If PTP synchronisation is done every second, then sporadic the interval
is higher than one second:
ptp4l[696.582]: master offset -17 s2 freq -1891 path delay 573
ptp4l[697.582]: master offset -22 s2 freq -1901 path delay 573
ptp4l[699.368]: master offset -1 s2 freq -1887 path delay 573
^^^^^^^ Should be 698.582!
This problem is caused by rotten packets, which are received after
polling but before interrupts are enabled again. This can be fixed by
checking for pending work and rescheduling if necessary after interrupts
has been enabled again.
Fixes: 403f69bbdb ("tsnep: Add TSN endpoint Ethernet MAC driver")
Signed-off-by: Gerhard Engleder <gerhard@engleder-embedded.com>
Link: https://lore.kernel.org/r/20221119211825.81805-1-gerhard@engleder-embedded.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
When using dash as /bin/sh, the CC_HAS_ASM_GOTO_TIED_OUTPUT test fails
with a syntax error which is not the one we are looking for:
<stdin>: In function ‘foo’:
<stdin>:1:29: warning: missing terminating " character
<stdin>:1:29: error: missing terminating " character
<stdin>:2:5: error: expected ‘:’ before ‘+’ token
<stdin>:2:7: warning: missing terminating " character
<stdin>:2:7: error: missing terminating " character
<stdin>:2:5: error: expected declaration or statement at end of input
Removing '\n' solves this.
Fixes: 1aa0e8b144 ("Kconfig: Add option for asm goto w/ tied outputs to workaround clang-13 bug")
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Reviewed-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
__io_cq_unlock_post() is identical to io_cq_unlock_post(), and
io_cqring_ev_posted() has a single caller so migth as well just inline
it there.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This reverts commit 7fdbc5f014.
This patch dealt with a subset of the real problem, which is a potential
circular dependency on the wakup path for io_uring itself. Outside of
io_uring, eventfd can also trigger this (see details in 03e02acda8)
and so can epoll (see details in caf1aeaffc). Now that we have a
generic solution to this problem, get rid of the io_uring specific
work-around.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Pass in EPOLL_URING_WAKE when signaling eventfd or doing poll related
wakups, so that we can check for a circular event dependency between
eventfd and epoll. If this flag is set when our wakeup handlers are
called, then we know we have a dependency that needs to terminate
multishot requests.
eventfd and epoll are the only such possible dependencies.
Cc: stable@vger.kernel.org # 6.0
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This is identical to eventfd_signal(), but it allows the caller to pass
in a mask to be used for the poll wakeup key. The use case is avoiding
repeated multishot triggers if we have a dependency between eventfd and
io_uring.
If we setup an eventfd context and register that as the io_uring eventfd,
and at the same time queue a multishot poll request for the eventfd
context, then any CQE posted will repeatedly trigger the multishot request
until it terminates when the CQ ring overflows.
In preparation for io_uring detecting this circular dependency, add the
mentioned helper so that io_uring can pass in EPOLL_URING as part of the
poll wakeup key.
Cc: stable@vger.kernel.org # 6.0
[axboe: fold in !CONFIG_EVENTFD fix from Zhang Qilong]
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Racing conflict could be:
task A task B
list_for_each_entry
strcmp(h->name))
list_for_each_entry
strcmp(h->name)
kzalloc kzalloc
...... .....
device_create device_create
list_add
list_add
The root cause is that task B has no idea about the fact someone
else(A) has inserted heap with same name when it calls list_add,
so a potential collision occurs.
Fixes: c02a81fba7 ("dma-buf: Add dma-buf heaps framework")
Signed-off-by: Dawei Li <set_pte_at@outlook.com>
Acked-by: Andrew Davis <afd@ti.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Sumit Semwal <sumit.semwal@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/TYCP286MB2323873BBDF88020781FB986CA3B9@TYCP286MB2323.JPNP286.PROD.OUTLOOK.COM
Commit 4581dd480c ("net: octeontx2-pf: mcs: consider MACSEC setting")
has already added "depends on MACSEC || !MACSEC", so remove it.
