i40e: Fix ping is lost after configuring ADq on VF

[ Upstream commit 9e0a603cb7dce2a19d98116d42de84b6db26d716 ]

Properly reconfigure VF VSIs after VF request ADQ.
Created new function to update queue mapping and queue pairs per TC
with AQ update VSI. This sets proper RSS size on NIC.
VFs num_queue_pairs should not be changed during setup of queue maps.
Previously, VF main VSI in ADQ had configured too many queues and had
wrong RSS size, which lead to packets not being consumed and drops in
connectivity.

Fixes: bc6d33c8d93f ("i40e: Fix the number of queues available to be mapped for use")
Co-developed-by: Przemyslaw Patynowski <przemyslawx.patynowski@intel.com>
Signed-off-by: Przemyslaw Patynowski <przemyslawx.patynowski@intel.com>
Signed-off-by: Eryk Rybak <eryk.roch.rybak@intel.com>
Tested-by: Tony Brelinski <tony.brelinski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

i40e: Fix changing previously set num_queue_pairs for PFs

[ Upstream commit d2a69fefd75683004ffe87166de5635b3267ee07 ]

Currently, the i40e_vsi_setup_queue_map is basing the count of queues in
TCs on a VSI's alloc_queue_pairs member which is not changed throughout
any user's action (for example via ethtool's set_channels callback).

This implies that vsi->tc_config.tc_info[n].qcount value that is given
to the kernel via netdev_set_tc_queue() that notifies about the count of
queues per particular traffic class is constant even if user has changed
the total count of queues.

This in turn caused the kernel warning after setting the queue count to
the lower value than the initial one:

$ ethtool -l ens801f0
Channel parameters for ens801f0:
Pre-set maximums:
RX:             0
TX:             0
Other:          1
Combined:       64
Current hardware settings:
RX:             0
TX:             0
Other:          1
Combined:       64

$ ethtool -L ens801f0 combined 40

[dmesg]
Number of in use tx queues changed invalidating tc mappings. Priority
traffic classification disabled!

Reason was that vsi->alloc_queue_pairs stayed at 64 value which was used
to set the qcount on TC0 (by default only TC0 exists so all of the
existing queues are assigned to TC0). we update the offset/qcount via
netdev_set_tc_queue() back to the old value but then the
netif_set_real_num_tx_queues() is using the vsi->num_queue_pairs as a
value which got set to 40.

Fix it by using vsi->req_queue_pairs as a queue count that will be
distributed across TCs. Do it only for non-zero values, which implies
that user actually requested the new count of queues.

For VSIs other than main, stay with the vsi->alloc_queue_pairs as we
only allow manipulating the queue count on main VSI.

Fixes: bc6d33c8d93f ("i40e: Fix the number of queues available to be mapped for use")
Co-developed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Co-developed-by: Przemyslaw Patynowski <przemyslawx.patynowski@intel.com>
Signed-off-by: Przemyslaw Patynowski <przemyslawx.patynowski@intel.com>
Signed-off-by: Eryk Rybak <eryk.roch.rybak@intel.com>
Tested-by: Tony Brelinski <tony.brelinski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

i40e: Fix NULL ptr dereference on VSI filter sync

[ Upstream commit 37d9e304acd903a445df8208b8a13d707902dea6 ]

Remove the reason of null pointer dereference in sync VSI filters.
Added new I40E_VSI_RELEASING flag to signalize deleting and releasing
of VSI resources to sync this thread with sync filters subtask.
Without this patch it is possible to start update the VSI filter list
after VSI is removed, that's causing a kernel oops.

Fixes: 41c445ff0f48 ("i40e: main driver core")
Signed-off-by: Grzegorz Szczurek <grzegorzx.szczurek@intel.com>
Signed-off-by: Michal Maloszewski <michal.maloszewski@intel.com>
Reviewed-by: Przemyslaw Patynowski <przemyslawx.patynowski@intel.com>
Reviewed-by: Witold Fijalkowski <witoldx.fijalkowski@intel.com>
Reviewed-by: Jaroslaw Gawin <jaroslawx.gawin@intel.com>
Reviewed-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>
Tested-by: Tony Brelinski <tony.brelinski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

i40e: Fix correct max_pkt_size on VF RX queue

[ Upstream commit 6afbd7b3c53cb7417189f476e99d431daccb85b0 ]

Setting VLAN port increasing RX queue max_pkt_size
by 4 bytes to take VLAN tag into account.
Trigger the VF reset when setting port VLAN for
VF to renegotiate its capabilities and reinitialize.

Fixes: ba4e003d29c1 ("i40e: don't hold spinlock while resetting VF")
Signed-off-by: Sylwester Dziedziuch <sylwesterx.dziedziuch@intel.com>
Signed-off-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>
Signed-off-by: Eryk Rybak <eryk.roch.rybak@intel.com>
Tested-by: Konrad Jankowski <konrad0.jankowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: virtio_net_hdr_to_skb: count transport header in UFO

[ Upstream commit cf9acc90c80ecbee00334aa85d92f4e74014bcff ]

virtio_net_hdr_to_skb does not set the skb's gso_size and gso_type
correctly for UFO packets received via virtio-net that are a little over
the GSO size. This can lead to problems elsewhere in the networking
stack, e.g. ovs_vport_send dropping over-sized packets if gso_size is
not set.

This is due to the comparison

  if (skb->len - p_off > gso_size)

not properly accounting for the transport layer header.

p_off includes the size of the transport layer header (thlen), so
skb->len - p_off is the size of the TCP/UDP payload.

gso_size is read from the virtio-net header. For UFO, fragmentation
happens at the IP level so does not need to include the UDP header.

Hence the calculation could be comparing a TCP/UDP payload length with
an IP payload length, causing legitimate virtio-net packets to have
lack gso_type/gso_size information.

Example: a UDP packet with payload size 1473 has IP payload size 1481.
If the guest used UFO, it is not fragmented and the virtio-net header's
flags indicate that it is a GSO frame (VIRTIO_NET_HDR_GSO_UDP), with
gso_size = 1480 for an MTU of 1500.  skb->len will be 1515 and p_off
will be 42, so skb->len - p_off = 1473.  Hence the comparison fails, and
shinfo->gso_size and gso_type are not set as they should be.

Instead, add the UDP header length before comparing to gso_size when
using UFO. In this way, it is the size of the IP payload that is
compared to gso_size.

Fixes: 6dd912f82680 ("net: check untrusted gso_size at kernel entry")
Signed-off-by: Jonathan Davies <jonathan.davies@nutanix.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: dpaa2-eth: fix use-after-free in dpaa2_eth_remove

[ Upstream commit 9b5a333272a48c2f8b30add7a874e46e8b26129c ]

Access to netdev after free_netdev() will cause use-after-free bug.
Move debug log before free_netdev() call to avoid it.

Fixes: 7472dd9f6499 ("staging: fsl-dpaa2/eth: Move print message")
Signed-off-by: Pavel Skripkin <paskripkin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: sched: act_mirred: drop dst for the direction from egress to ingress

[ Upstream commit f799ada6bf2397c351220088b9b0980125c77280 ]

Without dropping dst, the packets sent from local mirred/redirected
to ingress will may still use the old dst. ip_rcv() will drop it as
the old dst is for output and its .input is dst_discard.

This patch is to fix by also dropping dst for those packets that are
mirred or redirected from egress to ingress in act_mirred.

Note that we don't drop it for the direction change from ingress to
egress, as on which there might be a user case attaching a metadata
dst by act_tunnel_key that would be used later.

Fixes: b57dc7c13ea9 ("net/sched: Introduce action ct")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Cong Wang <cong.wang@bytedance.com>
Reviewed-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: mvmdio: fix compilation warning

[ Upstream commit 2460386bef0b9b98b71728d3c173e15558b78d82 ]

The kernel test robot reported a following issue:

>> drivers/net/ethernet/marvell/mvmdio.c:426:36: warning:
unused variable 'orion_mdio_acpi_match' [-Wunused-const-variable]
   static const struct acpi_device_id orion_mdio_acpi_match[] = {
                                      ^
   1 warning generated.

Fix that by surrounding the variable by appropriate ifdef.

Fixes: c54da4c1acb1 ("net: mvmdio: add ACPI support")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Marcin Wojtas <mw@semihalf.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20211115153024.209083-1-mw@semihalf.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

scsi: ufs: core: Fix another task management completion race

[ Upstream commit 5cb37a26355d79ab290220677b1b57d28e99a895 ]

hba->outstanding_tasks, which is read under host_lock spinlock, tells the
interrupt handler what task management tags are in use by the driver.  The
doorbell register bits indicate which tags are in use by the hardware.  A
doorbell bit that is 0 is because the bit has yet to be set by the driver,
or because the task is complete. It is only possible to disambiguate the 2
cases, if reading/writing the doorbell register is synchronized with
reading/writing hba->outstanding_tasks.

For that reason, reading REG_UTP_TASK_REQ_DOOR_BELL must be done under
spinlock.

Link: https://lore.kernel.org/r/20211108064815.569494-3-adrian.hunter@intel.com
Fixes: f5ef336fd2e4 ("scsi: ufs: core: Fix task management completion")
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

scsi: ufs: core: Fix task management completion timeout race

[ Upstream commit 886fe2915cce6658b0fc19e64b82879325de61ea ]

__ufshcd_issue_tm_cmd() clears req->end_io_data after timing out, which
races with the completion function ufshcd_tmc_handler() which expects
req->end_io_data to have a value.

Note __ufshcd_issue_tm_cmd() and ufshcd_tmc_handler() are already
synchronized using hba->tmf_rqs and hba->outstanding_tasks under the
host_lock spinlock.

It is also not necessary (nor typical) to clear req->end_io_data because
the block layer does it before allocating out requests e.g. via
blk_get_request().

So fix by not clearing it.

Link: https://lore.kernel.org/r/20211108064815.569494-2-adrian.hunter@intel.com
Fixes: f5ef336fd2e4 ("scsi: ufs: core: Fix task management completion")
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

scsi: core: sysfs: Fix hang when device state is set via sysfs

[ Upstream commit 4edd8cd4e86dd3047e5294bbefcc0a08f66a430f ]

This fixes a regression added with:

commit f0f82e2476f6 ("scsi: core: Fix capacity set to zero after
offlinining device")

The problem is that after iSCSI recovery, iscsid will call into the kernel
to set the dev's state to running, and with that patch we now call
scsi_rescan_device() with the state_mutex held. If the SCSI error handler
thread is just starting to test the device in scsi_send_eh_cmnd() then it's
going to try to grab the state_mutex.

