POWER8 and POWER9 machines have a hardware deviation where generation
of a hypervisor decrementer exception is suppressed if the HDICE bit
in the LPCR register is 0 at the time when the HDEC register
decrements from 0 to -1. When entering a guest, KVM first writes the
HDEC register with the time until it wants the CPU to exit the guest,
and then writes the LPCR with the guest value, which includes
HDICE = 1. If HDEC decrements from 0 to -1 during the interval
between those two events, it is possible that we can enter the guest
with HDEC already negative but no HDEC exception pending, meaning that
no HDEC interrupt will occur while the CPU is in the guest, or at
least not until HDEC wraps around. Thus it is possible for the CPU to
keep executing in the guest for a long time; up to about 4 seconds on
POWER8, or about 4.46 years on POWER9 (except that the host kernel
hard lockup detector will fire first).
To fix this, we set the LPCR[HDICE] bit before writing HDEC on guest
entry.
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
The current nested KVM code does not support HPT guests. This is
informed/enforced in some ways:
- Hosts < P9 will not be able to enable the nested HV feature;
- The nested hypervisor MMU capabilities will not contain
KVM_CAP_PPC_MMU_HASH_V3;
- QEMU reflects the MMU capabilities in the
'ibm,arch-vec-5-platform-support' device-tree property;
- The nested guest, at 'prom_parse_mmu_model' ignores the
'disable_radix' kernel command line option if HPT is not supported;
- The KVM_PPC_CONFIGURE_V3_MMU ioctl will fail if trying to use HPT.
There is, however, still a way to start a HPT guest by using
max-compat-cpu=power8 at the QEMU machine options. This leads to the
guest being set to use hash after QEMU calls the KVM_PPC_ALLOCATE_HTAB
ioctl.
With the guest set to hash, the nested hypervisor goes through the
entry path that has no knowledge of nesting (kvmppc_run_vcpu) and
crashes when it tries to execute an hypervisor-privileged (mtspr
HDEC) instruction at __kvmppc_vcore_entry:
root@L1:~ $ qemu-system-ppc64 -machine pseries,max-cpu-compat=power8 ...
<snip>
[ 538.543303] CPU: 83 PID: 25185 Comm: CPU 0/KVM Not tainted 5.9.0-rc4 #1
[ 538.543355] NIP: c00800000753f388 LR: c00800000753f368 CTR: c0000000001e5ec0
[ 538.543417] REGS: c0000013e91e33b0 TRAP: 0700 Not tainted (5.9.0-rc4)
[ 538.543470] MSR: 8000000002843033 <SF,VEC,VSX,FP,ME,IR,DR,RI,LE> CR: 22422882 XER: 20040000
[ 538.543546] CFAR: c00800000753f4b0 IRQMASK: 3
GPR00: c0080000075397a0 c0000013e91e3640 c00800000755e600 0000000080000000
GPR04: 0000000000000000 c0000013eab19800 c000001394de0000 00000043a054db72
GPR08: 00000000003b1652 0000000000000000 0000000000000000 c0080000075502e0
GPR12: c0000000001e5ec0 c0000007ffa74200 c0000013eab19800 0000000000000008
GPR16: 0000000000000000 c00000139676c6c0 c000000001d23948 c0000013e91e38b8
GPR20: 0000000000000053 0000000000000000 0000000000000001 0000000000000000
GPR24: 0000000000000001 0000000000000001 0000000000000000 0000000000000001
GPR28: 0000000000000001 0000000000000053 c0000013eab19800 0000000000000001
[ 538.544067] NIP [c00800000753f388] __kvmppc_vcore_entry+0x90/0x104 [kvm_hv]
[ 538.544121] LR [c00800000753f368] __kvmppc_vcore_entry+0x70/0x104 [kvm_hv]
[ 538.544173] Call Trace:
[ 538.544196] [c0000013e91e3640] [c0000013e91e3680] 0xc0000013e91e3680 (unreliable)
[ 538.544260] [c0000013e91e3820] [c0080000075397a0] kvmppc_run_core+0xbc8/0x19d0 [kvm_hv]
[ 538.544325] [c0000013e91e39e0] [c00800000753d99c] kvmppc_vcpu_run_hv+0x404/0xc00 [kvm_hv]
[ 538.544394] [c0000013e91e3ad0] [c0080000072da4fc] kvmppc_vcpu_run+0x34/0x48 [kvm]
[ 538.544472] [c0000013e91e3af0] [c0080000072d61b8] kvm_arch_vcpu_ioctl_run+0x310/0x420 [kvm]
[ 538.544539] [c0000013e91e3b80] [c0080000072c7450] kvm_vcpu_ioctl+0x298/0x778 [kvm]
[ 538.544605] [c0000013e91e3ce0] [c0000000004b8c2c] sys_ioctl+0x1dc/0xc90
[ 538.544662] [c0000013e91e3dc0] [c00000000002f9a4] system_call_exception+0xe4/0x1c0
[ 538.544726] [c0000013e91e3e20] [c00000000000d140] system_call_common+0xf0/0x27c
[ 538.544787] Instruction dump:
[ 538.544821] f86d1098 60000000 60000000 48000099 e8ad0fe8 e8c500a0 e9264140 75290002
[ 538.544886] 7d1602a6 7cec42a6 40820008 7d0807b4 <7d164ba6> 7d083a14 f90d10a0 480104fd
[ 538.544953] ---[ end trace 74423e2b948c2e0c ]---
This patch makes the KVM_PPC_ALLOCATE_HTAB ioctl fail when running in
the nested hypervisor, causing QEMU to abort.