Signed-off-by: Zheng Bin <zhengbin13@huawei.com>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Link: https://lore.kernel.org/r/20221119133616.3583538-1-zhengbin13@huawei.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
As comment of pci_get_domain_bus_and_slot() says, it returns
a pci device with refcount increment, when finish using it,
the caller must decrement the reference count by calling
pci_dev_put(). Call pci_dev_put() before returning from
bnx2x_vf_is_pcie_pending() to avoid refcount leak.
Fixes: b56e9670ff ("bnx2x: Prepare device and initialize VF database")
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Link: https://lore.kernel.org/r/20221119070202.1407648-1-yangyingliang@huawei.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Status is reported as always off in the 6032 case. Status
reporting now matches the logic in the setters. Once of
the differences to the 6030 is that there are no groups,
therefore the state needs to be read out in the lower bits.
Signed-off-by: Andreas Kemnade <andreas@kemnade.info>
Link: https://lore.kernel.org/r/20221120221208.3093727-3-andreas@kemnade.info
Signed-off-by: Mark Brown <broonie@kernel.org>
In former times, info->feature was populated via the parent driver
by pdata/regulator_init_data->driver_data for all regulators when
USB_PRODUCT_ID_LSB indicates a TWL6032.
Today, the information is not set, so re-add it at the regulator
definitions.
Fixes: 25d8233770 ("regulator: twl: make driver DT only")
Signed-off-by: Andreas Kemnade <andreas@kemnade.info>
Link: https://lore.kernel.org/r/20221120221208.3093727-2-andreas@kemnade.info
Signed-off-by: Mark Brown <broonie@kernel.org>
Current code re-calculates the size after aligning the starting and
ending physical addresses on a page boundary. But the re-calculation
also embeds the masking of high order bits that exceed the size of
the physical address space (via PHYSICAL_PAGE_MASK). If the masking
removes any high order bits, the size calculation results in a huge
value that is likely to immediately fail.
Fix this by re-calculating the page-aligned size first. Then mask any
high order bits using PHYSICAL_PAGE_MASK.
Fixes: ffa71f33a8 ("x86, ioremap: Fix incorrect physical address handling in PAE mode")
Signed-off-by: Michael Kelley <mikelley@microsoft.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Acked-by: Dave Hansen <dave.hansen@linux.intel.com>
Cc: <stable@kernel.org>
Link: https://lore.kernel.org/r/1668624097-14884-2-git-send-email-mikelley@microsoft.com
In commit f11fe1dae1 ("net/sched: Make NET_ACT_CT depends on NF_NAT"),
it fixed the build failure when NF_NAT is m and NET_ACT_CT is y by
adding depends on NF_NAT for NET_ACT_CT. However, it would also cause
NET_ACT_CT cannot be built without NF_NAT, which is not expected. This
patch fixes it by changing to use "(!NF_NAT || NF_NAT)" as the depend.
Fixes: f11fe1dae1 ("net/sched: Make NET_ACT_CT depends on NF_NAT")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Link: https://lore.kernel.org/r/b6386f28d1ba34721795fb776a91cbdabb203447.1668807183.git.lucien.xin@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
If phylink_of_phy_connect() fails, the port should be disabled.
If sparx5_serdes_set()/phy_power_on() fails, the port should be
disabled and the phylink should be stopped and disconnected.
Fixes: 946e7fd505 ("net: sparx5: add port module support")
Fixes: f3cad2611a ("net: sparx5: add hostmode with phylink support")
Signed-off-by: Liu Jian <liujian56@huawei.com>
Tested-by: Bjarni Jonasson <bjarni.jonasson@microchip.com>
Reviewed-by: Steen Hegelund <steen.hegelund@microchip.com>
Link: https://lore.kernel.org/r/20221117125918.203997-1-liujian56@huawei.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
acpi_evaluate_dsm() should be coupled with ACPI_FREE() to free the ACPI
memory, because we need to track the allocation of acpi_object when
ACPI_DBG_TRACK_ALLOCATIONS enabled, so use ACPI_FREE() instead of kfree().
Fixes: d38a648d2d ("net: wwan: iosm: fix memory leak in ipc_pcie_read_bios_cfg")
Signed-off-by: Wang ShaoBo <bobo.shaobowang@huawei.com>
Link: https://lore.kernel.org/r/20221118062447.2324881-1-bobo.shaobowang@huawei.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
When IPv6 module initializing in xfrm6_init(), register_pernet_subsys()
is possible to fail but its return value is ignored.