We are then stuck, because when scsi_rescan_device() tries to send its I/O
scsi_queue_rq() calls -> scsi_host_queue_ready() -> scsi_host_in_recovery()
which will return true (the host state is still in recovery) and I/O will
just be requeued. scsi_send_eh_cmnd() will then never be able to grab the
state_mutex to finish error handling.

To prevent the deadlock move the rescan-related code to after we drop the
state_mutex.

This also adds a check for if we are already in the running state. This
prevents extra scans and helps the iscsid case where if the transport class
has already onlined the device during its recovery process then we don't
need userspace to do it again plus possibly block that daemon.

Link: https://lore.kernel.org/r/20211105221048.6541-3-michael.christie@oracle.com
Fixes: f0f82e2476f6 ("scsi: core: Fix capacity set to zero after offlinining device")
Cc: Bart Van Assche <bvanassche@acm.org>
Cc: lijinlin <lijinlin3@huawei.com>
Cc: Wu Bo <wubo40@huawei.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Reviewed-by: Wu Bo <wubo40@huawei.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

scsi: ufs: core: Improve SCSI abort handling

[ Upstream commit 3ff1f6b6ba6f97f50862aa50e79959cc8ddc2566 ]

The following has been observed on a test setup:

WARNING: CPU: 4 PID: 250 at drivers/scsi/ufs/ufshcd.c:2737 ufshcd_queuecommand+0x468/0x65c
Call trace:
 ufshcd_queuecommand+0x468/0x65c
 scsi_send_eh_cmnd+0x224/0x6a0
 scsi_eh_test_devices+0x248/0x418
 scsi_eh_ready_devs+0xc34/0xe58
 scsi_error_handler+0x204/0x80c
 kthread+0x150/0x1b4
 ret_from_fork+0x10/0x30

That warning is triggered by the following statement:

WARN_ON(lrbp->cmd);

Fix this warning by clearing lrbp->cmd from the abort handler.

Link: https://lore.kernel.org/r/20211104181059.4129537-1-bvanassche@acm.org
Fixes: 7a3e97b0dc4b ("[SCSI] ufshcd: UFS Host controller driver")
Reviewed-by: Bean Huo <beanhuo@micron.com>
Reviewed-by: Stanley Chu <stanley.chu@mediatek.com>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5: E-Switch, return error if encap isn't supported

[ Upstream commit c4c3176739dfa6efcc5b1d1de4b3fd2b51b048c7 ]

On regular ConnectX HCAs getting encap mode isn't supported when the
E-Switch is in NONE mode. Current code would return no error code when
trying to get encap mode in such case which is wrong.

Fix by returning error value to indicate failure to caller in such case.

Fixes: 8e0aa4bc959c ("net/mlx5: E-switch, Protect eswitch mode changes")
Signed-off-by: Raed Salem <raeds@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5: Lag, update tracker when state change event received

[ Upstream commit ae396d85c01c7bdc9eeceecde1f493d03f793465 ]

Currently, In NETDEV_CHANGELOWERSTATE/NETDEV_CHANGEUPPERSTATE events
handling, tracking is not fully completed if the LAG device is not ready
at the time the events occur. But, we must keep track of the upper and
lower states after receiving the events because RoCE needs this info in
mlx5_lag_get_roce_netdev() - in order to return the corresponding port
that its running on. Returning the wrong (not most recent) port will lead
to gids table being incorrect.

For example: If during the attachment of a slave to the bond, the other
non-attached port performs pci_reload, then the LAG device is not ready,
but that should not result in dismissing attached slave tracker update
automatically (which is performed in mlx5_handle_changelowerstate()), Since
these events might not come later, which can lead to both bond ports
having tx_enabled=0 - which is not a valid state of LAG bond.

Fixes: 9b412cc35f00 ("net/mlx5e: Add LAG warning if bond slave is not lag master")
Signed-off-by: Maher Sanalla <msanalla@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Reviewed-by: Jianbo Liu <jianbol@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5e: CT, Fix multiple allocations and memleak of mod acts

[ Upstream commit 806401c20a0f9c51b6c8fd7035671e6ca841f6c2 ]

CT clear action offload adds additional mod hdr actions to the
flow's original mod actions in order to clear the registers which
hold ct_state.
When such flow also includes encap action, a neigh update event
can cause the driver to unoffload the flow and then reoffload it.

Each time this happens, the ct clear handling adds that same set
of mod hdr actions to reset ct_state until the max of mod hdr
actions is reached.

Also the driver never releases the allocated mod hdr actions and
causing a memleak.

Fix above two issues by moving CT clear mod acts allocation
into the parsing actions phase and only use it when offloading the rule.
The release of mod acts will be done in the normal flow_put().

 backtrace:
    [<000000007316e2f3>] krealloc+0x83/0xd0
    [<00000000ef157de1>] mlx5e_mod_hdr_alloc+0x147/0x300 [mlx5_core]
    [<00000000970ce4ae>] mlx5e_tc_match_to_reg_set_and_get_id+0xd7/0x240 [mlx5_core]
    [<0000000067c5fa17>] mlx5e_tc_match_to_reg_set+0xa/0x20 [mlx5_core]
    [<00000000d032eb98>] mlx5_tc_ct_entry_set_registers.isra.0+0x36/0xc0 [mlx5_core]
    [<00000000fd23b869>] mlx5_tc_ct_flow_offload+0x272/0x1f10 [mlx5_core]
    [<000000004fc24acc>] mlx5e_tc_offload_fdb_rules.part.0+0x150/0x620 [mlx5_core]
    [<00000000dc741c17>] mlx5e_tc_encap_flows_add+0x489/0x690 [mlx5_core]
    [<00000000e92e49d7>] mlx5e_rep_update_flows+0x6e4/0x9b0 [mlx5_core]
    [<00000000f60f5602>] mlx5e_rep_neigh_update+0x39a/0x5d0 [mlx5_core]

Fixes: 1ef3018f5af3 ("net/mlx5e: CT: Support clear action")
Signed-off-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Paul Blakey <paulb@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5: E-Switch, rebuild lag only when needed

[ Upstream commit 2eb0cb31bc4ce2ede5460cf3ef433b40cf5f040d ]

A user can enable VFs without changing E-Switch mode, this can happen
when a user moves straight to switchdev mode and only once in switchdev
VFs are enabled via the sysfs interface.

The cited commit assumed this isn't possible and exposed a single
API function where the E-switch calls into the lag code, breaks the lag
and prevents any other lag operations to take place until the
E-switch update has ended.

Breaking the hardware lag when it isn't needed can make it such that
hardware lag can't be enabled again.

In the sysfs call path check if the current E-Switch mode is NONE,
in the context of the function it can only mean the E-Switch is moving
out of NONE mode and the hardware lag should be disabled and enabled
once the mode change has ended. If the mode isn't NONE it means
VFs are about to be enabled and such operation doesn't require
toggling the hardware lag.

Fixes: cac1eb2cf2e3 ("net/mlx5: Lag, properly lock eswitch if needed")
Signed-off-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5: Update error handler for UCTX and UMEM

[ Upstream commit ba50cd9451f6c49cf0841c0a4a146ff6a2822699 ]

In the fast unload flow, the device state is set to internal error,
which indicates that the driver started the destroy process.
In this case, when a destroy command is being executed, it should return
MLX5_CMD_STAT_OK.
Fix MLX5_CMD_OP_DESTROY_UCTX and MLX5_CMD_OP_DESTROY_UMEM to return OK
instead of EIO.

This fixes a call trace in the umem release process -
[ 2633.536695] Call Trace:
[ 2633.537518]  ib_uverbs_remove_one+0xc3/0x140 [ib_uverbs]
[ 2633.538596]  remove_client_context+0x8b/0xd0 [ib_core]
[ 2633.539641]  disable_device+0x8c/0x130 [ib_core]
[ 2633.540615]  __ib_unregister_device+0x35/0xa0 [ib_core]
[ 2633.541640]  ib_unregister_device+0x21/0x30 [ib_core]
[ 2633.542663]  __mlx5_ib_remove+0x38/0x90 [mlx5_ib]
[ 2633.543640]  auxiliary_bus_remove+0x1e/0x30 [auxiliary]
[ 2633.544661]  device_release_driver_internal+0x103/0x1f0
[ 2633.545679]  bus_remove_device+0xf7/0x170
[ 2633.546640]  device_del+0x181/0x410
[ 2633.547606]  mlx5_rescan_drivers_locked.part.10+0x63/0x160 [mlx5_core]
[ 2633.548777]  mlx5_unregister_device+0x27/0x40 [mlx5_core]
[ 2633.549841]  mlx5_uninit_one+0x21/0xc0 [mlx5_core]
[ 2633.550864]  remove_one+0x69/0xe0 [mlx5_core]
[ 2633.551819]  pci_device_remove+0x3b/0xc0
[ 2633.552731]  device_release_driver_internal+0x103/0x1f0
[ 2633.553746]  unbind_store+0xf6/0x130
[ 2633.554657]  kernfs_fop_write+0x116/0x190
[ 2633.555567]  vfs_write+0xa5/0x1a0
[ 2633.556407]  ksys_write+0x4f/0xb0
[ 2633.557233]  do_syscall_64+0x5b/0x1a0
[ 2633.558071]  entry_SYSCALL_64_after_hwframe+0x65/0xca
[ 2633.559018] RIP: 0033:0x7f9977132648
[ 2633.559821] Code: 89 02 48 c7 c0 ff ff ff ff eb b3 0f 1f 80 00 00 00 00 f3 0f 1e fa 48 8d 05 55 6f 2d 00 8b 00 85 c0 75 17 b8 01 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 58 c3 0f 1f 80 00 00 00 00 41 54 49 89 d4 55
[ 2633.562332] RSP: 002b:00007fffb1a83888 EFLAGS: 00000246 ORIG_RAX: 0000000000000001
[ 2633.563472] RAX: ffffffffffffffda RBX: 000000000000000c RCX: 00007f9977132648
[ 2633.564541] RDX: 000000000000000c RSI: 000055b90546e230 RDI: 0000000000000001
[ 2633.565596] RBP: 000055b90546e230 R08: 00007f9977406860 R09: 00007f9977a54740
[ 2633.566653] R10: 0000000000000000 R11: 0000000000000246 R12: 00007f99774056e0
[ 2633.567692] R13: 000000000000000c R14: 00007f9977400880 R15: 000000000000000c
[ 2633.568725] ---[ end trace 10b4fe52945e544d ]---

Fixes: 6a6fabbfa3e8 ("net/mlx5: Update pci error handler entries and command translation")
Signed-off-by: Neta Ostrovsky <netao@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5e: nullify cq->dbg pointer in mlx5_debug_cq_remove()

[ Upstream commit 76ded29d3fcda4928da8849ffc446ea46871c1c2 ]

Prior to this patch in case mlx5_core_destroy_cq() failed it proceeds
to rest of destroy operations. mlx5_core_destroy_cq() could be called again
by user and cause additional call of mlx5_debug_cq_remove().
cq->dbg was not nullify in previous call and cause the crash.

