While debugging a situation where a delta for an event was calucalted wrong,
I realize there was nothing making sure that the delta of events are
correct. If a single event has an incorrect delta, then all events after it
will also have one. If the discrepency gets large enough, it could cause
the time stamps to go backwards when crossing sub buffers, that record a
full 64 bit time stamp, and the new deltas are added to that.
Add a way to validate the events at most events and when crossing a buffer
page. This will help make sure that the deltas are always correct. This test
will detect if they are ever corrupted.
The test adds a high overhead to the ring buffer recording, as it does the
audit for almost every event, and should only be used for testing the ring
buffer.
This will catch the bug that is fixed by commit 55ea4cf403 ("ring-buffer:
Update write stamp with the correct ts"), which is not applied when this
commit is applied.
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
On chip versions supporting tally counter reset we currently update
the counters after a reset although we know all counters are zero.
Skip this unnecessary step.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Link: https://lore.kernel.org/r/526618b2-b1bf-1844-b82a-dab2df7bdc8f@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Alex Elder says:
====================
net: ipa: IPA v4.5 aggregation and Qtime
This series updates some IPA register definitions that change in
substantive ways for IPA v4.5.
One register defines parameters used by an endpoint to aggregate
multiple packets into a buffer. The size and position of most
fields in that register have changed with this new hardware version,
and consequently the function that programs it needs to be done a
bit differently. The first patch takes care of this.
Second, IPA v4.5 introduces a unified time keeping component to be
used in several places by the IPA hardware. A main clock divider
provides a fundamental tick rate, and several timestamped features
now define their granularity based on that. There is also a set of
"pulse generators" derived from the main tick, and these are used
to implement timers used for aggregation and head-of-line block
avoidance. The second patch adds IPA register updates to support
Qtime along with its configuration, and the last two patches
configure the timers that use it.
====================
Link: https://lore.kernel.org/r/20201130233712.29113-1-elder@linaro.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Extend ipa_reg_init_hol_block_timer_val() so it properly calculates
the head-of-line block timeout to use for IPA v4.5.
Introduce hol_block_timer_qtime_val() to compute the value to use
for IPA v4.5, where Qtime is used as the basis of the timer. Call
that function from hol_block_timer_val() for IPA v4.5.
Both of these are private functions, so shorten their names a bit so
they don't take up so much space on the line.
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Change aggr_time_limit_encoded() to properly calculate the
aggregation time limit to use for IPA v4.5.
Older IPA versions program the AGGR_GRANULARITY field of the
of the COUNTER_CFG register to set the granularity of the
aggregation timer, which we configure to be 500 microseconds.
Instead, IPA v4.5 selects between two possible granularity values
derived from the 19.2 MHz Qtime clock. These granularities are
100 microseconds or 1 millisecond per tick. We use the smaller
granularity if possible, unless the desired period is too large
to be specified that way.
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
IPA v4.5 introduces a new unified timer architecture driven on the
19.2 MHz SoC crystal oscillator (XO). It is independent of the IPA
core clock and avoids some duplication.
Lower-resolution time stamps are derived from this by using only the
high-order bits of the 19.2 MHz Qtime clock. And timers are derived
from this based on "pulse generators" configured to fire at a fixed
rate based on the Qtime clock.
This patch introduces ipa_qtime_config(), which configures the Qtime
mechanism for use. It also adds to the IPA register definitions
related to timers and time stamping.
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
IPA v4.5 significantly changes the format of the configuration
register used for endpoint aggregation. The AGGR_BYTE_LIMIT field
is now larger, and the positions of other fields are shifted. This
complicates the way we have to access this register because functions
like u32_encode_bits() require their field mask argument to be constant.
A further complication is that we want to know the maximum value
representable by at least one of these fields, and that too requires
a constant field mask.
This patch adds support for IPA v4.5 endpoint aggregation registers
in a way that continues to support "legacy" IPA hardware. It does
so in a way that keeps field masks constant.
First, for each variable field mask, we define an inline function
whose return value is either the legacy value or the IPA v4.5 value.
Second, we define functions for these fields that encode a value
to use in each field based on the IPA version (this approach is
already used elsewhere). The field mask provided is supplied by
the function mentioned above.
Finally, for the aggregation byte limit fields where we want to
know the maximum representable value, we define a function that
returns that maximum, computed from the appropriate field mask.
