
normally be the case, since this ioctl is normally used immediately
after writing out the file.
But in any other case this implementation performs very poorly, since
only one page is read at a time.
Fix this by implementing readahead using the functions from
mm/readahead.c.
This improves performance in the uncached case by about 20x, as seen in
the following benchmarks done on a 250MB file (on x86_64 with SHA-NI):
FS_IOC_ENABLE_VERITY uncached (before) 3.299s
FS_IOC_ENABLE_VERITY uncached (after) 0.160s
FS_IOC_ENABLE_VERITY cached 0.147s
sha256sum uncached 0.191s
sha256sum cached 0.145s
Note: we could instead switch to kernel_read(). But that would mean
we'd no longer be hashing the data directly from the pagecache, which is
a nice optimization of its own. And using kernel_read() would require
allocating another temporary buffer, hashing the data and tree pages
separately, and explicitly zero-padding the last page -- so it wouldn't
really be any simpler than direct pagecache access, at least for now.
Link: https://lore.kernel.org/r/20200106205410.136707-1-ebiggers@kernel.org
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Just as commit 0566e40ce7 ("tracing: initcall: Ordered comparison of
function pointers"), this patch fixes another remaining one in xen.h
found by clang-9.
In file included from arch/x86/xen/trace.c:21:
In file included from ./include/trace/events/xen.h:475:
In file included from ./include/trace/define_trace.h:102:
In file included from ./include/trace/trace_events.h:473:
./include/trace/events/xen.h:69:7: warning: ordered comparison of function \
pointers ('xen_mc_callback_fn_t' (aka 'void (*)(void *)') and 'xen_mc_callback_fn_t') [-Wordered-compare-function-pointers]
__field(xen_mc_callback_fn_t, fn)
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
./include/trace/trace_events.h:421:29: note: expanded from macro '__field'
^
./include/trace/trace_events.h:407:6: note: expanded from macro '__field_ext'
is_signed_type(type), filter_type); \
^
./include/linux/trace_events.h:554:44: note: expanded from macro 'is_signed_type'
^
Fixes: c796f213a6 ("xen/trace: add multicall tracing")
Signed-off-by: Changbin Du <changbin.du@gmail.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Currently fscrypt_zeroout_range() issues and waits on a bio for each
block it writes, which makes it very slow.
Optimize it to write up to 16 pages at a time instead.
Also add a function comment, and improve reliability by allowing the
allocations of the bio and the first ciphertext page to wait on the
corresponding mempools.
Link: https://lore.kernel.org/r/20191226160813.53182-1-ebiggers@kernel.org
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Eric Biggers <ebiggers@google.com>
We've had generic code handling module sysconfig and OCP reset registers
for omap variants for many years now and all the drivers really needs to
do is just call runtime PM functions.
Looks like the omap-hdq driver got only partially updated over the years
to use runtime PM, and still has lots of custom PM code left.
We can replace all the custom code for sysconfig, OCP reset, and PM with
just a few lines of runtime PM autosuspend code.
In order to set the device mode properly when pm_runtime_get_sync() is
called during probe, we need to also move parsing of "ti,mode" to happen
earlier before we call pm_runtime_enable().
Since we now disable interrupts lazily in omap_hdq_runtime_suspend(), we
must remove the call to hdq_disable_interrupt() in omap_w1_read_byte().
And we must clear irqstatus calling wait_event_timeout() on it, so let's
add hdq_reset_irqstatus() for that.
Note that the earlier driver specific usage count limit of four seems
completely artificial and should not be an issue in normal use.
Cc: Adam Ford <aford173@gmail.com>
Cc: Andrew F. Davis <afd@ti.com>
Cc: Andreas Kemnade <andreas@kemnade.info>
Cc: H. Nikolaus Schaller <hns@goldelico.com>
Cc: Vignesh R <vigneshr@ti.com>
Tested-by: Andreas Kemnade <andreas@kemnade.info> # gta04
Tested-by: Adam Ford <aford173@gmail.com> #logicpd-torpedo-37xx-devkit
Signed-off-by: Tony Lindgren <tony@atomide.com>
Link: https://lore.kernel.org/r/20191217004048.46298-1-tony@atomide.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently this driver is loaded if the DMI string matches coreboot
and has a proper smi_command in the ACPI FADT table, but a GSMI handler in
SMM is an optional feature in coreboot.
So probe for a SMM GSMI handler before initializing the driver.
If the smihandler leaves the calling argument in %eax in the SMM save state
untouched that generally means the is no handler for GSMI.