Fix it by nullify cq->dbg pointer after removal.

Also proceed to destroy operations only if FW return 0
for MLX5_CMD_OP_DESTROY_CQ command.

general protection fault, probably for non-canonical address 0x2000300004058: 0000 [#1] SMP PTI
CPU: 5 PID: 1228 Comm: python Not tainted 5.15.0-rc5_for_upstream_min_debug_2021_10_14_11_06 #1
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014
RIP: 0010:lockref_get+0x1/0x60
Code: 5d e9 53 ff ff ff 48 8d 7f 70 e8 0a 2e 48 00 c7 85 d0 00 00 00 02
00 00 00 c6 45 70 00 fb 5d c3 c3 cc cc cc cc cc cc cc cc 53 <48> 8b 17
48 89 fb 85 d2 75 3d 48 89 d0 bf 64 00 00 00 48 89 c1 48
RSP: 0018:ffff888137dd7a38 EFLAGS: 00010206
RAX: 0000000000000000 RBX: ffff888107d5f458 RCX: 00000000fffffffe
RDX: 000000000002c2b0 RSI: ffffffff8155e2e0 RDI: 0002000300004058
RBP: ffff888137dd7a88 R08: 0002000300004058 R09: ffff8881144a9f88
R10: 0000000000000000 R11: 0000000000000000 R12: ffff8881141d4000
R13: ffff888137dd7c68 R14: ffff888137dd7d58 R15: ffff888137dd7cc0
FS:  00007f4644f2a4c0(0000) GS:ffff8887a2d40000(0000)
knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 000055b4500f4380 CR3: 0000000114f7a003 CR4: 0000000000170ea0
Call Trace:
  simple_recursive_removal+0x33/0x2e0
  ? debugfs_remove+0x60/0x60
  debugfs_remove+0x40/0x60
  mlx5_debug_cq_remove+0x32/0x70 [mlx5_core]
  mlx5_core_destroy_cq+0x41/0x1d0 [mlx5_core]
  devx_obj_cleanup+0x151/0x330 [mlx5_ib]
  ? __pollwait+0xd0/0xd0
  ? xas_load+0x5/0x70
  ? xa_load+0x62/0xa0
  destroy_hw_idr_uobject+0x20/0x80 [ib_uverbs]
  uverbs_destroy_uobject+0x3b/0x360 [ib_uverbs]
  uobj_destroy+0x54/0xa0 [ib_uverbs]
  ib_uverbs_cmd_verbs+0xaf2/0x1160 [ib_uverbs]
  ? uverbs_finalize_object+0xd0/0xd0 [ib_uverbs]
  ib_uverbs_ioctl+0xc4/0x1b0 [ib_uverbs]
  __x64_sys_ioctl+0x3e4/0x8e0

Fixes: 94b960b9deff ("net/mlx5e: Fix memory leak in mlx5_core_destroy_cq() error path")
Signed-off-by: Valentine Fatiev <valentinef@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5: E-Switch, Fix resetting of encap mode when entering switchdev

[ Upstream commit d7751d6476185ff754b9dad2cba0c0a6e43ecadc ]

E-Switch encap mode is relevant only when in switchdev mode.
The RDMA driver can query the encap configuration via
mlx5_eswitch_get_encap_mode(). Make sure it returns the currently
used mode and not the set one.

This reverts the cited commit which reset the encap mode
on entering switchdev and fixes the original issue properly.

Fixes: 9a64144d683a ("net/mlx5: E-Switch, Fix default encap mode")
Signed-off-by: Paul Blakey <paulb@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5e: Wait for concurrent flow deletion during neigh/fib events

[ Upstream commit 362980eada85b5ea691e5e0d9257a991aa7ade47 ]

Function mlx5e_take_tmp_flow() skips flows with zero reference count. This
can cause syndrome 0x179e84 when the called from neigh or route update code
and the skipped flow is not removed from the hardware by the time
underlying encap/decap resource is deleted. Add new completion
'del_hw_done' that is completed when flow is unoffloaded. This is safe to
do because flow with reference count zero needs to be detached from
encap/decap entry before its memory is deallocated, which requires taking
the encap_tbl_lock mutex that is held by the event handlers code.

Fixes: 8914add2c9e5 ("net/mlx5e: Handle FIB events to update tunnel endpoint device")
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5e: kTLS, Fix crash in RX resync flow

[ Upstream commit cc4a9cc03faa6d8db1a6954bb536f2c1e63bdff6 ]

For the TLS RX resync flow, we maintain a list of TLS contexts
that require some attention, to communicate their resync information
to the HW.
Here we fix list corruptions, by protecting the entries against
movements coming from resync_handle_seq_match(), until their resync
handling in napi is fully completed.

Fixes: e9ce991bce5b ("net/mlx5e: kTLS, Add resiliency to RX resync failures")
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

RDMA/core: Set send and receive CQ before forwarding to the driver

[ Upstream commit 6cd7397d01c4a3e09757840299e4f114f0aa5fa0 ]

Preset both receive and send CQ pointers prior to call to the drivers and
overwrite it later again till the mlx4 is going to be changed do not
overwrite ibqp properties.

This change is needed for mlx5, because in case of QP creation failure, it
will go to the path of QP destroy which relies on proper CQ pointers.

 BUG: KASAN: use-after-free in create_qp.cold+0x164/0x16e [mlx5_ib]
 Write of size 8 at addr ffff8880064c55c0 by task a.out/246

 CPU: 0 PID: 246 Comm: a.out Not tainted 5.15.0+ #291
 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014
 Call Trace:
  dump_stack_lvl+0x45/0x59
  print_address_description.constprop.0+0x1f/0x140
  kasan_report.cold+0x83/0xdf
  create_qp.cold+0x164/0x16e [mlx5_ib]
  mlx5_ib_create_qp+0x358/0x28a0 [mlx5_ib]
  create_qp.part.0+0x45b/0x6a0 [ib_core]
  ib_create_qp_user+0x97/0x150 [ib_core]
  ib_uverbs_handler_UVERBS_METHOD_QP_CREATE+0x92c/0x1250 [ib_uverbs]
  ib_uverbs_cmd_verbs+0x1c38/0x3150 [ib_uverbs]
  ib_uverbs_ioctl+0x169/0x260 [ib_uverbs]
  __x64_sys_ioctl+0x866/0x14d0
  do_syscall_64+0x3d/0x90
  entry_SYSCALL_64_after_hwframe+0x44/0xae

 Allocated by task 246:
  kasan_save_stack+0x1b/0x40
  __kasan_kmalloc+0xa4/0xd0
  create_qp.part.0+0x92/0x6a0 [ib_core]
  ib_create_qp_user+0x97/0x150 [ib_core]
  ib_uverbs_handler_UVERBS_METHOD_QP_CREATE+0x92c/0x1250 [ib_uverbs]
  ib_uverbs_cmd_verbs+0x1c38/0x3150 [ib_uverbs]
  ib_uverbs_ioctl+0x169/0x260 [ib_uverbs]
  __x64_sys_ioctl+0x866/0x14d0
  do_syscall_64+0x3d/0x90
  entry_SYSCALL_64_after_hwframe+0x44/0xae

 Freed by task 246:
  kasan_save_stack+0x1b/0x40
  kasan_set_track+0x1c/0x30
  kasan_set_free_info+0x20/0x30
  __kasan_slab_free+0x10c/0x150
  slab_free_freelist_hook+0xb4/0x1b0
  kfree+0xe7/0x2a0
  create_qp.part.0+0x52b/0x6a0 [ib_core]
  ib_create_qp_user+0x97/0x150 [ib_core]
  ib_uverbs_handler_UVERBS_METHOD_QP_CREATE+0x92c/0x1250 [ib_uverbs]
  ib_uverbs_cmd_verbs+0x1c38/0x3150 [ib_uverbs]
  ib_uverbs_ioctl+0x169/0x260 [ib_uverbs]
  __x64_sys_ioctl+0x866/0x14d0
  do_syscall_64+0x3d/0x90
  entry_SYSCALL_64_after_hwframe+0x44/0xae

Fixes: 514aee660df4 ("RDMA: Globally allocate and release QP memory")
Link: https://lore.kernel.org/r/2dbb2e2cbb1efb188a500e5634be1d71956424ce.1636631035.git.leonro@nvidia.com
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

btrfs: make 1-bit bit-fields of scrub_page unsigned int

[ Upstream commit d08e38b62327961295be1c63b562cd46ec97cd07 ]

The bitfields have_csum and io_error are currently signed which is not
recommended as the representation is an implementation defined
behaviour. Fix this by making the bit-fields unsigned ints.

Fixes: 2c36395430b0 ("btrfs: scrub: remove the anonymous structure from scrub_page")
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

udp: Validate checksum in udp_read_sock()

[ Upstream commit 099f896f498a2b26d84f4ddae039b2c542c18b48 ]

It turns out the skb's in sock receive queue could have bad checksums, as
both ->poll() and ->recvmsg() validate checksums. We have to do the same
for ->read_sock() path too before they are redirected in sockmap.

Fixes: d7f571188ecf ("udp: Implement ->read_sock() for sockmap")
Reported-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20211115044006.26068-1-xiyou.wangcong@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

platform/x86: think-lmi: Abort probe on analyze failure

[ Upstream commit 812fcc609502096e98cc3918a4b807722dba8fd9 ]

A Lenovo ThinkStation S20 (4157CTO BIOS 60KT41AUS) fails to boot on
recent kernels including the think-lmi driver, due to the fact that
errors returned by the tlmi_analyze() function are ignored by
tlmi_probe(), where  tlmi_sysfs_init() is called unconditionally.
This results in making use of an array of already freed, non-null
pointers and other uninitialized globals, causing all sorts of nasty
kobject and memory faults.