We can no longer verify at build time that our buffer size is
in the range that can be represented by the aggregation byte
limit field. So remove the test done by a BUILD_BUG_ON() call
in ipa_endpoint_validate_build(), and implement a comparable check
at the top of ipa_endpoint_data_valid().
Doing that makes ipa_endpoint_validate_build() contain a single
line BUILD_BUG_ON() call, so just remove that function and move
the remaining line into ipa_endpoint_data_valid().
One final note: the aggregation time limit value for IPA v4.5 needs
to be computed differently. That is handled in an upcoming patch.
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Karsten Graul says:
====================
net/smc: Add support for generic netlink API
Up to version 4 this patch series was using the sock_diag netlink
infrastructure. This version is using the generic netlink API. Generic
netlink API offers a better type safety between kernel and userspace
communication.
Using the generic netlink API the smc module can now provide information
about SMC linkgroups, links and devices (both for SMC-R and SMC-D).
v2: Add missing include to uapi header smc_diag.h.
v3: Apply code style recommendations from review comments.
Instead of using EXPORTs to allow the smc_diag module to access
data of the smc module, introduce struct smc_diag_ops and let
smc_diag access the required data using function pointers.
v4: Address checkpatch.pl warnings. Do not use static inline for
functions.
v5: Use generic netlink API instead of the sock_diag netlink
infrastructure.
v6: Integrate more review comments from Jakub.
v7: Use nla_nest_start() with the new family. Use .maxattr=1 in the
genl family and define one entry for attribute 1 in the policy to
reject this attritbute for all commands. All other possible attributes
are rejected because NL_VALIDATE_STRICT is set for the policy
implicitely, which includes NL_VALIDATE_MAXTYPE.
Setting policy[0].strict_start_type=1 does not work here because there
is no valid attribute defined for this family, only plain commands. For
any type > maxtype (which is .maxattr) validate_nla() would return 0 to
userspace instead of -EINVAL. What helps here is __nla_validate_parse()
which checks for type > maxtype and returns -EINVAL when NL_VALIDATE_MAXTYPE
is set. This requires the one entry for type == .maxattr with
.type = NLA_REJECT in the nla_policy.
When a future command wants to allow attributes then it can easily specify a
dedicated .policy for this new command in the genl_ops array. This dedicated
policy overlays the global policy specified in the genl_family structure.
====================
Link: https://lore.kernel.org/r/20201201192049.53517-1-kgraul@linux.ibm.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Introduce get link command which loops through
all available links of all available link groups. It
uses the SMC-R linkgroup list as entry point, not
the socket list, which makes linkgroup diagnosis
possible, in case linkgroup does not contain active
connections anymore.
Signed-off-by: Guvenc Gulce <guvenc@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Introduce get linkgroup command which loops through
all available SMCR linkgroups. It uses the SMC-R linkgroup
list as entry point, not the socket list, which makes
linkgroup diagnosis possible, in case linkgroup does not
contain active connections anymore.
Signed-off-by: Guvenc Gulce <guvenc@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add new netlink command to obtain system information
of the smc module.
Signed-off-by: Guvenc Gulce <guvenc@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Encapsulate the smc ism v2 capability boolean value
in a function for better information hiding.
Signed-off-by: Guvenc Gulce <guvenc@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
During link creation add net-device ifindex and ib-device
name to link structure. This is needed for diagnostic purposes.
When diagnostic information is gathered, we need to traverse
device, linkgroup and link structures, to be able to do that
we need to hold a spinlock for the linkgroup list, without this
diagnostic information in link structure, another device list
mutex holding would be necessary to dereference the device
pointer in the link structure which would be impossible when
holding a spinlock already.
Signed-off-by: Guvenc Gulce <guvenc@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
During smc ib-device creation, add network device ifindex to smc
ib-device structure. Register for netdevice changes and update ib-device
accordingly. This is needed for diagnostic purposes.
Signed-off-by: Guvenc Gulce <guvenc@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add link counters to the structure of the smc ib device, one counter per
ib port. Increase/decrease the counters as needed in the corresponding
routines.
Signed-off-by: Guvenc Gulce <guvenc@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add connection counters to the structure of the link.
Increase/decrease the counters as needed in the corresponding
routines.
Signed-off-by: Guvenc Gulce <guvenc@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Use active link of the connection directly and not
via linkgroup array structure when obtaining link
data of the connection.
Signed-off-by: Guvenc Gulce <guvenc@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The helper smc_connect_abort() can be used by the listen processing
functions, too. And rename this helper to smc_conn_abort() to make the
purpose clearer.