Reported-by: Satheesh Rajendran <sathnaga@linux.vnet.ibm.com>
Signed-off-by: Fabiano Rosas <farosas@linux.ibm.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
ENOTSUPP is a linux only thingy, the value of which is unknown to
userspace, not to be confused with ENOTSUP which linux maps to
EOPNOTSUPP, as permitted by POSIX [1]:
[EOPNOTSUPP]
Operation not supported on socket. The type of socket (address family
or protocol) does not support the requested operation. A conforming
implementation may assign the same values for [EOPNOTSUPP] and [ENOTSUP].
Return -EOPNOTSUPP instead of -ENOTSUPP for the following ioctls:
- KVM_GET_FPU for Book3s and BookE
- KVM_SET_FPU for Book3s and BookE
- KVM_GET_DIRTY_LOG for BookE
This doesn't affect QEMU which doesn't call the KVM_GET_FPU and
KVM_SET_FPU ioctls on POWER anyway since they are not supported,
and _buggily_ ignores anything but -EPERM for KVM_GET_DIRTY_LOG.
[1] https://pubs.opengroup.org/onlinepubs/9699919799/functions/V2_chap02.html
Signed-off-by: Greg Kurz <groug@kaod.org>
Acked-by: Thadeu Lima de Souza Cascardo <cascardo@canonical.com>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
The KSZ9477 and KSZ8795 use the port_cnt field differently: For the
KSZ9477, it includes the CPU port(s), while for the KSZ8795, it doesn't.
It would be a good cleanup to make the handling of both drivers match,
but as a first step, fix the recently broken assignment of num_ports in
the KSZ8795 driver (which completely broke probing, as the CPU port
index was always failing the num_ports check).
Fixes: af199a1a9c ("net: dsa: microchip: set the correct number of ports")
Signed-off-by: Matthias Schiffer <matthias.schiffer@ew.tq-group.com>
Reviewed-by: Codrin Ciubotariu <codrin.ciubotariu@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds transport ports information for route lookup so that
IPsec can select Geneve tunnel traffic to do encryption. This is
needed for OVS/OVN IPsec with encrypted Geneve tunnels.
This can be tested by configuring a host-host VPN using an IKE
daemon and specifying port numbers. For example, for an
Openswan-type configuration, the following parameters should be
configured on both hosts and IPsec set up as-per normal:
$ cat /etc/ipsec.conf
conn in
...
left=$IP1
right=$IP2
...
leftprotoport=udp/6081
rightprotoport=udp
...
conn out
...
left=$IP1
right=$IP2
...
leftprotoport=udp
rightprotoport=udp/6081
...
The tunnel can then be setup using "ip" on both hosts (but
changing the relevant IP addresses):
$ ip link add tun type geneve id 1000 remote $IP2
$ ip addr add 192.168.0.1/24 dev tun
$ ip link set tun up
This can then be tested by pinging from $IP1:
$ ping 192.168.0.2
Without this patch the traffic is unencrypted on the wire.
Fixes: 2d07dc79fe ("geneve: add initial netdev driver for GENEVE tunnels")
Signed-off-by: Qiuyu Xiao <qiuyu.xiao.qyx@gmail.com>
Signed-off-by: Mark Gray <mark.d.gray@redhat.com>
Reviewed-by: Greg Rose <gvrose8192@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use DEFINE_SEQ_ATTRIBUTE macro to simplify the code.
Signed-off-by: Liu Shixin <liushixin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use the dim library to manage dynamic interrupt
moderation in ionic.
v3: rebase
v2: untangled declarations in ionic_dim_work()
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
smc->clcsock and smc->clcsock->sk are used before the check if they can
be dereferenced. Fix this by checking the variables first.
Fixes: a60a2b1e0a ("net/smc: reduce active tcp_listen workers")
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When we're handling TO_EXCLUDE report in EXCLUDE filter mode we should
not ignore the return value of __grp_src_toex_excl() as we'll miss
sending notifications about group changes.
Fixes: 5bf1e00b68 ("net: bridge: mcast: support for IGMPV3/MLDv2 CHANGE_TO_INCLUDE/EXCLUDE report")
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The rcu_tasks_trace_postgp() function uses for_each_process_thread()
to scan the task list without the benefit of RCU read-side protection,
which can result in use-after-free errors on task_struct structures.