Make use of the analyze function return value, free a couple leaked
allocations, and remove the settings_count field, which is incremented
but never consumed.

Fixes: a40cd7ef22fb ("platform/x86: think-lmi: Add WMI interface support on Lenovo platforms")
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Reviewed-by: Mark Gross <markgross@kernel.org>
Reviewed-by: Mark Pearson <markpearson@lenovo.com>
Link: https://lore.kernel.org/r/163639463588.1330483.15850167112490200219.stgit@omen
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

platform/x86: hp_accel: Fix an error handling path in 'lis3lv02d_probe()'

[ Upstream commit c961a7d2aa23ae19e0099fbcdf1040fb760eea83 ]

If 'led_classdev_register()' fails, some additional resources should be
released.

Add the missing 'i8042_remove_filter()' and 'lis3lv02d_remove_fs()' calls
that are already in the remove function but are missing here.

Fixes: a4c724d0723b ("platform: hp_accel: add a i8042 filter to remove HPQ6000 data from kb bus stream")
Fixes: 9e0c79782143 ("lis3lv02d: merge with leds hp disk")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Link: https://lore.kernel.org/r/5a4f218f8f16d2e3a7906b7ca3654ffa946895f8.1636314074.git.christophe.jaillet@wanadoo.fr
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

gpio: rockchip: needs GENERIC_IRQ_CHIP to fix build errors

[ Upstream commit d6912b1251b47e6b04ea8c8881dfb35a6e7a3e29 ]

gpio-rockchip uses interfaces that are provided by the Kconfig
symbol GENERIC_IRQ_CHIP, so the driver should select that symbol
in order to prevent build errors.

Fixes these build errors (and more):

aarch64-linux-ld: drivers/gpio/gpio-rockchip.o: in function `rockchip_irq_disable':
gpio-rockchip.c:(.text+0x454): undefined reference to `irq_gc_mask_set_bit'
aarch64-linux-ld: drivers/gpio/gpio-rockchip.o: in function `rockchip_irq_enable':
gpio-rockchip.c:(.text+0x478): undefined reference to `irq_gc_mask_clr_bit'
aarch64-linux-ld: drivers/gpio/gpio-rockchip.o: in function `rockchip_interrupts_register':
gpio-rockchip.c:(.text+0x518): undefined reference to `irq_generic_chip_ops'
aarch64-linux-ld: gpio-rockchip.c:(.text+0x594): undefined reference to `__irq_alloc_domain_generic_chips'
aarch64-linux-ld: gpio-rockchip.c:(.text+0x5cc): undefined reference to `irq_get_domain_generic_chip'
aarch64-linux-ld: gpio-rockchip.c:(.text+0x5e0): undefined reference to `irq_gc_ack_set_bit'
aarch64-linux-ld: gpio-rockchip.c:(.text+0x604): undefined reference to `irq_gc_set_wake'

Fixes: 936ee2675eee ("gpio/rockchip: add driver for rockchip gpio")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Bartosz Golaszewski <brgl@bgdev.pl>
Signed-off-by: Sasha Levin <sashal@kernel.org>

mips: lantiq: add support for clk_get_parent()

[ Upstream commit fc1aabb088860d6cf9dd03612b7a6f0de91ccac2 ]

Provide a simple implementation of clk_get_parent() in the
lantiq subarch so that callers of it will build without errors.

Fixes this build error:
ERROR: modpost: "clk_get_parent" [drivers/iio/adc/ingenic-adc.ko] undefined!

Fixes: 171bb2f19ed6 ("MIPS: Lantiq: Add initial support for Lantiq SoCs")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Suggested-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Cc: linux-mips@vger.kernel.org
Cc: John Crispin <john@phrozen.org>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Jonathan Cameron <jic23@kernel.org>
Cc: linux-iio@vger.kernel.org
Cc: Russell King <linux@armlinux.org.uk>
Cc: Andy Shevchenko <andy.shevchenko@gmail.com>
Acked-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: John Crispin <john@phrozen.org>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

mips: bcm63xx: add support for clk_get_parent()

[ Upstream commit e8f67482e5a4bc8d0b65d606d08cb60ee123b468 ]

BCM63XX selects HAVE_LEGACY_CLK but does not provide/support
clk_get_parent(), so add a simple implementation of that
function so that callers of it will build without errors.

Fixes these build errors:

mips-linux-ld: drivers/iio/adc/ingenic-adc.o: in function `jz4770_adc_init_clk_div':
ingenic-adc.c:(.text+0xe4): undefined reference to `clk_get_parent'
mips-linux-ld: drivers/iio/adc/ingenic-adc.o: in function `jz4725b_adc_init_clk_div':
ingenic-adc.c:(.text+0x1b8): undefined reference to `clk_get_parent'

Fixes: e7300d04bd08 ("MIPS: BCM63xx: Add support for the Broadcom BCM63xx family of SOCs." )
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Reported-by: kernel test robot <lkp@intel.com>
Suggested-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Cc: Artur Rojek <contact@artur-rojek.eu>
Cc: Paul Cercueil <paul@crapouillou.net>
Cc: linux-mips@vger.kernel.org
Cc: Jonathan Cameron <jic23@kernel.org>
Cc: Lars-Peter Clausen <lars@metafoo.de>
Cc: linux-iio@vger.kernel.org
Cc: Florian Fainelli <f.fainelli@gmail.com>
Cc: Andy Shevchenko <andy.shevchenko@gmail.com>
Cc: Russell King <linux@armlinux.org.uk>
Cc: bcm-kernel-feedback-list@broadcom.com
Cc: Jonas Gorski <jonas.gorski@gmail.com>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Acked-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

MIPS: generic/yamon-dt: fix uninitialized variable error

[ Upstream commit 255e51da15baed47531beefd02f222e4dc01f1c1 ]

In the case where fw_getenv returns an error when fetching values
for ememsizea and memsize then variable phys_memsize is not assigned
a variable and will be uninitialized on a zero check of phys_memsize.
Fix this by initializing phys_memsize to zero.

Cleans up cppcheck error:
arch/mips/generic/yamon-dt.c:100:7: error: Uninitialized variable: phys_memsize [uninitvar]

Fixes: f41d2430bbd6 ("MIPS: generic/yamon-dt: Support > 256MB of RAM")
Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

bpf: Fix toctou on read-only map's constant scalar tracking

[ Upstream commit 353050be4c19e102178ccc05988101887c25ae53 ]

Commit a23740ec43ba ("bpf: Track contents of read-only maps as scalars") is
checking whether maps are read-only both from BPF program side and user space
side, and then, given their content is constant, reading out their data via
map->ops->map_direct_value_addr() which is then subsequently used as known
scalar value for the register, that is, it is marked as __mark_reg_known()
with the read value at verification time. Before a23740ec43ba, the register
content was marked as an unknown scalar so the verifier could not make any
assumptions about the map content.

The current implementation however is prone to a TOCTOU race, meaning, the
value read as known scalar for the register is not guaranteed to be exactly
the same at a later point when the program is executed, and as such, the
prior made assumptions of the verifier with regards to the program will be
invalid which can cause issues such as OOB access, etc.

While the BPF_F_RDONLY_PROG map flag is always fixed and required to be
specified at map creation time, the map->frozen property is initially set to
false for the map given the map value needs to be populated, e.g. for global
data sections. Once complete, the loader "freezes" the map from user space
such that no subsequent updates/deletes are possible anymore. For the rest
of the lifetime of the map, this freeze one-time trigger cannot be undone
anymore after a successful BPF_MAP_FREEZE cmd return. Meaning, any new BPF_*
cmd calls which would update/delete map entries will be rejected with -EPERM
since map_get_sys_perms() removes the FMODE_CAN_WRITE permission. This also
means that pending update/delete map entries must still complete before this
guarantee is given. This corner case is not an issue for loaders since they
create and prepare such program private map in successive steps.

However, a malicious user is able to trigger this TOCTOU race in two different
ways: i) via userfaultfd, and ii) via batched updates. For i) userfaultfd is
used to expand the competition interval, so that map_update_elem() can modify
the contents of the map after map_freeze() and bpf_prog_load() were executed.
This works, because userfaultfd halts the parallel thread which triggered a
map_update_elem() at the time where we copy key/value from the user buffer and
this already passed the FMODE_CAN_WRITE capability test given at that time the
map was not "frozen". Then, the main thread performs the map_freeze() and
bpf_prog_load(), and once that had completed successfully, the other thread
is woken up to complete the pending map_update_elem() which then changes the
map content. For ii) the idea of the batched update is similar, meaning, when
there are a large number of updates to be processed, it can increase the
competition interval between the two. It is therefore possible in practice to
modify the contents of the map after executing map_freeze() and bpf_prog_load().

One way to fix both i) and ii) at the same time is to expand the use of the
map's map->writecnt. The latter was introduced in fc9702273e2e ("bpf: Add mmap()
support for BPF_MAP_TYPE_ARRAY") and further refined in 1f6cb19be2e2 ("bpf:
Prevent re-mmap()'ing BPF map as writable for initially r/o mapping") with
the rationale to make a writable mmap()'ing of a map mutually exclusive with
read-only freezing. The counter indicates writable mmap() mappings and then
prevents/fails the freeze operation. Its semantics can be expanded beyond
just mmap() by generally indicating ongoing write phases. This would essentially
span any parallel regular and batched flavor of update/delete operation and
then also have map_freeze() fail with -EBUSY. For the check_mem_access() in
the verifier we expand upon the bpf_map_is_rdonly() check ensuring that all
last pending writes have completed via bpf_map_write_active() test. Once the
map->frozen is set and bpf_map_write_active() indicates a map->writecnt of 0
only then we are really guaranteed to use the map's data as known constants.
For map->frozen being set and pending writes in process of still being completed
we fall back to marking that register as unknown scalar so we don't end up
making assumptions about it. With this, both TOCTOU reproducers from i) and
ii) are fixed.