If IPv6 initialization fails later and xfrm6_fini() is called,
removing uninitialized list in xfrm6_net_ops will cause null-ptr-deref:
KASAN: null-ptr-deref in range [0x0000000000000008-0x000000000000000f]
CPU: 1 PID: 330 Comm: insmod
RIP: 0010:unregister_pernet_operations+0xc9/0x450
Call Trace:
<TASK>
unregister_pernet_subsys+0x31/0x3e
xfrm6_fini+0x16/0x30 [ipv6]
ip6_route_init+0xcd/0x128 [ipv6]
inet6_init+0x29c/0x602 [ipv6]
...
Fix it by catching the error return value of register_pernet_subsys().
Fixes: 8d068875ca ("xfrm: make gc_thresh configurable in all namespaces")
Signed-off-by: Chen Zhongjin <chenzhongjin@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Kernel 5.14 added a new "byseq" index to speed
up xfrm_state lookups by sequence number in commit
fe9f1d8779 ("xfrm: add state hashtable keyed by seq")
While the patch was thorough, the function pfkey_send_new_mapping()
in net/af_key.c also modifies x->km.seq and never added
the current xfrm_state to the "byseq" index.
This leads to the following kernel Ooops:
BUG: kernel NULL pointer dereference, address: 0000000000000000
..
RIP: 0010:__xfrm_state_delete+0xc9/0x1c0
..
Call Trace:
<TASK>
xfrm_state_delete+0x1e/0x40
xfrm_del_sa+0xb0/0x110 [xfrm_user]
xfrm_user_rcv_msg+0x12d/0x270 [xfrm_user]
? remove_entity_load_avg+0x8a/0xa0
? copy_to_user_state_extra+0x580/0x580 [xfrm_user]
netlink_rcv_skb+0x51/0x100
xfrm_netlink_rcv+0x30/0x50 [xfrm_user]
netlink_unicast+0x1a6/0x270
netlink_sendmsg+0x22a/0x480
__sys_sendto+0x1a6/0x1c0
? __audit_syscall_entry+0xd8/0x130
? __audit_syscall_exit+0x249/0x2b0
__x64_sys_sendto+0x23/0x30
do_syscall_64+0x3a/0x90
entry_SYSCALL_64_after_hwframe+0x61/0xcb
Exact location of the crash in __xfrm_state_delete():
if (x->km.seq)
hlist_del_rcu(&x->byseq);
The hlist_node "byseq" was never populated.
The bug only triggers if a new NAT traversal mapping (changed IP or port)
is detected in esp_input_done2() / esp6_input_done2(), which in turn
indirectly calls pfkey_send_new_mapping() *if* the kernel is compiled
with CONFIG_NET_KEY and "af_key" is active.
The PF_KEYv2 message SADB_X_NAT_T_NEW_MAPPING is not part of RFC 2367.
Various implementations have been examined how they handle
the "sadb_msg_seq" header field:
- racoon (Android): does not process SADB_X_NAT_T_NEW_MAPPING
- strongswan: does not care about sadb_msg_seq
- openswan: does not care about sadb_msg_seq
There is no standard how PF_KEYv2 sadb_msg_seq should be populated
for SADB_X_NAT_T_NEW_MAPPING and it's not used in popular
implementations either. Herbert Xu suggested we should just
use the current km.seq value as is. This fixes the root cause
of the oops since we no longer modify km.seq itself.
The update of "km.seq" looks like a copy'n'paste error
from pfkey_send_acquire(). SADB_ACQUIRE must indeed assign a unique km.seq
number according to RFC 2367. It has been verified that code paths
involving pfkey_send_acquire() don't cause the same Oops.
PF_KEYv2 SADB_X_NAT_T_NEW_MAPPING support was originally added here:
https://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git
commit cbc3488685b20e7b2a98ad387a1a816aada569d8
Author: Derek Atkins <derek@ihtfp.com>
AuthorDate: Wed Apr 2 13:21:02 2003 -0800
[IPSEC]: Implement UDP Encapsulation framework.
In particular, implement ESPinUDP encapsulation for IPsec
Nat Traversal.
A note on triggering the bug: I was not able to trigger it using VMs.
There is one VPN using a high latency link on our production VPN server
that triggered it like once a day though.