This error was missed because the TRACE01 rcutorture scenario enables
lockdep, but also builds with CONFIG_PREEMPT_NONE=y. In this situation,
preemption is disabled everywhere, so lockdep thinks everywhere can
be a legitimate RCU reader. This commit therefore adds the needed
rcu_read_lock() and rcu_read_unlock().
Note that this bug can occur only after an RCU Tasks Trace CPU stall
warning, which by default only happens after a grace period has extended
for ten minutes (yes, not a typo, minutes).
Fixes: 4593e772b5 ("rcu-tasks: Add stall warnings for RCU Tasks Trace")
Cc: Alexei Starovoitov <alexei.starovoitov@gmail.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: <bpf@vger.kernel.org>
Cc: <stable@vger.kernel.org> # 5.7.x
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
When rcu_tasks_trace_postgp() function detects an RCU Tasks Trace
CPU stall, it adds all tasks blocking the current grace period to
a list, invoking get_task_struct() on each to prevent them from
being freed while on the list. It then traverses that list,
printing stall-warning messages for each one that is still blocking
the current grace period and removing it from the list. The list
removal invokes the matching put_task_struct().
This of course means that in the admittedly unlikely event that some
task executes its outermost rcu_read_unlock_trace() in the meantime, it
won't be removed from the list and put_task_struct() won't be executing,
resulting in a task_struct leak. This commit therefore makes the list
removal and put_task_struct() unconditional, stopping the leak.
Note further that this bug can occur only after an RCU Tasks Trace CPU
stall warning, which by default only happens after a grace period has
extended for ten minutes (yes, not a typo, minutes).
Fixes: 4593e772b5 ("rcu-tasks: Add stall warnings for RCU Tasks Trace")
Cc: Alexei Starovoitov <alexei.starovoitov@gmail.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: <bpf@vger.kernel.org>
Cc: <stable@vger.kernel.org> # 5.7.x
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
The more intense grace-period processing resulting from the 50x RCU
Tasks Trace grace-period speedups exposed the following race condition:
o Task A running on CPU 0 executes rcu_read_lock_trace(),
entering a read-side critical section.
o When Task A eventually invokes rcu_read_unlock_trace()
to exit its read-side critical section, this function
notes that the ->trc_reader_special.s flag is zero and
and therefore invoke wil set ->trc_reader_nesting to zero
using WRITE_ONCE(). But before that happens...
o The RCU Tasks Trace grace-period kthread running on some other
CPU interrogates Task A, but this fails because this task is
currently running. This kthread therefore sends an IPI to CPU 0.
o CPU 0 receives the IPI, and thus invokes trc_read_check_handler().
Because Task A has not yet cleared its ->trc_reader_nesting
counter, this function sees that Task A is still within its
read-side critical section. This function therefore sets the
->trc_reader_nesting.b.need_qs flag, AKA the .need_qs flag.
Except that Task A has already checked the .need_qs flag, which
is part of the ->trc_reader_special.s flag. The .need_qs flag
therefore remains set until Task A's next rcu_read_unlock_trace().
o Task A now invokes synchronize_rcu_tasks_trace(), which cannot
start a new grace period until the current grace period completes.
And thus cannot return until after that time.
But Task A's .need_qs flag is still set, which prevents the current
grace period from completing. And because Task A is blocked, it
will never execute rcu_read_unlock_trace() until its call to
synchronize_rcu_tasks_trace() returns.
We are therefore deadlocked.
This race is improbable, but 80 hours of rcutorture made it happen twice.
The race was possible before the grace-period speedup, but roughly 50x
less probable. Several thousand hours of rcutorture would have been
necessary to have a reasonable chance of making this happen before this
50x speedup.
This commit therefore eliminates this deadlock by setting
->trc_reader_nesting to a large negative number before checking the
.need_qs and zeroing (or decrementing with respect to its initial
value) ->trc_reader_nesting. For its part, the IPI handler's
trc_read_check_handler() function adds a check for negative values,
deferring evaluation of the task in this case. Taken together, these
changes avoid this deadlock scenario.
Fixes: 276c410448 ("rcu-tasks: Split ->trc_reader_need_end")
Cc: Alexei Starovoitov <alexei.starovoitov@gmail.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: <bpf@vger.kernel.org>
Cc: <stable@vger.kernel.org> # 5.7.x
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
The various RCU tasks flavors currently wait 100 milliseconds between each
grace period in order to prevent CPU-bound loops and to favor efficiency
over latency. However, RCU Tasks Trace needs to have a grace-period
latency of roughly 25 milliseconds, which is completely infeasible given
the 100-millisecond per-grace-period sleep. This commit therefore reduces
this sleep duration to 5 milliseconds (or one jiffy, whichever is longer)
in kernels built with CONFIG_TASKS_TRACE_RCU_READ_MB=y.