Note that the map->writecnt has been converted into a atomic64 in the fix in
order to avoid a double freeze_mutex mutex_{un,}lock() pair when updating
map->writecnt in the various map update/delete BPF_* cmd flavors. Spanning
the freeze_mutex over entire map update/delete operations in syscall side
would not be possible due to then causing everything to be serialized.
Similarly, something like synchronize_rcu() after setting map->frozen to wait
for update/deletes to complete is not possible either since it would also
have to span the user copy which can sleep. On the libbpf side, this won't
break d66562fba1ce ("libbpf: Add BPF object skeleton support") as the
anonymous mmap()-ed "map initialization image" is remapped as a BPF map-backed
mmap()-ed memory where for .rodata it's non-writable.

Fixes: a23740ec43ba ("bpf: Track contents of read-only maps as scalars")
Reported-by: w1tcher.bupt@gmail.com
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

iavf: Restore VLAN filters after link down

[ Upstream commit 4293014230b887d94b68aa460ff00153454a3709 ]

Restore VLAN filters after the link is brought down, and up - since all
filters are deleted from HW during the netdev link down routine.

Fixes: ed1f5b58ea01 ("i40evf: remove VLAN filters on close")
Signed-off-by: Akeem G Abodunrin <akeem.g.abodunrin@intel.com>
Tested-by: George Kuruvinakunnel <george.kuruvinakunnel@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

iavf: Fix for setting queues to 0

[ Upstream commit 9a6e9e483a9684a34573fd9f9e30ecfb047cb8cb ]

Now setting combine to 0 will be rejected with the
appropriate error code.
This has been implemented by adding a condition that checks
the value of combine equal to zero.
Without this patch, when the user requested it, no error was
returned and combine was set to the default value for VF.

Fixes: 5520deb15326 ("iavf: Enable support for up to 16 queues")
Signed-off-by: Grzegorz Szczurek <grzegorzx.szczurek@intel.com>
Tested-by: Tony Brelinski <tony.brelinski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

iavf: Fix for the false positive ASQ/ARQ errors while issuing VF reset

[ Upstream commit 321421b57a12e933f92b228e0e6d0b2c6541f41d ]

While issuing VF Reset from the guest OS, the VF driver prints
logs about critical / Overflow error detection. This is not an
actual error since the VF_MBX_ARQLEN register is set to all FF's
for a short period of time and the VF would catch the bits set if
it was reading the register during that spike of time.
This patch introduces an additional check to ignore this condition
since the VF is in reset.

Fixes: 19b73d8efaa4 ("i40evf: Add additional check for reset")
Signed-off-by: Surabhi Boob <surabhi.boob@intel.com>
Tested-by: Tony Brelinski <tony.brelinski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

iavf: validate pointers

[ Upstream commit 131b0edc4028bb88bb472456b1ddba526cfb7036 ]

In some cases, the ethtool get_rxfh handler may be called with a null
key or indir parameter. So check these pointers, or you will have a very
bad day.

Fixes: 43a3d9ba34c9 ("i40evf: Allow PF driver to configure RSS")
Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Tested-by: Tony Brelinski <tony.brelinski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

iavf: prevent accidental free of filter structure

[ Upstream commit 4f0400803818f2642f066d3eacaf013f23554cc7 ]

In iavf_config_clsflower, the filter structure could be accidentally
released at the end, if iavf_parse_cls_flower or iavf_handle_tclass ever
return a non-zero but positive value.

In this case, the function continues through to the end, and will call
kfree() on the filter structure even though it has been added to the
linked list.

This can actually happen because iavf_parse_cls_flower will return
a positive IAVF_ERR_CONFIG value instead of the traditional negative
error codes.

Fix this by ensuring that the kfree() check and error checks are
similar. Use the more idiomatic "if (err)" to catch all non-zero error
codes.

Fixes: 0075fa0fadd0 ("i40evf: Add support to apply cloud filters")
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Tony Brelinski <tony.brelinski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

iavf: Fix failure to exit out from last all-multicast mode

[ Upstream commit 8905072a192fffe9389255489db250c73ecab008 ]

The driver could only quit allmulti when allmulti and promisc modes are
turn on at the same time. If promisc had been off there was no way to turn
off allmulti mode.
The patch corrects this behavior. Switching allmulti does not depends on
promisc state mode anymore

Fixes: f42a5c74da99 ("i40e: Add allmulti support for the VF")
Signed-off-by: Piotr Marczak <piotr.marczak@intel.com>
Tested-by: Tony Brelinski <tony.brelinski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

iavf: don't clear a lock we don't hold

[ Upstream commit 2135a8d5c8186bc92901dc00f179ffd50e54c2ac ]

In iavf_configure_clsflower() the function will bail out if it is unable
to obtain the crit_section lock in a reasonable time. However, it will
clear the lock when exiting, so fix this.

Fixes: 640a8af5841f ("i40evf: Reorder configure_clsflower to avoid deadlock on error")
Signed-off-by: Nicholas Nunley <nicholas.d.nunley@intel.com>
Tested-by: Tony Brelinski <tony.brelinski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

iavf: free q_vectors before queues in iavf_disable_vf

[ Upstream commit 89f22f129696ab53cfbc608e0a2184d0fea46ac1 ]

iavf_free_queues() clears adapter->num_active_queues, which
iavf_free_q_vectors() relies on, so swap the order of these two function
calls in iavf_disable_vf(). This resolves a panic encountered when the
interface is disabled and then later brought up again after PF
communication is restored.

Fixes: 65c7006f234c ("i40evf: assign num_active_queues inside i40evf_alloc_queues")
Signed-off-by: Nicholas Nunley <nicholas.d.nunley@intel.com>
Tested-by: Tony Brelinski <tony.brelinski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

iavf: check for null in iavf_fix_features

[ Upstream commit 8a4a126f4be88eb8b5f00a165ab58c35edf4ef76 ]

If the driver has lost contact with the PF then it enters a disabled state
and frees adapter->vf_res. However, ndo_fix_features can still be called on
the interface, so we need to check for this condition first. Since we have
no information on the features at this time simply leave them unmodified
and return.

Fixes: c4445aedfe09 ("i40evf: Fix VLAN features")
Signed-off-by: Nicholas Nunley <nicholas.d.nunley@intel.com>
Tested-by: Tony Brelinski <tony.brelinski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

iavf: Fix return of set the new channel count

[ Upstream commit 4e5e6b5d9d1334d3490326b6922a2daaf56a867f ]

Fixed return correct code from set the new channel count.
Implemented by check if reset is done in appropriate time.
This solution give a extra time to pf for reset vf in case
when user want set new channel count for all vfs.
Without this patch it is possible to return misleading output
code to user and vf reset not to be correctly performed by pf.

Fixes: 5520deb15326 ("iavf: Enable support for up to 16 queues")
Signed-off-by: Grzegorz Szczurek <grzegorzx.szczurek@intel.com>
Signed-off-by: Mateusz Palczewski <mateusz.palczewski@intel.com>
Tested-by: Konrad Jankowski <konrad0.jankowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

NFSD: Fix exposure in nfsd4_decode_bitmap()

[ Upstream commit c0019b7db1d7ac62c711cda6b357a659d46428fe ]

rtm@csail.mit.edu reports:
> nfsd4_decode_bitmap4() will write beyond bmval[bmlen-1] if the RPC
> directs it to do so. This can cause nfsd4_decode_state_protect4_a()
> to write client-supplied data beyond the end of
> nfsd4_exchange_id.spo_must_allow[] when called by
> nfsd4_decode_exchange_id().

Rewrite the loops so nfsd4_decode_bitmap() cannot iterate beyond
@bmlen.

Reported by: rtm@csail.mit.edu
Fixes: d1c263a031e8 ("NFSD: Replace READ* macros in nfsd4_decode_fattr()")
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/smc: Make sure the link_id is unique

[ Upstream commit cf4f5530bb55ef7d5a91036b26676643b80b1616 ]

The link_id is supposed to be unique, but smcr_next_link_id() doesn't
skip the used link_id as expected. So the patch fixes this.

Fixes: 026c381fb477 ("net/smc: introduce link_idx for link group array")
Signed-off-by: Wen Gu <guwen@linux.alibaba.com>
Reviewed-by: Tony Lu <tonylu@linux.alibaba.com>
Acked-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

sock: fix /proc/net/sockstat underflow in sk_clone_lock()

[ Upstream commit 938cca9e4109b30ee1d476904538225a825e54eb ]

sk_clone_lock() needs to call sock_inuse_add(1) before entering the
sk_free_unlock_clone() error path, for __sk_free() from sk_free() from
sk_free_unlock_clone() calls sock_inuse_add(-1).

Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Fixes: 648845ab7e200993 ("sock: Move the socket inuse to namespace.")
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

tipc: only accept encrypted MSG_CRYPTO msgs

[ Upstream commit 271351d255b09e39c7f6437738cba595f9b235be ]

The MSG_CRYPTO msgs are always encrypted and sent to other nodes
for keys' deployment. But when receiving in peers, if those nodes
do not validate it and make sure it's encrypted, one could craft
a malicious MSG_CRYPTO msg to deploy its key with no need to know
other nodes' keys.

This patch is to do that by checking TIPC_SKB_CB(skb)->decrypted
and discard it if this packet never got decrypted.

Note that this is also a supplementary fix to CVE-2021-43267 that
can be triggered by an unencrypted malicious MSG_CRYPTO msg.

Fixes: 1ef6f7c9390f ("tipc: add automatic session key exchange")
Acked-by: Ying Xue <ying.xue@windriver.com>
Acked-by: Jon Maloy <jmaloy@redhat.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

bnxt_en: reject indirect blk offload when hw-tc-offload is off

[ Upstream commit b0757491a118ae5727cf9f1c3a11544397d46596 ]

The driver does not check if hw-tc-offload is enabled for the device
before offloading a flow in the context of indirect block callback.
Fix this by checking NETIF_F_HW_TC in the features flag and rejecting
the offload request.  This will avoid unnecessary dmesg error logs when
hw-tc-offload is disabled, such as these:

bnxt_en 0000:19:00.1 eno2np1: dev(ifindex=294) not on same switch
bnxt_en 0000:19:00.1 eno2np1: Error: bnxt_tc_add_flow: cookie=0xffff8dace1c88000 error=-22
bnxt_en 0000:19:00.0 eno1np0: dev(ifindex=294) not on same switch
bnxt_en 0000:19:00.0 eno1np0: Error: bnxt_tc_add_flow: cookie=0xffff8dace1c88000 error=-22

Reported-by: Marcelo Ricardo Leitner <mleitner@redhat.com>
Fixes: 627c89d00fb9 ("bnxt_en: flow_offload: offload tunnel decap rules via indirect callbacks")
Signed-off-by: Sriharsha Basavapatna <sriharsha.basavapatna@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: bnx2x: fix variable dereferenced before check

[ Upstream commit f8885ac89ce310570e5391fe0bf0ec9c7c9b4fdc ]

Smatch says:
bnx2x_init_ops.h:640 bnx2x_ilt_client_mem_op()
warn: variable dereferenced before check 'ilt' (see line 638)

Move ilt_cli variable initialization _after_ ilt validation, because
it's unsafe to deref the pointer before validation check.