No functional change.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Recent changes made to remove AES constants started using protocol
aware salt_size. ctx->prot_info's salt_size is filled in tls sw case,
but not in tls offload mode, and was working so far because of the
hard coded value was used.
Fixes: 6942a284fb ("net/tls: make inline helpers protocol-aware")
Signed-off-by: Rohit Maheshwari <rohitm@chelsio.com>
Link: https://lore.kernel.org/r/20201201090752.27355-1-rohitm@chelsio.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Now that the driver can be enabled by COMPILE_TEST, we see warnings on
64bit platforms when void pointers are cast to unsigned int (and
vice versa).
warning: cast to smaller integer type 'unsigned int' from 'void *'
unsigned int chan = (unsigned int)link->con_priv;
...
warning: cast to 'void *' from smaller integer type 'unsigned int'
ipcc->controller.chans[i].con_priv = (void *)i;
Update these casts to use unsigned long variables, which are the same
size as pointers on all platforms.
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Martin Kaiser <martin@kaiser.cx>
Reviewed-by: Fabien Dessenne <fabien.dessenne@st.com>
Signed-off-by: Jassi Brar <jaswinder.singh@linaro.org>
platform_get_irq_byname already prints an error message if the requested
irq was not found. Don't print another message in the driver.
Signed-off-by: Martin Kaiser <martin@kaiser.cx>
Reviewed-by: Fabien Dessenne <fabien.dessenne@st.com>
Signed-off-by: Jassi Brar <jaswinder.singh@linaro.org>
This allows compiling the driver on architectures where the hardware is not
available. Most other mailbox drivers support this as well.
Signed-off-by: Martin Kaiser <martin@kaiser.cx>
Reviewed-by: Fabien Dessenne <fabien.dessenne@st.com>
Signed-off-by: Jassi Brar <jaswinder.singh@linaro.org>
The mhu_db_channel info is allocated per channel using devm_kzalloc from
mhu_db_mbox_xlate which gets called from mbox_request_channel. However
we are releasing the allocated mhu_db_channel info using plain kfree from
mhu_db_shutdown which is called from mbox_free_channel.
This leads to random crashes when the channel is freed like below one:
Unable to handle kernel paging request at virtual address 0080000400000008
[0080000400000008] address between user and kernel address ranges
Internal error: Oops: 96000044 [#1] PREEMPT SMP
Modules linked in: scmi_module(-)
CPU: 1 PID: 2212 Comm: rmmod Not tainted 5.10.0-rc5 #31
Hardware name: ARM LTD ARM Juno Development Platform/ARM Juno
Development Platform, BIOS EDK II Nov 19 2020
pstate: 20000085 (nzCv daIf -PAN -UAO -TCO BTYPE=--)
pc : release_nodes+0x74/0x230
lr : devres_release_all+0x40/0x68
Call trace:
release_nodes+0x74/0x230
devres_release_all+0x40/0x68
device_release_driver_internal+0x12c/0x1f8
driver_detach+0x58/0xe8
bus_remove_driver+0x64/0xe0
driver_unregister+0x38/0x68
platform_driver_unregister+0x1c/0x28
scmi_driver_exit+0x38/0x44 [scmi_module]
__arm64_sys_delete_module+0x188/0x260
el0_svc_common.constprop.0+0x80/0x1a8
do_el0_svc+0x2c/0x98
el0_sync_handler+0x160/0x168
el0_sync+0x174/0x180
Code: 1400000d eb07009f 54000460 f9400486 (f90004a6)
---[ end trace c55ffd306c140233 ]---
Fix it by replacing kfree with devm_kfree as required.
Fixes: 7002ca237b ("mailbox: arm_mhu: Add ARM MHU doorbell driver")
Reported-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Tested-by: Cristian Marussi <cristian.marussi@arm.com>
Reviewed-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Jassi Brar <jaswinder.singh@linaro.org>
When adding support for propagating ECT(1) marking in IP headers it seems I
suffered from endianness-confusion in the checksum update calculation: In
fact the ECN field is in the *lower* bits of the first 16-bit word of the
IP header when calculating in network byte order. This means that the
addition performed to update the checksum field was wrong; let's fix that.