Link: https://lore.kernel.org/bpf/CAADnVQK_AiX+S_L_A4CQWT11XyveppBbQSQgH_qWGyzu_E8Yeg@mail.gmail.com/
Reported-by: Alexei Starovoitov <alexei.starovoitov@gmail.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: <bpf@vger.kernel.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Many workloads are quite sensitive to IPIs, and such workloads should
build kernels with CONFIG_TASKS_TRACE_RCU_READ_MB=y to prevent RCU
Tasks Trace from using them under normal conditions. However, other
workloads are quite happy to permit more IPIs if doing so makes BPF
program updates go faster. This commit therefore sets the default
value for the rcupdate.rcu_task_ipi_delay kernel parameter to zero for
kernels that have been built with CONFIG_TASKS_TRACE_RCU_READ_MB=n,
while retaining the old default of (HZ / 10) for kernels that have
indicated an aversion to IPIs via CONFIG_TASKS_TRACE_RCU_READ_MB=y.
Link: https://lore.kernel.org/bpf/CAADnVQK_AiX+S_L_A4CQWT11XyveppBbQSQgH_qWGyzu_E8Yeg@mail.gmail.com/
Reported-by: Alexei Starovoitov <alexei.starovoitov@gmail.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: <bpf@vger.kernel.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Commit 8344496e8b ("rcu-tasks: Conditionally compile
show_rcu_tasks_gp_kthreads()") introduced conditional
compilation of several functions, but forgot one occurrence of
show_rcu_tasks_classic_gp_kthread() that causes the compiler to warn of
an unused static function. This commit uses "static inline" to avoid
these complaints and possibly also to avoid emitting an actual definition
of this function.
Fixes: 8344496e8b ("rcu-tasks: Conditionally compile show_rcu_tasks_gp_kthreads()")
Cc: <stable@vger.kernel.org> # 5.8.x
Reported-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
The RCU Tasks Trace grace periods are too slow, as in 40x slower than
those of RCU Tasks. This is due to my having assumed a one-second grace
period was OK, and thus not having optimized any further. This commit
provides the first step in this optimization process, namely by allowing
the task_list scan backoff interval to be specified on a per-flavor basis,
and then speeding up the scans for RCU Tasks Trace. However, kernels
built with CONFIG_TASKS_TRACE_RCU_READ_MB=y continue to use the old slower
backoff, consistent with that Kconfig option's goal of reducing IPIs.
Link: https://lore.kernel.org/bpf/CAADnVQK_AiX+S_L_A4CQWT11XyveppBbQSQgH_qWGyzu_E8Yeg@mail.gmail.com/
Reported-by: Alexei Starovoitov <alexei.starovoitov@gmail.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: <bpf@vger.kernel.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
The n_heavy_reader_attempts, n_heavy_reader_updates, and
n_heavy_reader_ofl_updates variables are not used outside of their
translation unit, so this commit marks them static.
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
The O_NDELAY flag occurs twice in the VALID_OPEN_FLAGS definition, this
change removes the duplicate. There is no change to the functionality.
Note, that the flags O_NONBLOCK and O_NDELAY are not duplicates, as
values of these flags are platform dependent, and on platforms like
Sparc O_NONBLOCK and O_NDELAY are not the same.
This has been done that way to maintain the ABI compatibility with
Solaris since the Sparc port was first introduced.
This change resolves the following Coccinelle warning:
include/linux/fcntl.h:11:13-21: duplicated argument to & or |
Signed-off-by: Krzysztof Wilczyński <kw@linux.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
if of_find_device_by_node() succeed, mtk_drm_kms_init() doesn't have
a corresponding put_device(). Thus add jump target to fix the exception
handling for this function implementation.
Fixes: 8f83f26891 ("drm/mediatek: Add HDMI support")
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Signed-off-by: Chun-Kuang Hu <chunkuang.hu@kernel.org>
if of_find_device_by_node() succeed, mtk_drm_kms_init() doesn't have
a corresponding put_device(). Thus add jump target to fix the exception
handling for this function implementation.
Fixes: 119f517362 ("drm/mediatek: Add DRM Driver for Mediatek SoC MT8173.")
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Signed-off-by: Chun-Kuang Hu <chunkuang.hu@kernel.org>
mtk_ddp_comp_init() is called in a loop in mtk_drm_probe(), if it
fail, previous successive init component is not proccessed.
Thus uninitialize valid component and put their device if component
init failed.
Fixes: 119f517362 ("drm/mediatek: Add DRM Driver for Mediatek SoC MT8173.")
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Signed-off-by: Chun-Kuang Hu <chunkuang.hu@kernel.org>
if of_find_device_by_node() succeed, mtk_ddp_comp_init() doesn't have
a corresponding put_device(). Thus add put_device() to fix the exception
handling for this function implementation.