Fixes: 523224a3b3cd ("bnx2x, cnic, bnx2i: use new FW/HSI")
Signed-off-by: Pavel Skripkin <paskripkin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests: gpio: fix gpio compiling error

[ Upstream commit 92a59d7f381d2caf69385bfa00590028e32eea26 ]

The gpio selftests build against the system includes rather than the
headers from the linux tree.  This results in the compile failing if
the system includes are outdated.

Prefer the headers from the linux tree, as per other selftests.

Fixes: 8bc395a6a2e2 ("selftests: gpio: rework and simplify test implementation")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Li Zhijian <lizhijian@cn.fujitsu.com>
[Kent: reworded commit comment and added Fixes:]
Signed-off-by: Kent Gibson <warthog618@gmail.com>
Signed-off-by: Bartosz Golaszewski <brgl@bgdev.pl>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: ipa: disable HOLB drop when updating timer

[ Upstream commit 816316cacad2b5abd5b41423cf04e4845239abd4 ]

The head-of-line blocking timer should only be modified when
head-of-line drop is disabled.

One of the steps in recovering from a modem crash is to enable
dropping of packets with timeout of 0 (immediate).  We don't know
how the modem configured its endpoints, so before we program the
timer, we need to ensure HOL_BLOCK is disabled.

Fixes: 84f9bd12d46db ("soc: qcom: ipa: IPA endpoints")
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: ipa: HOLB register sometimes must be written twice

[ Upstream commit 6e228d8cbb1cc6ba78022d406340e901e08d26e0 ]

Starting with IPA v4.5, the HOL_BLOCK_EN register must be written
twice when enabling head-of-line blocking avoidance.

Fixes: 84f9bd12d46db ("soc: qcom: ipa: IPA endpoints")
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

mac80211: fix monitor_sdata RCU/locking assertions

[ Upstream commit 6dd2360334f3cb3b45fc1b8194c670090474b87c ]

Since commit a05829a7222e ("cfg80211: avoid holding the RTNL when
calling the driver") we've not only been protecting the pointer
to monitor_sdata with the RTNL, but also with the wiphy->mtx. This
is relevant in a number of lockdep assertions, e.g. the one we hit
in ieee80211_set_monitor_channel(). However, we're now protecting
all the assignments/dereferences, even the one in interface iter,
with the wiphy->mtx, so switch over the lockdep assertions to that
lock.

Fixes: a05829a7222e ("cfg80211: avoid holding the RTNL when calling the driver")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Link: https://lore.kernel.org/r/20211112135143.cb8e8ceffef3.Iaa210f16f6904c8a7a24954fb3396da0ef86ec08@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

nl80211: fix radio statistics in survey dump

[ Upstream commit ce6b69749961426c6d822215ded9e67154e1ad4f ]

Even if userspace specifies the NL80211_ATTR_SURVEY_RADIO_STATS
attribute, we cannot get the statistics because we're not really
parsing the incoming attributes properly any more.

Fix this by passing the attrbuf to nl80211_prepare_wdev_dump()
and filling it there, if given, and using a local version only
if no output is desired.

Since I'm touching it anyway, make nl80211_prepare_wdev_dump()
static.

Fixes: 50508d941c18 ("cfg80211: use parallel_ops for genl")
Reported-by: Jan Fuchs <jf@simonwunderlich.de>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Tested-by: Sven Eckelmann <sven@narfation.org>
Link: https://lore.kernel.org/r/20211029092539.2851b4799386.If9736d4575ee79420cbec1bd930181e1d53c7317@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

tracing: Add length protection to histogram string copies

[ Upstream commit 938aa33f14657c9ed9deea348b7d6f14b6d69cb7 ]

The string copies to the histogram storage has a max size of 256 bytes
(defined by MAX_FILTER_STR_VAL). Only the string size of the event field
needs to be copied to the event storage, but no more than what is in the
event storage. Although nothing should be bigger than 256 bytes, there's
no protection against overwriting of the storage if one day there is.

Copy no more than the destination size, and enforce it.

Also had to turn MAX_FILTER_STR_VAL into an unsigned int, to keep the
min() comparison of the string sizes of comparable types.

Link: https://lore.kernel.org/all/CAHk-=wjREUihCGrtRBwfX47y_KrLCGjiq3t6QtoNJpmVrAEb1w@mail.gmail.com/
Link: https://lkml.kernel.org/r/20211114132834.183429a4@rorschach.local.home
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Tom Zanussi <zanussi@kernel.org>
Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Reviewed-by: Masami Hiramatsu <mhiramat@kernel.org>
Fixes: 63f84ae6b82b ("tracing/histogram: Do not copy the fixed-size char array field over the field size")
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

tcp: Fix uninitialized access in skb frags array for Rx 0cp.

[ Upstream commit 70701b83e208767f2720d8cd3e6a62cddafb3a30 ]

TCP Receive zerocopy iterates through the SKB queue via
tcp_recv_skb(), acquiring a pointer to an SKB and an offset within
that SKB to read from. From there, it iterates the SKB frags array to
determine which offset to start remapping pages from.

However, this is built on the assumption that the offset read so far
within the SKB is smaller than the SKB length. If this assumption is
violated, we can attempt to read an invalid frags array element, which
would cause a fault.

tcp_recv_skb() can cause such an SKB to be returned when the TCP FIN
flag is set. Therefore, we must guard against this occurrence inside
skb_advance_frag().

One way that we can reproduce this error follows:
1) In a receiver program, call getsockopt(TCP_ZEROCOPY_RECEIVE) with:
char some_array[32 * 1024];
struct tcp_zerocopy_receive zc = {
  .copybuf_address  = (__u64) &some_array[0],
  .copybuf_len = 32 * 1024,
};

2) In a sender program, after a TCP handshake, send the following
sequence of packets:
  i) Seq = [X, X+4000]
  ii) Seq = [X+4000, X+5000]
  iii) Seq = [X+4000, X+5000], Flags = FIN | URG, urgptr=1000

(This can happen without URG, if we have a signal pending, but URG is
a convenient way to reproduce the behaviour).

In this case, the following event sequence will occur on the receiver:

tcp_zerocopy_receive():
-> receive_fallback_to_copy() // copybuf_len >= inq
-> tcp_recvmsg_locked() // reads 5000 bytes, then breaks due to URG
-> tcp_recv_skb() // yields skb with skb->len == offset
-> tcp_zerocopy_set_hint_for_skb()
-> skb_advance_to_frag() // will returns a frags ptr. >= nr_frags
-> find_next_mappable_frag() // will dereference this bad frags ptr.

With this patch, skb_advance_to_frag() will no longer return an
invalid frags pointer, and will return NULL instead, fixing the issue.

Signed-off-by: Arjun Roy <arjunroy@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Fixes: 05255b823a61 ("tcp: add TCP_ZEROCOPY_RECEIVE support for zerocopy receive")
Link: https://lore.kernel.org/r/20211111235215.2605384-1-arjunroy.kdev@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/ipa: ipa_resource: Fix wrong for loop range

[ Upstream commit 27df68d579c67ef6c39a5047559b6a7c08c96219 ]

The source group count was mistakenly assigned to both dst and src loops.
Fix it to make IPA probe and work again.

Fixes: 4fd704b3608a ("net: ipa: record number of groups in data")
Acked-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@somainline.org>
Reviewed-by: Marijn Suijten <marijn.suijten@somainline.org>
Signed-off-by: Konrad Dybcio <konrad.dybcio@somainline.org>
Reviewed-by: Alex Elder <elder@linaro.org>
Link: https://lore.kernel.org/r/20211111183724.593478-1-konrad.dybcio@somainline.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests: net: switch to socat in the GSO GRE test

[ Upstream commit 0cda7d4bac5fd29dceb13df26083333fa99d6bb4 ]

Commit a985442fdecb ("selftests: net: properly support IPv6 in GSO GRE test")
is not compatible with:

  Ncat: Version 7.80 ( https://nmap.org/ncat )

(which is distributed with Fedora/Red Hat), tests fail with:

  nc: invalid option -- 'N'

Let's switch to socat which is far more dependable.

Fixes: 025efa0a82df ("selftests: add simple GSO GRE test")
Fixes: a985442fdecb ("selftests: net: properly support IPv6 in GSO GRE test")
Tested-by: Andrea Righi <andrea.righi@canonical.com>
Link: https://lore.kernel.org/r/20211111162929.530470-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

samples/bpf: Fix incorrect use of strlen in xdp_redirect_cpu

[ Upstream commit 2453afe3845523d9dfe89dbfb3d71abfa095e260 ]

Commit b599015f044d ("samples/bpf: Fix application of sizeof to pointer")
tried to fix a bug where sizeof was incorrectly applied to a pointer instead
of the array string was being copied to, to find the destination buffer size,
but ended up using strlen, which is still incorrect. However, on closer look
ifname_buf has no other use, hence directly use optarg.

Fixes: b599015f044d ("samples/bpf: Fix application of sizeof to pointer")
Fixes: e531a220cc59 ("samples: bpf: Convert xdp_redirect_cpu to XDP samples helper")
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Alexander Lobakin <alexandr.lobakin@intel.com>
Tested-by: Alexander Lobakin <alexandr.lobakin@intel.com>
Link: https://lore.kernel.org/bpf/20211112020301.528357-1-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

samples/bpf: Fix summary per-sec stats in xdp_sample_user

[ Upstream commit dc14ca4644f48b1cfa93631e35c28bdc011ad109 ]

sample_summary_print() uses accumulated period to calculate and display
per-sec averages. This period gets incremented by sampling interval each
time a new sample is formed, and thus equals to the number of samples
collected multiplied by this interval.

However, the totals are being calculated differently, they receive current
sample statistics already divided by the interval gotten as a difference
between sample timestamps for better precision -- in other words, they are
being incremented by the per-sec values each sample.