Link: https://github.com/strongswan/strongswan/issues/992
Link: https://lore.kernel.org/netdev/00959f33ee52c4b3b0084d42c430418e502db554.1652340703.git.antony.antony@secunet.com/T/
Link: https://lore.kernel.org/netdev/20221027142455.3975224-1-chenzhihao@meizu.com/T/
Fixes: fe9f1d8779 ("xfrm: add state hashtable keyed by seq")
Reported-by: Roth Mark <rothm@mail.com>
Reported-by: Zhihao Chen <chenzhihao@meizu.com>
Tested-by: Roth Mark <rothm@mail.com>
Signed-off-by: Thomas Jarosch <thomas.jarosch@intra2net.com>
Acked-by: Antony Antony <antony.antony@secunet.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
There is a race between modprobe and mount as below:
modprobe zonefs | mount -t zonefs
--------------------------------|-------------------------
zonefs_init |
register_filesystem [1] |
| zonefs_fill_super [2]
zonefs_sysfs_init [3] |
1. register zonefs suceess, then
2. user can mount the zonefs
3. if sysfs initialize failed, the module initialize failed.
Then the mount process maybe some error happened since the module
initialize failed.
Let's register zonefs after all dependency resource ready. And
reorder the dependency resource release in module exit.
Fixes: 9277a6d4fb ("zonefs: Export open zone resource information through sysfs")
Signed-off-by: Zhang Xiaoxu <zhangxiaoxu5@huawei.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Commit 1229b33973 ("ice: Add low latency Tx timestamp read") refactored
PTP timestamping logic to use a threaded IRQ instead of a separate kthread.
This implementation introduced ice_misc_intr_thread_fn and redefined the
ice_ptp_process_ts function interface to return a value of whether or not
the timestamp processing was complete.
ice_misc_intr_thread_fn would take the return value from ice_ptp_process_ts
and convert it into either IRQ_HANDLED if there were no more timestamps to
be processed, or IRQ_WAKE_THREAD if the thread should continue processing.
This is not correct, as the kernel does not re-schedule threaded IRQ
functions automatically. IRQ_WAKE_THREAD can only be used by the main IRQ
function.
This results in the ice_ptp_process_ts function (and in turn the
ice_ptp_tx_tstamp function) from only being called exactly once per
interrupt.
If an application sends a burst of Tx timestamps without waiting for a
response, the interrupt will trigger for the first timestamp. However,
later timestamps may not have arrived yet. This can result in dropped or
discarded timestamps. Worse, on E822 hardware this results in the interrupt
logic getting stuck such that no future interrupts will be triggered. The
result is complete loss of Tx timestamp functionality.
Fix this by modifying the ice_misc_intr_thread_fn to perform its own
polling of the ice_ptp_process_ts function. We sleep for a few microseconds
between attempts to avoid wasting significant CPU time. The value was
chosen to allow time for the Tx timestamps to complete without wasting so
much time that we overrun application wait budgets in the worst case.
The ice_ptp_process_ts function also currently returns false in the event
that the Tx tracker is not initialized. This would result in the threaded
IRQ handler never exiting if it gets started while the tracker is not
initialized.
Fix the function to appropriately return true when the tracker is not
initialized.
Note that this will not reproduce with default ptp4l behavior, as the
program always synchronously waits for a timestamp response before sending
another timestamp request.
Reported-by: Siddaraju DH <siddaraju.dh@intel.com>
Fixes: 1229b33973 ("ice: Add low latency Tx timestamp read")
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Gurucharan G <gurucharanx.g@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Link: https://lore.kernel.org/r/20221118222729.1565317-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
If skb_linearize() fails in tipc_disc_rcv(), we need to free the skb instead of
handle it.
Fixes: 25b0b9c4e8 ("tipc: handle collisions of 32-bit node address hash values")
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Acked-by: Jon Maloy <jmaloy@redhat.com>
Link: https://lore.kernel.org/r/20221119072832.7896-1-yuehaibing@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Tony Nguyen says:
====================
Intel Wired LAN Driver Updates 2022-11-18 (iavf)
Ivan Vecera resolves issues related to reset by adding back call to
netif_tx_stop_all_queues() and adding calls to dev_close() to ensure
device is properly closed during reset.