Fixes: d0afe37f52 ("drm/mediatek: support CMDQ interface in ddp component")
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Signed-off-by: Chun-Kuang Hu <chunkuang.hu@kernel.org>
Even though cmdq client is created successfully, without the cmdq event,
cmdq could not work correctly, so use CPU when fail to get cmdq event.
Fixes: 60fa8c13ab ("drm/mediatek: Move gce event property to mutex device node")
Signed-off-by: Chun-Kuang Hu <chunkuang.hu@kernel.org>
Remove mtk_drm_ddp.h which is included more than once
Fixes: 9aef5867c8 ("drm/mediatek: drop use of drmP.h")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wang Hai <wanghai38@huawei.com>
Signed-off-by: Chun-Kuang Hu <chunkuang.hu@kernel.org>
This patch add support to --show-ring & --set-ring Ethtool functions:
- Adding min, max, power of two check to new ring parameter's value.
- Bring down the network interface before changing the value of ring
parameters.
- Bring up the network interface after changing the value of ring
parameters.
Signed-off-by: Song, Yoong Siang <yoong.siang.song@intel.com>
Signed-off-by: Voon Weifeng <weifeng.voon@intel.com>
Signed-off-by: Ong Boon Leong <boon.leong.ong@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel says:
====================
mlxsw: Refactor headroom management
Petr says:
On Spectrum, port buffers, also called port headroom, is where packets are
stored while they are parsed and the forwarding decision is being made. For
lossless traffic flows, in case shared buffer admission is not allowed,
headroom is also where to put the extra traffic received before the sent
PAUSE takes effect. Another aspect of the port headroom is the so called
internal buffer, which is used for egress mirroring.
Linux supports two DCB interfaces related to the headroom: dcbnl_setbuffer
for configuration, and dcbnl_getbuffer for inspection. In order to make it
possible to implement these interfaces, it is first necessary to clean up
headroom handling, which is currently strewn in several places in the
driver.
The end goal is an architecture whereby it is possible to take a copy of
the current configuration, adjust parameters, and then hand the proposed
configuration over to the system to implement it. When everything works,
the proposed configuration is accepted and saved. First, this centralizes
the reconfiguration handling to one function, which takes care of
coordinating buffer size changes and priority map changes to avoid
introducing drops. Second, the fact that the configuration is all in one
place makes it easy to keep a backup and handle error path rollbacks, which
were previously hard to understand.
Patch #1 introduces struct mlxsw_sp_hdroom, which will keep port headroom
configuration.
Patch #2 unifies handling of delay provision between PFC and PAUSE. From
now on, delay is to be measured in bytes of extra space, and will not
include MTU. PFC handler sets the delay directly from the parameter it gets
through the DCB interface. For PAUSE, MLXSW_SP_PAUSE_DELAY is converted to
have the same meaning.
In patches #3-#5, MTU, lossiness and priorities are gradually moved over to
struct mlxsw_sp_hdroom.
In patches #6-#11, handling of buffer resizing and priority maps is moved
from spectrum.c and spectrum_dcb.c to spectrum_buffers.c. The API is
gradually adapted so that struct mlxsw_sp_hdroom becomes the main interface
through which the various clients express how the headroom should be
configured.
Patch #12 is a small cleanup that the previous transformation made
possible.
In patch #13, the port init code becomes a boring client of the headroom
code, instead of rolling its own thing.
Patches #14 and #15 move handling of internal mirroring buffer to the new
headroom code as well. Previously, this code was in the SPAN module. This
patchset converts the SPAN module to another boring client of the headroom
code.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Traffic mirroring modes that are in-chip implemented on egress need an
internal buffer to work. As the only client, the SPAN module was managing
the buffer so far. However logically it belongs to the buffers module. E.g.
buffer size validation needs to take the size of the internal buffer into
account.
Therefore move the related code from SPAN to spectrum_buffers. Move over
the callbacks that determine the minimum buffer size as a function of
maximum speed and MTU. Add a field describing the internal buffer to struct
mlxsw_sp_hdroom. Extend mlxsw_sp_hdroom_bufs_reset_sizes() to take care of
sizing the internal buffer as well. Change the SPAN module to invoke that
function and mlxsw_sp_hdroom_configure() like all the other hdroom clients.
Drop the now-unnecessary mlxsw_sp_span_port_buffer_disable().
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The size of the internal buffer is currently calculated in the SPAN module.
Logically it belongs to the spectrum_buffers module, where it should be
moved. However, that being a chip-specific operation, it needs dynamic
dispatch. There currently is a chip-specific structure for description of
shared buffer values, struct mlxsw_sp_sb_vals. However placing ops into
this structure would be confusing. Therefore introduce a new per-chip
structure, currently empty, and initialize the ops pointer as appropriate.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently mlxsw_sp_port_headroom_init() configures both priomap and buffers
by hand. Additionally, for port buffers, it configures buffer 0 with a size
that it will never again have if PFC configuration is touched.