This leads to the excessive division of summary per-secs when interval != 1
sec. It is obvious pps couldn't become two times lower just from picking a
different sampling interval value:

  $ samples/bpf/xdp_redirect_cpu -p xdp_prognum_n1_inverse_qnum -c all
    -s -d 6 -i 1
  < snip >
    Packets received    : 2,197,230,321
    Average packets/s   : 22,887,816
    Packets redirected  : 2,197,230,472
    Average redir/s     : 22,887,817
  $ samples/bpf/xdp_redirect_cpu -p xdp_prognum_n1_inverse_qnum -c all
    -s -d 6 -i 2
  < snip >
    Packets received    : 159,566,498
    Average packets/s   : 11,397,607
    Packets redirected  : 159,566,995
    Average redir/s     : 11,397,642

This can be easily fixed by treating the divisor not as a period, but rather
as a total number of samples, and thus incrementing it by 1 instead of
interval. As a nice side effect, we can now remove so-named argument from a
couple of functions. Let us also create an "alias" for sample_output::rx_cnt::pps
named 'num' using a union since this field is used to store this number (period
previously) as well, and the resulting counter-intuitive code might've been a
reason for this bug.

Fixes: 156f886cf697 ("samples: bpf: Add basic infrastructure for XDP samples")
Signed-off-by: Alexander Lobakin <alexandr.lobakin@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Reviewed-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/bpf/20211111215703.690-1-alexandr.lobakin@intel.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

bpf: Fix inner map state pruning regression.

[ Upstream commit 34d11a440c6167133201b7374065b59f259730d7 ]

Introduction of map_uid made two lookups from outer map to be distinct.
That distinction is only necessary when inner map has an embedded timer.
Otherwise it will make the verifier state pruning to be conservative
which will cause complex programs to hit 1M insn_processed limit.
Tighten map_uid logic to apply to inner maps with timers only.

Fixes: 3e8ce29850f1 ("bpf: Prevent pointer mismatch in bpf_timer_init.")
Reported-by: Lorenz Bauer <lmb@cloudflare.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Tested-by: Lorenz Bauer <lmb@cloudflare.com>
Link: https://lore.kernel.org/bpf/CACAyw99hVEJFoiBH_ZGyy=+oO-jyydoz6v1DeKPKs2HVsUH28w@mail.gmail.com
Link: https://lore.kernel.org/bpf/20211110172556.20754-1-alexei.starovoitov@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/nouveau: hdmigv100.c: fix corrupted HDMI Vendor InfoFrame

[ Upstream commit 3cc1ae1fa70ab369e4645e38ce335a19438093ad ]

gv100_hdmi_ctrl() writes vendor_infoframe.subpack0_high to 0x6f0110, and
then overwrites it with 0. Just drop the overwrite with 0, that's clearly
a mistake.

Because of this issue the HDMI VIC is 0 instead of 1 in the HDMI Vendor
InfoFrame when transmitting 4kp30.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Fixes: 290ffeafcc1a ("drm/nouveau/disp/gv100: initial support")
Reviewed-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/3d3bd0f7-c150-2479-9350-35d394ee772d@xs4all.nl
Signed-off-by: Sasha Levin <sashal@kernel.org>

perf tests: Remove bash construct from record+zstd_comp_decomp.sh

[ Upstream commit a9cdc1c5e3700a5200e5ca1f90b6958b6483845b ]

Commit 463538a383a2 ("perf tests: Fix test 68 zstd compression for
s390") inadvertently removed the -g flag from all platforms rather than
just s390, because the [[ ]] construct fails in sh. Changing to single
brackets restores testing of call graphs and removes the following error
from the output:

  $ ./perf test -v 85
  85: Zstd perf.data compression/decompression                        :
  --- start ---
  test child forked, pid 50643
  Collecting compressed record file:
  ./tests/shell/record+zstd_comp_decomp.sh: 15: [[: not found

Fixes: 463538a383a2 ("perf tests: Fix test 68 zstd compression for s390")
Signed-off-by: James Clark <james.clark@arm.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Florian Fainelli <f.fainelli@gmail.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: John Fastabend <john.fastabend@gmail.com>
Cc: KP Singh <kpsingh@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Martin KaFai Lau <kafai@fb.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Sumanth Korikkar <sumanthk@linux.ibm.com>
Cc: Thomas Richter <tmricht@linux.ibm.com>
Cc: Yonghong Song <yhs@fb.com>
Cc: bpf@vger.kernel.org
Cc: netdev@vger.kernel.org
Link: https://lore.kernel.org/r/20211028134828.65774-3-james.clark@arm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

perf bench futex: Fix memory leak of perf_cpu_map__new()

[ Upstream commit 88e48238d53682281c9de2a0b65d24d3b64542a0 ]

ASan reports memory leaks while running:

  $ sudo ./perf bench futex all

The leaks are caused by perf_cpu_map__new not being freed.
This patch adds the missing perf_cpu_map__put since it calls
cpu_map_delete implicitly.

Fixes: 9c3516d1b850ea93 ("libperf: Add perf_cpu_map__new()/perf_cpu_map__read() functions")
Signed-off-by: Sohaib Mohamed <sohaib.amhmd@gmail.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: André Almeida <andrealmeid@collabora.com>
Cc: Darren Hart <dvhart@infradead.org>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Sohaib Mohamed <sohaib.amhmd@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lore.kernel.org/lkml/20211112201134.77892-1-sohaib.amhmd@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

perf bpf: Avoid memory leak from perf_env__insert_btf()

[ Upstream commit 4924b1f7c46711762fd0e65c135ccfbcfd6ded1f ]

perf_env__insert_btf() doesn't insert if a duplicate BTF id is
encountered and this causes a memory leak. Modify the function to return
a success/error value and then free the memory if insertion didn't
happen.

v2. Adds a return -1 when the insertion error occurs in
    perf_env__fetch_btf. This doesn't affect anything as the result is
    never checked.

Fixes: 3792cb2ff43b1b19 ("perf bpf: Save BTF in a rbtree in perf_env")
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Andrii Nakryiko <andrii@kernel.org>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: John Fastabend <john.fastabend@gmail.com>
Cc: KP Singh <kpsingh@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Martin KaFai Lau <kafai@fb.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tiezhu Yang <yangtiezhu@loongson.cn>
Cc: Yonghong Song <yhs@fb.com>
Cc: bpf@vger.kernel.org
Cc: netdev@vger.kernel.org
Link: http://lore.kernel.org/lkml/20211112074525.121633-1-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

tracing/histogram: Do not copy the fixed-size char array field over the field size

[ Upstream commit 63f84ae6b82bb4dff672f76f30c6fd7b9d3766bc ]

Do not copy the fixed-size char array field of the events over
the field size. The histogram treats char array as a string and
there are 2 types of char array in the event, fixed-size and
dynamic string. The dynamic string (__data_loc) field must be
null terminated, but the fixed-size char array field may not
be null terminated (not a string, but just a data).
In that case, histogram can copy the data after the field.
This uses the original field size for fixed-size char array
field to restrict the histogram not to access over the original
field size.

Link: https://lkml.kernel.org/r/163673292822.195747.3696966210526410250.stgit@devnote2
Fixes: 02205a6752f2 (tracing: Add support for 'field variables')
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

blkcg: Remove extra blkcg_bio_issue_init

[ Upstream commit b781d8db580c058ecd54ed7d5dde7f8270b25f5b ]

KASAN reports a use-after-free report when doing block test:

==================================================================
[10050.967049] BUG: KASAN: use-after-free in
submit_bio_checks+0x1539/0x1550

[10050.977638] Call Trace:
[10050.978190]  dump_stack+0x9b/0xce
[10050.979674]  print_address_description.constprop.6+0x3e/0x60
[10050.983510]  kasan_report.cold.9+0x22/0x3a
[10050.986089]  submit_bio_checks+0x1539/0x1550
[10050.989576]  submit_bio_noacct+0x83/0xc80
[10050.993714]  submit_bio+0xa7/0x330
[10050.994435]  mpage_readahead+0x380/0x500
[10050.998009]  read_pages+0x1c1/0xbf0
[10051.002057]  page_cache_ra_unbounded+0x4c2/0x6f0
[10051.007413]  do_page_cache_ra+0xda/0x110
[10051.008207]  force_page_cache_ra+0x23d/0x3d0
[10051.009087]  page_cache_sync_ra+0xca/0x300
[10051.009970]  generic_file_buffered_read+0xbea/0x2130
[10051.012685]  generic_file_read_iter+0x315/0x490
[10051.014472]  blkdev_read_iter+0x113/0x1b0
[10051.015300]  aio_read+0x2ad/0x450
[10051.023786]  io_submit_one+0xc8e/0x1d60
[10051.029855]  __se_sys_io_submit+0x125/0x350
[10051.033442]  do_syscall_64+0x2d/0x40
[10051.034156]  entry_SYSCALL_64_after_hwframe+0x44/0xa9

[10051.048733] Allocated by task 18598:
[10051.049482]  kasan_save_stack+0x19/0x40
[10051.050263]  __kasan_kmalloc.constprop.1+0xc1/0xd0
[10051.051230]  kmem_cache_alloc+0x146/0x440
[10051.052060]  mempool_alloc+0x125/0x2f0
[10051.052818]  bio_alloc_bioset+0x353/0x590
[10051.053658]  mpage_alloc+0x3b/0x240
[10051.054382]  do_mpage_readpage+0xddf/0x1ef0
[10051.055250]  mpage_readahead+0x264/0x500
[10051.056060]  read_pages+0x1c1/0xbf0
[10051.056758]  page_cache_ra_unbounded+0x4c2/0x6f0
[10051.057702]  do_page_cache_ra+0xda/0x110
[10051.058511]  force_page_cache_ra+0x23d/0x3d0
[10051.059373]  page_cache_sync_ra+0xca/0x300
[10051.060198]  generic_file_buffered_read+0xbea/0x2130
[10051.061195]  generic_file_read_iter+0x315/0x490
[10051.062189]  blkdev_read_iter+0x113/0x1b0
[10051.063015]  aio_read+0x2ad/0x450
[10051.063686]  io_submit_one+0xc8e/0x1d60
[10051.064467]  __se_sys_io_submit+0x125/0x350
[10051.065318]  do_syscall_64+0x2d/0x40
[10051.066082]  entry_SYSCALL_64_after_hwframe+0x44/0xa9