Stefan Assmann removes waiting for setting of MAC address as this breaks
ARP.
Slawomir adds setting of __IAVF_IN_REMOVE_TASK bit to prevent deadlock
between remove and shutdown.
* '40GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
iavf: Fix race condition between iavf_shutdown and iavf_remove
iavf: remove INITIAL_MAC_SET to allow gARP to work properly
iavf: Do not restart Tx queues after reset task failure
iavf: Fix a crash during reset task
====================
Link: https://lore.kernel.org/r/20221118222439.1565245-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Xin Long says:
====================
tipc: fix two race issues in tipc_conn_alloc
The race exists beteen tipc_topsrv_accept() and tipc_conn_close(),
one is allocating the con while the other is freeing it and there
is no proper lock protecting it. Therefore, a null-pointer-defer
and a use-after-free may be triggered, see details on each patch.
====================
Link: https://lore.kernel.org/r/cover.1668807842.git.lucien.xin@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
One extra conn_get() is needed in tipc_conn_alloc(), as after
tipc_conn_alloc() is called, tipc_conn_close() may free this
con before deferencing it in tipc_topsrv_accept():
tipc_conn_alloc();
newsk = newsock->sk;
<---- tipc_conn_close();
write_lock_bh(&sk->sk_callback_lock);
newsk->sk_data_ready = tipc_conn_data_ready;
Then an uaf issue can be triggered:
BUG: KASAN: use-after-free in tipc_topsrv_accept+0x1e7/0x370 [tipc]
Call Trace:
<TASK>
dump_stack_lvl+0x33/0x46
print_report+0x178/0x4b0
kasan_report+0x8c/0x100
kasan_check_range+0x179/0x1e0
tipc_topsrv_accept+0x1e7/0x370 [tipc]
process_one_work+0x6a3/0x1030
worker_thread+0x8a/0xdf0
This patch fixes it by holding it in tipc_conn_alloc(), then after
all accessing in tipc_topsrv_accept() releasing it. Note when does
this in tipc_topsrv_kern_subscr(), as tipc_conn_rcv_sub() returns
0 or -1 only, we don't need to check for "> 0".
Fixes: c5fa7b3cf3 ("tipc: introduce new TIPC server infrastructure")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Jon Maloy <jmaloy@redhat.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
A crash was reported by Wei Chen:
BUG: kernel NULL pointer dereference, address: 0000000000000018
RIP: 0010:tipc_conn_close+0x12/0x100
Call Trace:
tipc_topsrv_exit_net+0x139/0x320
ops_exit_list.isra.9+0x49/0x80
cleanup_net+0x31a/0x540
process_one_work+0x3fa/0x9f0
worker_thread+0x42/0x5c0
It was caused by !con->sock in tipc_conn_close(). In tipc_topsrv_accept(),
con is allocated in conn_idr then its sock is set:
con = tipc_conn_alloc();
... <----[1]
con->sock = newsock;
If tipc_conn_close() is called in anytime of [1], the null-pointer-def
is triggered by con->sock->sk due to con->sock is not yet set.
This patch fixes it by moving the con->sock setting to tipc_conn_alloc()
under s->idr_lock. So that con->sock can never be NULL when getting the
con from s->conn_idr. It will be also safer to move con->server and flag
CF_CONNECTED setting under s->idr_lock, as they should all be set before
tipc_conn_alloc() is called.
Fixes: c5fa7b3cf3 ("tipc: introduce new TIPC server infrastructure")
Reported-by: Wei Chen <harperchen1110@gmail.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Jon Maloy <jmaloy@redhat.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Fix to return a negative error code from the ccr read error handling
case instead of 0, as done elsewhere in this function.
Fixes: 3265f42188 ("net: phy: at803x: add fiber support")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20221118103635.254256-1-weiyongjun@huaweicloud.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currenty extended packet number (EPN) update routine is accessing
macsec object without holding the general macsec lock hence facing
a possible race condition when an EPN update occurs while updating
or deleting the SA.
Fix by holding the general macsec lock before accessing the object.
Fixes: 4411a6c0ab ("net/mlx5e: Support MACsec offload extended packet number (EPN)")
Signed-off-by: Emeel Hakim <ehakim@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Currently updating SecY destroys and re-creates RX SA objects,
the re-created RX SA objects are not identical to the destroyed
objects and it disagree on the encryption enabled property which
holds the value false after recreation, this value is not
supported with offload which leads to no traffic after an update.