Rewrite the init code to become a client of the new hdroom code. The only
difference in invocation is that the configuration is forced, so that it is
issued even if the desired configuration happens to match what is contained
in (hitherto not initialized with meaningful values) mlxsw_sp_port->hdroom.
Since now mlxsw_sp_port_headroom_init() initializes all the PG buffers to
meaningful values, mlxsw_sp_hdroom_configure_buffers() can avoid querying
the current configuration, and can fill the whole PBMC itself.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This function is now only used from the buffers module, and is a trivial
field reference. Just inline it and drop the related artifacts.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move all the headroom code to the spectrum_buffers module, where it
belongs.
Rename mlxsw_sp_pg_buf_threshold_get() and mlxsw_sp_pg_buf_pack() to
..._hdroom_... to match the naming convention of the new headroom code.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The ETS handler performs the headroom configuration in three steps: first
it resizes the buffers and adds any new ones. Then it redirects priorities
to the new buffers. And finally it sets the size of the now-unused buffers
to zero. This way no packet drops are introduced.
This sort of careful approach will also be useful for configuring port
buffer sizes and priority map by hand, through dcbnl_setbuffer. Therefore
move the code from the DCB handler to the generic headroom function.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The new hdroom code has certain conventions: iteration over priorities is
done through a variable named `prio', configuration is not pushed unless it
is dirty, but a `force' flag can be used to override this, updated
configuration is written to port. Convert the function
mlxsw_sp_port_pg_prio_map() to use these conventions and rename
appropriately to fit in.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The ETS handler performs the headroom configuration in three steps: first
it resizes the buffers and adds any new ones. Then it redirects priorities
to the new buffers. And finally it sets the size of the now-unused buffers
to zero. This way no packet drops are introduced.
Both of the buffer size configuration operations are simply buffer size
configurations, there is no material difference between setting buffers to
zero and any other value. Therefore simply invoke the same
mlxsw_sp_hdroom_configure(), and drop mlxsw_sp_port_pg_destroy() and
mlxsw_sp_ets_has_pg() which are now unused.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Split mlxsw_sp_port_headroom_set() to three functions.
mlxsw_sp_hdroom_bufs_reset_sizes() changes the sizes of the individual PG
buffers, and mlxsw_sp_hdroom_configure_buffers() will actually apply the
configuration. A third function, mlxsw_sp_hdroom_bufs_fit(), verifies that
the requested buffer configuration matches total headroom size
requirements.
Add wrappers, mlxsw_sp_hdroom_configure() and __..., that will eventually
perform full headroom configuration, but for now, only have them verify the
configured headroom size, and invoke mlxsw_sp_hdroom_configure_buffers().
Have them take the `force` argument to prepare for a later patch, even
though it is currently unused.
Note that the loop in mlxsw_sp_hdroom_configure_buffers() only goes through
DCBX_MAX_BUFFERS. Since there is no logic to configure the control buffer,
it needs to keep the values queried from the FW. Eventually this function
should configure all the PGs.
Note that conversion of __mlxsw_sp_dcbnl_ieee_setets() is not trivial. That
function performs the headroom configuration in three steps: first it
resizes the buffers and adds any new ones. Then it redirects priorities to
the new buffers. And finally it sets the size of the now-unused buffers to
zero. This way no packet drops are introduced.
So after invoking mlxsw_sp_hdroom_bufs_reset_sizes(), tweak the
configuration to keep the old sizes of PG buffers for those buffers whose
size was set to zero.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
So far, port buffers were always autoconfigured. When dcbnl_setbuffer
callback is implemented, it will allow the user to change the buffer size
configuration by hand. The sizes therefore need to be a configuration
parameter, not always deduced, and therefore belong to struct
mlxsw_sp_hdroom, where the configuration routine should take them from.
Update mlxsw_sp_port_headroom_set() to update these sizes. Have the
function update the sizes even for the case that a given buffer is not
used.
Additionally, change the loop iteration end to DCBX_MAX_BUFFERS instead of
IEEE_8021QAZ_MAX_TCS. The value is the same, but the semantics differ.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Client-side configuration has lossiness as an attribute of a priority.
Therefore add a "lossy" attribute to struct mlxsw_sp_hdroom_prio.
To a Spectrum ASIC, lossiness is a feature of a port buffer. Therefore add
struct mlxsw_sp_hdroom_buf, which in the following patches will get more
attributes, but right now only use it to track port buffer lossiness.
Instead of passing around the primary indicators of PFC and pause_en, add a
function mlxsw_sp_hdroom_bufs_reset_lossiness() to compute the buffer
lossiness from the priority map and priority lossiness. Change
mlxsw_sp_port_headroom_set() to take the buffer lossy flag from the
headroom configuration. Have the PFC and pause handlers configure priority
lossiness in mlxsw_sp_hdroom, from where it will propagate.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The mapping from priorities to buffers determines which buffers should be
configured. Lossiness of these priorities combined with the mapping
determines whether a given buffer should be lossy.