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Signed-off-by: Patrick Rudolph <patrick.rudolph@9elements.com>
Link: https://lore.kernel.org/r/20191118101934.22526-4-patrick.rudolph@9elements.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
With CONFIG_CC_OPTIMIZE_FOR_PERFORMANCE_O3, the stack usage in vme_fake
grows above the warning limit:
drivers/vme/bridges/vme_fake.c: In function 'fake_master_read':
drivers/vme/bridges/vme_fake.c:610:1: error: the frame size of 1160 bytes is larger than 1024 bytes [-Werror=frame-larger-than=]
drivers/vme/bridges/vme_fake.c: In function 'fake_master_write':
drivers/vme/bridges/vme_fake.c:797:1: error: the frame size of 1160 bytes is larger than 1024 bytes [-Werror=frame-larger-than=]
The problem is that in some configurations, each call to
fake_vmereadX() puts another variable on the stack.
Reduce the amount of inlining to get back to the previous state,
with no function using more than 200 bytes each.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Link: https://lore.kernel.org/r/20200107200610.3482901-1-arnd@arndb.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit 429120f3df starts to take account of segment's start dma address
when computing max segment size, and data type of 'unsigned long'
is used to do that. However, the segment mask may be 0xffffffff, so
the figured out segment size may be overflowed in case of zero physical
address on 32bit arch.
Fix the issue by returning queue_max_segment_size() directly when that
happens.
Fixes: 429120f3df ("block: fix splitting segments on boundary masks")
Reported-by: Guenter Roeck <linux@roeck-us.net>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Cc: Christoph Hellwig <hch@lst.de>
Tested-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Currently, hv_sock restricts the port the guest socket can accept
connections on. hv_sock divides the socket port namespace into two parts
for server side (listening socket), 0-0x7FFFFFFF & 0x80000000-0xFFFFFFFF
(there are no restrictions on client port namespace). The first part
(0-0x7FFFFFFF) is reserved for sockets where connections can be accepted.
The second part (0x80000000-0xFFFFFFFF) is reserved for allocating ports
for the peer (host) socket, once a connection is accepted.
This reservation of the port namespace is specific to hv_sock and not
known by the generic vsock library (ex: af_vsock). This is problematic
because auto-binds/ephemeral ports are handled by the generic vsock
library and it has no knowledge of this port reservation and could
allocate a port that is not compatible with hv_sock (and legitimately so).
The issue hasn't surfaced so far because the auto-bind code of vsock
(__vsock_bind_stream) prior to the change 'VSOCK: bind to random port for
VMADDR_PORT_ANY' would start walking up from LAST_RESERVED_PORT (1023) and
start assigning ports. That will take a large number of iterations to hit
0x7FFFFFFF. But, after the above change to randomize port selection, the
issue has started coming up more frequently.
There has really been no good reason to have this port reservation logic
in hv_sock from the get go. Reserving a local port for peer ports is not
how things are handled generally. Peer ports should reflect the peer port.
This fixes the issue by lifting the port reservation, and also returns the
right peer port. Since the code converts the GUID to the peer port (by
using the first 4 bytes), there is a possibility of conflicts, but that
seems like a reasonable risk to take, given this is limited to vsock and
that only applies to all local sockets.
Signed-off-by: Sunil Muthuswamy <sunilmut@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jason A. Donenfeld says:
====================
skb_list_walk_safe refactoring for net/*'s skb_gso_segment usage
This patchset adjusts all return values of skb_gso_segment in net/* to
use the new skb_list_walk_safe helper.
First we fix a minor bug in the helper macro that didn't come up in the
last patchset's uses. Then we adjust several cases throughout net/. The
xfrm changes were a bit hairy, but doable. Reading and thinking about
the code in mac80211 indicates a memory leak, which the commit
addresses. All the other cases were pretty trivial.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
This is a conversion case for the new function, keeping the flow of the
existing code as intact as possible. We also switch over to using
skb_mark_not_on_list instead of a null write to skb->next.
Finally, this code appeared to have a memory leak in the case where
header building fails before the last gso segment. In that case, the
remaining segments are not freed. So this commit also adds the proper
kfree_skb_list call for the remainder of the skbs.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is a straight-forward conversion case for the new function, keeping
the flow of the existing code as intact as possible.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is a straight-forward conversion case for the new function, keeping
the flow of the existing code as intact as possible.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is a straight-forward conversion case for the new function, keeping
the flow of the existing code as intact as possible.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