Fix by recreating an identical objects.
Fixes: 5a39816a75 ("net/mlx5e: Add MACsec offload SecY support")
Signed-off-by: Emeel Hakim <ehakim@nvidia.com>
Reviewed-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Currently as part of MACsec SA initialization routine
extended packet number (EPN) object attribute is always
being set without checking if EPN is actually enabled,
the above could lead to a NULL dereference.
Fix by adding such a check.
Fixes: 4411a6c0ab ("net/mlx5e: Support MACsec offload extended packet number (EPN)")
Signed-off-by: Emeel Hakim <ehakim@nvidia.com>
Reviewed-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Before the cited commit, for N channels, a dedicated set of N queues was
created to support XSK, in indices [N, 2N-1], doubling the number of
queues.
In addition, changing the number of channels was prohibited, as it would
shift the indices.
Remove these two leftovers, as we moved XSK to a new queueing scheme,
starting from index 0.
Fixes: 3db4c85cde ("net/mlx5e: xsk: Use queue indices starting from 0 for XSK queues")
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Gal Pressman <gal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
The cited commit adds a for loop to support multiple encapsulations.
But it only checks if the last encap is valid.
Fix it by setting slow path flag when one of the encap is invalid.
Fixes: f493f15534 ("net/mlx5e: Move flow attr reformat action bit to per dest flags")
Signed-off-by: Chris Mi <cmi@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
When sync reset now event handling fails on mlx5_pci_link_toggle() then
no reset was done. However, since mlx5_cmd_fast_teardown_hca() was
already done, the firmware function is closed and the driver is left
without firmware functionality.
Fix it by setting device error state and reopen the firmware resources.
Reopening is done by the thread that was called for devlink reload
fw_activate as it already holds the devlink lock.
Fixes: 5ec697446f ("net/mlx5: Add support for devlink reload action fw activate")
Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Aya Levin <ayal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
The cited commit moved from using reformat_id integer to packet_reformat
pointer which introduced the possibility to null pointer dereference.
When setting packet reformat flag and pkt_reformat pointer must
exists so checking MLX5_ESW_DEST_ENCAP is not enough, we need
to make sure the pkt_reformat is valid and check for MLX5_ESW_DEST_ENCAP_VALID.
If the dest encap valid flag does not exists then pkt_reformat can be
either invalid address or null.
Also, to make sure we don't try to access invalid pkt_reformat set it to
null when invalidated and invalidate it before calling add flow code as
its logically more correct and to be safe.
Fixes: 2b688ea5ef ("net/mlx5: Add flow steering actions to fs_cmd shim layer")
Signed-off-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Chris Mi <cmi@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
In case command interface is down, or the command is not allowed, driver
did not increment the entry refcount, but might have decrement as part
of forced completion handling.
Fix that by always increment and decrement the refcount to make it
symmetric for all flows.
Fixes: 50b2412b7e ("net/mlx5: Avoid possible free of command entry while timeout comp handler")
Signed-off-by: Eran Ben Elisha <eranbe@nvidia.com>
Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Reported-by: Jack Wang <jinpu.wang@ionos.com>
Tested-by: Jack Wang <jinpu.wang@ionos.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
While moving to new CMD API (quiet API), some pre-existing flows may call the new API
function that in case of error, returns the error instead of printing it as previously done.
For such flows we bring back the print but to tracepoint this time for sys admins to
have the ability to check for errors especially for commands using the new quiet API.
Tracepoint output example:
devlink-1333 [001] ..... 822.746922: mlx5_cmd: ACCESS_REG(0x805) op_mod(0x0) failed, status bad resource(0x5), syndrome (0xb06e1f), err(-22)
Fixes: f23519e542 ("net/mlx5: cmdif, Add new api for command execution")
Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Maor Gottlieb <maorg@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
When SF devices and SF port representors are located on different
functions, unloading and reloading of SF parent driver doesn't recreate
the existing SF present in the device.
Fix it by querying SFs and probe active SFs during driver probe phase.