Currently this configuration is stored implicitly in DCB ETS, PFC and
ethtool PAUSE configuration. Keeping it together with the rest of the
headroom configuration and deriving it as needed from PFC / ETS / PAUSE
will make things clearer. To that end, add a field "prios" to struct
mlxsw_sp_hdroom.
Previously, __mlxsw_sp_port_headroom_set() took prio_tc as an argument, and
assumed that the same mapping as we use on the egress should be used on
ingress as well. Instead, track this configuration at each priority, so
that it can be adjusted flexibly.
In the following patches, as dcbnl_setbuffer is implemented, it will need
to store its own mapping, and it will also be sometimes necessary to revert
back to the original ETS mapping. Therefore track two buffer indices: the
one for chip configuration (buf_idx), and the source one (ets_buf_idx).
Introduce a function to configure the chip-level buffer index, and for now
have it simply copy the ETS mapping over to the chip mapping.
Update the ETS handler to project prio_tc to the ets_buf_idx and invoke the
buf_idx recomputation.
Now that there is a canonical place to look for this configuration,
mlxsw_sp_port_headroom_set() does not need to invent def_prio_tc to use if
DCB is compiled out.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
MTU influences sizes of auto-allocated buffers. Make it a part of port
buffer configuration and have __mlxsw_sp_port_headroom_set() take it from
there, instead of as an argument.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When a priority is marked as lossless using DCB PFC, or when pause frames
are enabled on a port, mlxsw adds to port buffers an extra space to cover
the traffic that will arrive between the time that a pause or PFC frame is
emitted, and the time traffic actually stops. This is called the delay. The
concept is the same in PFC and pause, however the way the extra buffer
space is calculated differs.
In this patch, unify this handling. Delay is to be measured in bytes of
extra space, and will not include MTU. PFC handler sets the delay directly
from the parameter it gets through the DCB interface.
To convert pause handler, move MLXSW_SP_PAUSE_DELAY to ethtool module,
convert to bytes, and reduce it by maximum MTU, and divide by two. Then it
has the same meaning as the delay_bytes set by the PFC handler.
Keep the delay_bytes value in struct mlxsw_sp_hdroom introduced in the
previous patch. Change PFC and pause handlers to store the new delay value
there and have __mlxsw_sp_port_headroom_set() take it from there.
Instead of mlxsw_sp_pfc_delay_get() and mlxsw_sp_pg_buf_delay_get(),
introduce mlxsw_sp_hdroom_buf_delay_get() to calculate the delay provision.
Drop the unnecessary MLXSW_SP_CELL_FACTOR, and instead add an explanatory
comment describing the formula used.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The port headroom handling is currently strewn across several modules and
tricky to follow: MTU, DCB PFC, DCB ETS and ethtool pause all influence the
settings, and then there is the completely separate initial configuraion in
spectrum_buffers. A following patch will implement the dcbnl_setbuffer
callback, which is going to further complicate the landscape.
In order to simplify work with port buffers, the following patches are
going to centralize all port-buffer handling in spectrum_buffers. As a
first step, introduce a (currently empty) struct mlxsw_sp_hdroom that will
keep the configuration parameters, and allocate and free it in appropriate
places.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Various updates to mlx5 driver,
1) Eli adds support for TC trap action.
2) Eran, minor improvements to clock.c code structure
3) Better handling of error reporting in LAG from Jianbo
4) IPv6 traffic class (DSCP) header rewrite support from Maor
5) Ofer Levi adds support for CQE compression of multi-strides packets
6) Vu, Enables use of vport meta data by default.
7) Some minor code cleanup
-----BEGIN PGP SIGNATURE-----
iQEzBAABCAAdFiEEGhZs6bAKwk/OTgTpSD+KveBX+j4FAl9hDykACgkQSD+KveBX
+j6qAQgAn4HWJp7Bu7S7okRbv1bg+uj7mQgU1oEU7P1xzpx2gfZcD0ejjwoxGV/8
iK/FC2KQeuBKqIkLPnQC1o4CH8Fk9kr2HuhmX46Gkn07ohyObf6w8fFVrGv/5QrB
fWUWhu+TQJNA/qnMlCfQ5t5Jt+XYL0m7VdfhCHE3R5rmpcZ2PHhxmvoG/NlBLUUK
kjggjtjX6Vv1CRit0w08FJwsJbqHy3wqpciX4Xc+wZp9A+D5VAyVtXP6ngaDIsAA
RcUzGyH8x4gphnplySkvj/LXboaqiMtd8sPeXCOax2HlYarFAAnNG//7fwhfYIHe
c/509buvfjSFsIwQYRem7d/abkU5Rw==
=4r5e
-----END PGP SIGNATURE-----
Merge tag 'mlx5-updates-2020-09-15' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux
Saeed Mahameed says:
====================
mlx5-updates-2020-09-15
Various updates to mlx5 driver,
1) Eli adds support for TC trap action.