Fixes: b723748750 ("tunnel: Propagate ECT(1) when decapsulating as recommended by RFC6040")
Reported-by: Jonathan Morton <chromatix99@gmail.com>
Tested-by: Pete Heist <pete@heistp.net>
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Link: https://lore.kernel.org/r/20201130183705.17540-1-toke@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Order of operations is slightly more correct in the driver
to change the netdev->mtu after the queues have been stopped
rather than before.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Remove memory allocation fail messages where the OOM stack
trace will make it obvious which allocation request failed.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Use dma_set_mask_and_coherent() instead of pci_set_dma_mask() followed by
a pci_set_consistent_dma_mask().
Link: https://lore.kernel.org/r/20201201091811.37984-1-galpress@amazon.com
Reviewed-by: Firas JahJah <firasj@amazon.com>
Reviewed-by: Yossi Leybovich <sleybo@amazon.com>
Signed-off-by: Gal Pressman <galpress@amazon.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
The variable 'cnt' is used to represent the max number of sge an SQ WQE
can use at first, then it means how many extended sge an SQ has. In
addition, this function has no need to return a value. So refactor and
encapsulate the parts of getting number of extended sge a WQE can use to
make it easier to understand.
Link: https://lore.kernel.org/r/1606558959-48510-4-git-send-email-liweihang@huawei.com
Signed-off-by: Weihang Li <liweihang@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Page alignment is required when setting the number of extended sge
according to the hardware's achivement. If the space of needed extended
sge is greater than one page, the roundup_pow_of_two() can ensure
that. But if the needed extended sge isn't 0 and can not be filled in a
whole page, the driver should align it specifically.
Fixes: 54d6638765 ("RDMA/hns: Optimize WQE buffer size calculating process")
Link: https://lore.kernel.org/r/1606558959-48510-3-git-send-email-liweihang@huawei.com
Signed-off-by: Yangyang Li <liyangyang20@huawei.com>
Signed-off-by: Weihang Li <liweihang@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
One RC SQ WQE can store 2 sges but UD can't, so ignore 2 valid sges of
wr.sglist for RC which have been filled in WQE before setting extended
sge. Either of RC and UD can not contain 0-length sges, so these 0-length
sges should be skipped.
Fixes: 54d6638765 ("RDMA/hns: Optimize WQE buffer size calculating process")
Link: https://lore.kernel.org/r/1606558959-48510-2-git-send-email-liweihang@huawei.com
Signed-off-by: Lang Cheng <chenglang@huawei.com>
Signed-off-by: Weihang Li <liweihang@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Julian Wiedmann says:
====================
s390/ctcm: updates 2020-11-30
Some rare ctcm updates by Sebastian, who cleans up all places where
in_interrupt() was used to determine the correct GFP_* mask for
allocations.
In the first three patches we can get rid of those allocations entirely,
as they just end up being copied into the skb.
====================
Link: https://lore.kernel.org/r/20201130100950.42051-1-jwi@linux.ibm.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
gfp_type() uses in_interrupt() to figure out the correct GFP mask.
The usage of in_interrupt() in drivers is phased out and Linus clearly
requested that code which changes behaviour depending on context should
either be separated or the context be conveyed in an argument passed by the
caller, which usually knows the context.
ctcmpc_tx() is used as net_device_ops::ndo_start_xmit. This callback is
invoked with disabled bottom halves.
Use GFP_ATOMIC for memory allocation in ctcmpc_tx().
Remove gfp_type() since the last user is gone.
Reviewed-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
gfp_type() uses in_interrupt() to figure out the correct GFP mask.
The usage of in_interrupt() in drivers is phased out and Linus clearly
requested that code which changes behaviour depending on context should
either be separated or the context be conveyed in an argument passed by the
caller, which usually knows the context.
The memory allocation of `ch' a few lines above is using GFP_KERNEL,
also an allocation a few lines later is using GFP_KERNEL.
Use GFP_KERNEL for the memory allocation.
Reviewed-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
gfp_type() uses in_interrupt() to figure out the correct GFP mask.
The usage of in_interrupt() in drivers is phased out and Linus clearly
requested that code which changes behaviour depending on context should
either be separated or the context be conveyed in an argument passed by the
caller, which usually knows the context.
The call chain of ctcmpc_unpack_skb():
ctcmpc_bh()
-> ctcmpc_unpack_skb()
ctcmpc_bh() is a tasklet handler so GFP_ATOMIC is needed.
Use GFP_ATOMIC as allocation type in ctcmpc_unpack_skb().
Reviewed-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The size of struct pdu is 8 byte. The memory is allocated, initialized,
used and deallocated a few lines later.