Fixes: 90d010b863 ("net/mlx5: SF, Add auxiliary device support")
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Parav Pandit <parav@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Fix a bug in calculation of FW tracer timestamp. Decreasing one in the
calculation should effect only bits 52_7 and not effect bits 6_0 of the
timestamp, otherwise bits 6_0 are always set in this calculation.
Fixes: 70dd6fdb89 ("net/mlx5: FW tracer, parse traces and kernel tracing support")
Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Feras Daoud <ferasda@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
The driver should not interact with PCI while PCI is disabled. Trying to
do so may result in being unable to get vital signs during PCI reset,
driver gets timed out and fails to recover.
Fixes: fad1783a6d ("net/mlx5: Print more info on pci error handlers")
Signed-off-by: Roy Novich <royno@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Aya Levin <ayal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
If mes enabled, reserve VM invalidation engine 5 for firmware.
Signed-off-by: Jack Xiao <Jack.Xiao@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org # 6.0.x
Allow user to know number of compute units (CU) that are in use at any
given moment. Enable access to the method kgd_gfx_v9_get_cu_occupancy
that computes CU occupancy.
Signed-off-by: Ramesh Errabolu <Ramesh.Errabolu@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
The basic problem here is that it's not allowed to page fault while
holding the reservation lock.
So it can happen that multiple processes try to validate an userptr
at the same time.
Work around that by putting the HMM range object into the mutex
protected bo list for now.
v2: make sure range is set to NULL in case of an error
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
CC: stable@vger.kernel.org
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Since switching to HMM we always need that because we no longer grab
references to the pages.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
CC: stable@vger.kernel.org
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Coverity noticed this one, so let's fix it.
Fixes: ba891436c2 ("drm/amdgpu/mst: Stop ignoring error codes and deadlocking")
Signed-off-by: Lyude Paul <lyude@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Cc: stable@vger.kernel.org # v5.6+
Now that we've fixed the issue with using the incorrect topology manager,
we're actually grabbing the topology manager's lock - and consequently
deadlocking. Luckily for us though, there's actually nothing in AMD's DSC
state computation code that really should need this lock. The one exception
is the mutex_lock() in dm_dp_mst_is_port_support_mode(), however we grab no
locks beneath &mgr->lock there so that should be fine to leave be.
Gitlab issue: https://gitlab.freedesktop.org/drm/amd/-/issues/2171
Signed-off-by: Lyude Paul <lyude@redhat.com>
Fixes: 8c20a1ed9b ("drm/amd/display: MST DSC compute fair share")
Cc: <stable@vger.kernel.org> # v5.6+
Reviewed-by: Wayne Lin <Wayne.Lin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This bug hurt me. Basically, it appears that we've been grabbing the
entirely wrong mutex in the MST DSC computation code for amdgpu! While
we've been grabbing:
amdgpu_dm_connector->mst_mgr
That's zero-initialized memory, because the only connectors we'll ever
actually be doing DSC computations for are MST ports. Which have mst_mgr
zero-initialized, and instead have the correct topology mgr pointer located
at:
amdgpu_dm_connector->mst_port->mgr;
I'm a bit impressed that until now, this code has managed not to crash
anyone's systems! It does seem to cause a warning in LOCKDEP though:
[ 66.637670] DEBUG_LOCKS_WARN_ON(lock->magic != lock)
This was causing the problems that appeared to have been introduced by:
commit 4d07b0bc40 ("drm/display/dp_mst: Move all payload info into the atomic state")
This wasn't actually where they came from though. Presumably, before the
only thing we were doing with the topology mgr pointer was attempting to
grab mst_mgr->lock. Since the above commit however, we grab much more
information from mst_mgr including the atomic MST state and respective
modesetting locks.
This patch also implies that up until now, it's quite likely we could be
susceptible to race conditions when going through the MST topology state
for DSC computations since we technically will not have grabbed any lock
when going through it.
So, let's fix this by adjusting all the respective code paths to look at
the right pointer and skip things that aren't actual MST connectors from a
topology.
Gitlab issue: https://gitlab.freedesktop.org/drm/amd/-/issues/2171
Signed-off-by: Lyude Paul <lyude@redhat.com>
Fixes: 8c20a1ed9b ("drm/amd/display: MST DSC compute fair share")
Cc: <stable@vger.kernel.org> # v5.6+
Reviewed-by: Wayne Lin <Wayne.Lin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Gerhard Engleder