2) Eran, minor improvements to clock.c code structure
3) Better handling of error reporting in LAG from Jianbo
4) IPv6 traffic class (DSCP) header rewrite support from Maor
5) Ofer Levi adds support for CQE compression of multi-strides packets
6) Vu, Enables use of vport meta data by default.
7) Some minor code cleanup
====================
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
ima_file_hash can be called when there is no iint->ima_hash available
even though the inode exists in the integrity cache. It is fairly
common for a file to not have a hash. (e.g. an mknodat, prior to the
file being closed).
Another example where this can happen (suggested by Jann Horn):
Process A does:
while(1) {
unlink("/tmp/imafoo");
fd = open("/tmp/imafoo", O_RDWR|O_CREAT|O_TRUNC, 0700);
if (fd == -1) {
perror("open");
continue;
}
write(fd, "A", 1);
close(fd);
}
and Process B does:
while (1) {
int fd = open("/tmp/imafoo", O_RDONLY);
if (fd == -1)
continue;
char *mapping = mmap(NULL, 0x1000, PROT_READ|PROT_EXEC,
MAP_PRIVATE, fd, 0);
if (mapping != MAP_FAILED)
munmap(mapping, 0x1000);
close(fd);
}
Due to the race to get the iint->mutex between ima_file_hash and
process_measurement iint->ima_hash could still be NULL.
Fixes: 6beea7afcc ("ima: add the ability to query the cached hash of a given file")
Signed-off-by: KP Singh <kpsingh@google.com>
Reviewed-by: Florent Revest <revest@chromium.org>
Signed-off-by: Mimi Zohar <zohar@linux.ibm.com>
Translation Blocking is a required feature for Downstream Ports (Root
Ports or Switch Downstream Ports) that implement ACS. When enabled, the
Port checks the Address Type (AT) of each upstream Memory Request it
receives.
The default AT (00b) means "untranslated" and the IOMMU can decide whether
to treat the address as I/O virtual or physical.
If AT is not the default, i.e., if the Memory Request contains an
already-translated (physical) address, the Port blocks the request and
reports an ACS error.
When enabling ACS, enable Translation Blocking for external-facing ports
and untrusted (external) devices. This is to help prevent attacks from
external devices that initiate DMA with physical addresses that bypass the
IOMMU.
[bhelgaas: commit log, simplify setting bit and drop warning; TB is
required for Downstream Ports with ACS, so we should never see the warning]
Link: https://lore.kernel.org/r/20200707224604.3737893-4-rajatja@google.com
Signed-off-by: Rajat Jain <rajatja@google.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
In the skb Tx path, transmission of a packet is performed with
dev_direct_xmit(). When NETDEV_TX_BUSY is set in the drivers, it
signifies that it was not possible to send the packet right now,
please try later. Unfortunately, the xsk transmit code discarded the
packet and returned EBUSY to the application. Fix this unnecessary
packet loss, by not discarding the packet in the Tx ring and return
EAGAIN. As EAGAIN is returned to the application, it can then retry
the send operation later and the packet will then likely be sent as
the driver will then likely have space/resources to send the packet.
In summary, EAGAIN tells the application that the packet was not
discarded from the Tx ring and that it needs to call send()
again. EBUSY, on the other hand, signifies that the packet was not
sent and discarded from the Tx ring. The application needs to put
the packet on the Tx ring again if it wants it to be sent.
Fixes: 35fcde7f8d ("xsk: support for Tx")
Reported-by: Arkadiusz Zema <A.Zema@falconvsystems.com>
Suggested-by: Arkadiusz Zema <A.Zema@falconvsystems.com>
Suggested-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Link: https://lore.kernel.org/bpf/1600257625-2353-1-git-send-email-magnus.karlsson@gmail.com
On A20R machines the interrupt pending bits in cause register need to be
updated by requesting the chipset to do it. This needs to be done to
find the interrupt cause and after interrupt service. In
commit 0b888c7f3a ("MIPS: SNI: Convert to new irq_chip functions") the
function to do after service update got lost, which caused spurious
interrupts.
Fixes: 0b888c7f3a ("MIPS: SNI: Convert to new irq_chip functions")
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Since we store a pointer to the fake iommu device that is allocated on
the stack, as soon as we leave the function it goes out of scope and any
future dereference is undefined behaviour. Just in case we may need to
look at the fake iommu device after initialiation, move the allocation
from the stack into the data.
Fixes: 01b9d4e211 ("iommu/vt-d: Use dev_iommu_priv_get/set()")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200916105022.28316-2-chris@chris-wilson.co.uk