It is more efficient to avoid the allocation/free dance and assign the
values directly to skb's data part instead of using memcpy() for it.
Avoid an allocation of struct pdu and use the resulting skb pointer
instead.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
[jwi: Fix-up the pdu_offset, adjust skb->len for the pushed length.
Reflow the commit msg.]
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The size of struct qllc is 2 byte. The memory for is allocated,
initialized, used and deallocated a few lines later.
It is more efficient to avoid the allocation/free dance and assign the
values directly to skb's data part instead of using memcpy() for it.
Avoid an allocation of struct qllc and use the resulting skb pointer
instead.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
[jwi: remove a newline, reflow the commit msg]
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The size of struct th_header is 8 byte and the size of struct th_sweep
is 16 byte. The memory for is allocated, initialized, used and
deallocated a few lines later.
It is more efficient to avoid the allocation/free dance and assign the
values directly to skb's data part instead of using memcpy() for it.
Avoid an allocation of struct th_sweep/th_header and use the resulting
skb pointer instead.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
[jwi: use skb_put_zero(), instead of skb_put() + memset to 0]
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The local variables cur_state and new_state hold the state that should be
used for the modify QP operation instead of the ones in the ib_qp_attr
struct.
Fixes: 40909f664d ("RDMA/efa: Add EFA verbs implementation")
Link: https://lore.kernel.org/r/20201201091724.37016-1-galpress@amazon.com
Reviewed-by: Firas JahJah <firasj@amazon.com>
Reviewed-by: Yossi Leybovich <sleybo@amazon.com>
Signed-off-by: Gal Pressman <galpress@amazon.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
The push feature does not work as expected in x722 and has historically
been disabled in the driver.
Purge all remaining code related to the push feature in i40iw.
Link: https://lore.kernel.org/r/20201125005616.1800-3-shiraz.saleem@intel.com
Signed-off-by: Shiraz Saleem <shiraz.saleem@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
-----BEGIN PGP SIGNATURE-----
iQFSBAABCAA8FiEEq68RxlopcLEwq+PEeb4+QwBBGIYFAl/EM9oeHHRvcnZhbGRz
QGxpbnV4LWZvdW5kYXRpb24ub3JnAAoJEHm+PkMAQRiG/3kH/RNkFyTlHlUkZpJx
8Ks2yWgUln7YhZcmOaG/IcIyWnhCgo3l35kiaH7XxM+rPMZzidp51MHUllaTAQDc
u+5EFHMJsmTWUfE8ocHPb1cPdYEDSoVr6QUsixbL9+uADpRz+VZVtWMb89EiyMrC
wvLIzpnqY5UNriWWBxD0hrmSsT4g9XCsauer4k2KB+zvebwg6vFOMCFLFc2qz7fb
ABsrPFqLZOMp+16chGxyHP7LJ6ygI/Hwf7tPW8ppv4c+hes4HZg7yqJxXhV02QbJ
s10s6BTcEWMqKg/T6L/VoScsMHWUcNdvrr3uuPQhgup240XdmB1XO8rOKddw27e7
VIjrjNw=
=4ZaP
-----END PGP SIGNATURE-----
Merge tag 'v5.10-rc6' into rdma.git for-next
For dependencies in following patches
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
If an OS has not been granted AER control via _OSC, it should not make
changes to PCI_ERR_ROOT_COMMAND and PCI_ERR_ROOT_STATUS related registers.
Per section 4.5.1 of the System Firmware Intermediary (SFI) _OSC and DPC
Updates ECN [1], this bit also covers these aspects of the PCI Express
Advanced Error Reporting. Based on the above and earlier discussion [2],
make the following changes:
Add a check for the native case (i.e., AER control via _OSC)
Note that the previous "clear, reset, enable" order suggests that the reset
might cause errors that we should ignore. After this commit, those errors
(if any) will remain logged in the PCI_ERR_ROOT_STATUS register.
[1] System Firmware Intermediary (SFI) _OSC and DPC Updates ECN, Feb 24,
2020, affecting PCI Firmware Specification, Rev. 3.2
https://members.pcisig.com/wg/PCI-SIG/document/14076
[2] https://lore.kernel.org/linux-pci/20201020162820.GA370938@bjorn-Precision-5520/
Link: https://lore.kernel.org/r/20201121001036.8560-2-sean.v.kelley@intel.com
Tested-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> # non-native/no RCEC
Signed-off-by: Sean V Kelley <sean.v.kelley@intel.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