[10051.067455] Freed by task 13307:
[10051.068136]  kasan_save_stack+0x19/0x40
[10051.068931]  kasan_set_track+0x1c/0x30
[10051.069726]  kasan_set_free_info+0x1b/0x30
[10051.070621]  __kasan_slab_free+0x111/0x160
[10051.071480]  kmem_cache_free+0x94/0x460
[10051.072256]  mempool_free+0xd6/0x320
[10051.072985]  bio_free+0xe0/0x130
[10051.073630]  bio_put+0xab/0xe0
[10051.074252]  bio_endio+0x3a6/0x5d0
[10051.074984]  blk_update_request+0x590/0x1370
[10051.075870]  scsi_end_request+0x7d/0x400
[10051.076667]  scsi_io_completion+0x1aa/0xe50
[10051.077503]  scsi_softirq_done+0x11b/0x240
[10051.078344]  blk_mq_complete_request+0xd4/0x120
[10051.079275]  scsi_mq_done+0xf0/0x200
[10051.080036]  virtscsi_vq_done+0xbc/0x150
[10051.080850]  vring_interrupt+0x179/0x390
[10051.081650]  __handle_irq_event_percpu+0xf7/0x490
[10051.082626]  handle_irq_event_percpu+0x7b/0x160
[10051.083527]  handle_irq_event+0xcc/0x170
[10051.084297]  handle_edge_irq+0x215/0xb20
[10051.085122]  asm_call_irq_on_stack+0xf/0x20
[10051.085986]  common_interrupt+0xae/0x120
[10051.086830]  asm_common_interrupt+0x1e/0x40

==================================================================

Bio will be checked at beginning of submit_bio_noacct(). If bio needs
to be throttled, it will start the timer and stop submit bio directly.
Bio will submit in blk_throtl_dispatch_work_fn() when the timer expires.
But in the current process, if bio is throttled, it will still set bio
issue->value by blkcg_bio_issue_init(). This is redundant and may cause
the above use-after-free.

CPU0                                   CPU1
submit_bio
submit_bio_noacct
  submit_bio_checks
    blk_throtl_bio()
      <=mod_timer(&sq->pending_timer
                                      blk_throtl_dispatch_work_fn
                                        submit_bio_noacct() <= bio have
                                        throttle tag, will throw directly
                                        and bio issue->value will be set
                                        here

                                      bio_endio()
                                      bio_put()
                                      bio_free() <= free this bio

    blkcg_bio_issue_init(bio)
      <= bio has been freed and
      will lead to UAF
  return BLK_QC_T_NONE

Fix this by remove extra blkcg_bio_issue_init.

Fixes: e439bedf6b24 (blkcg: consolidate bio_issue_init() to be a part of core)
Signed-off-by: Laibin Qiu <qiulaibin@huawei.com>
Link: https://lore.kernel.org/r/20211112093354.3581504-1-qiulaibin@huawei.com
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>

perf/x86/vlbr: Add c->flags to vlbr event constraints

[ Upstream commit 5863702561e625903ec678551cb056a4b19e0b8a ]

Just like what we do in the x86_get_event_constraints(), the
PERF_X86_EVENT_LBR_SELECT flag should also be propagated
to event->hw.flags so that the host lbr driver can save/restore
MSR_LBR_SELECT for the special vlbr event created by KVM or BPF.

Fixes: 097e4311cda9 ("perf/x86: Add constraint to create guest LBR event without hw counter")
Reported-by: Wanpeng Li <wanpengli@tencent.com>
Signed-off-by: Like Xu <likexu@tencent.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Wanpeng Li <wanpengli@tencent.com>
Link: https://lore.kernel.org/r/20211103091716.59906-1-likexu@tencent.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

sched/fair: Prevent dead task groups from regaining cfs_rq's

[ Upstream commit b027789e5e50494c2325cc70c8642e7fd6059479 ]

Kevin is reporting crashes which point to a use-after-free of a cfs_rq
in update_blocked_averages(). Initial debugging revealed that we've
live cfs_rq's (on_list=1) in an about to be kfree()'d task group in
free_fair_sched_group(). However, it was unclear how that can happen.

His kernel config happened to lead to a layout of struct sched_entity
that put the 'my_q' member directly into the middle of the object
which makes it incidentally overlap with SLUB's freelist pointer.
That, in combination with SLAB_FREELIST_HARDENED's freelist pointer
mangling, leads to a reliable access violation in form of a #GP which
made the UAF fail fast.

Michal seems to have run into the same issue[1]. He already correctly
diagnosed that commit a7b359fc6a37 ("sched/fair: Correctly insert
cfs_rq's to list on unthrottle") is causing the preconditions for the
UAF to happen by re-adding cfs_rq's also to task groups that have no
more running tasks, i.e. also to dead ones. His analysis, however,
misses the real root cause and it cannot be seen from the crash
backtrace only, as the real offender is tg_unthrottle_up() getting
called via sched_cfs_period_timer() via the timer interrupt at an
inconvenient time.

When unregister_fair_sched_group() unlinks all cfs_rq's from the dying
task group, it doesn't protect itself from getting interrupted. If the
timer interrupt triggers while we iterate over all CPUs or after
unregister_fair_sched_group() has finished but prior to unlinking the
task group, sched_cfs_period_timer() will execute and walk the list of
task groups, trying to unthrottle cfs_rq's, i.e. re-add them to the
dying task group. These will later -- in free_fair_sched_group() -- be
kfree()'ed while still being linked, leading to the fireworks Kevin
and Michal are seeing.

To fix this race, ensure the dying task group gets unlinked first.
However, simply switching the order of unregistering and unlinking the
task group isn't sufficient, as concurrent RCU walkers might still see
it, as can be seen below:

    CPU1:                                      CPU2:
      :                                        timer IRQ:
      :                                          do_sched_cfs_period_timer():
      :                                            :
      :                                            distribute_cfs_runtime():
      :                                              rcu_read_lock();
      :                                              :
      :                                              unthrottle_cfs_rq():
    sched_offline_group():                             :
      :                                                walk_tg_tree_from(…,tg_unthrottle_up,…):
      list_del_rcu(&tg->list);                           :
 (1)  :                                                  list_for_each_entry_rcu(child, &parent->children, siblings)
      :                                                    :
 (2)  list_del_rcu(&tg->siblings);                         :
      :                                                    tg_unthrottle_up():
      unregister_fair_sched_group():                         struct cfs_rq *cfs_rq = tg->cfs_rq[cpu_of(rq)];
        :                                                    :
        list_del_leaf_cfs_rq(tg->cfs_rq[cpu]);               :
        :                                                    :
        :                                                    if (!cfs_rq_is_decayed(cfs_rq) || cfs_rq->nr_running)
 (3)    :                                                        list_add_leaf_cfs_rq(cfs_rq);
      :                                                      :
      :                                                    :
      :                                                  :
      :                                                :
      :                                              :
 (4)  :                                              rcu_read_unlock();

CPU 2 walks the task group list in parallel to sched_offline_group(),
specifically, it'll read the soon to be unlinked task group entry at
(1). Unlinking it on CPU 1 at (2) therefore won't prevent CPU 2 from
still passing it on to tg_unthrottle_up(). CPU 1 now tries to unlink
all cfs_rq's via list_del_leaf_cfs_rq() in
unregister_fair_sched_group().  Meanwhile CPU 2 will re-add some of
these at (3), which is the cause of the UAF later on.

To prevent this additional race from happening, we need to wait until
walk_tg_tree_from() has finished traversing the task groups, i.e.
after the RCU read critical section ends in (4). Afterwards we're safe
to call unregister_fair_sched_group(), as each new walk won't see the
dying task group any more.

On top of that, we need to wait yet another RCU grace period after
unregister_fair_sched_group() to ensure print_cfs_stats(), which might
run concurrently, always sees valid objects, i.e. not already free'd
ones.

This patch survives Michal's reproducer[2] for 8h+ now, which used to
trigger within minutes before.

  [1] https://lore.kernel.org/lkml/20211011172236.11223-1-mkoutny@suse.com/
  [2] https://lore.kernel.org/lkml/20211102160228.GA57072@blackbody.suse.cz/

Fixes: a7b359fc6a37 ("sched/fair: Correctly insert cfs_rq's to list on unthrottle")
[peterz: shuffle code around a bit]
Reported-by: Kevin Tanguy <kevin.tanguy@corp.ovh.com>
Signed-off-by: Mathias Krause <minipli@grsecurity.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

sched/core: Mitigate race cpus_share_cache()/update_top_cache_domain()

[ Upstream commit 42dc938a590c96eeb429e1830123fef2366d9c80 ]

Nothing protects the access to the per_cpu variable sd_llc_id. When testing
the same CPU (i.e. this_cpu == that_cpu), a race condition exists with
update_top_cache_domain(). One scenario being:

              CPU1                            CPU2
  ==================================================================

  per_cpu(sd_llc_id, CPUX) => 0
                                    partition_sched_domains_locked()
             detach_destroy_domains()
  cpus_share_cache(CPUX, CPUX)          update_top_cache_domain(CPUX)
    per_cpu(sd_llc_id, CPUX) => 0
                                          per_cpu(sd_llc_id, CPUX) = CPUX
    per_cpu(sd_llc_id, CPUX) => CPUX
    return false

ttwu_queue_cond() wouldn't catch smp_processor_id() == cpu and the result
is a warning triggered from ttwu_queue_wakelist().

Avoid a such race in cpus_share_cache() by always returning true when
this_cpu == that_cpu.

Fixes: 518cd6234178 ("sched: Only queue remote wakeups when crossing cache boundaries")
Reported-by: Jing-Ting Wu <jing-ting.wu@mediatek.com>
Signed-off-by: Vincent Donnefort <vincent.donnefort@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Valentin Schneider <valentin.schneider@arm.com>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Link: https://lore.kernel.org/r/20211104175120.857087-1-vincent.donnefort@arm.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

MIPS: boot/compressed/: add __bswapdi2() to target for ZSTD decompression

[ Upstream commit e2f4b3be1d3c73176db734565b160250cc1300dd ]

For MIPS pre-boot, when CONFIG_KERNEL_ZSTD=y, the decompressor
function uses __bswapdi2(), so this object file should be added to
the target object file.

Fixes these build errors:

mips-linux-ld: arch/mips/boot/compressed/decompress.o: in function `xxh64':
decompress.c:(.text+0x8be0): undefined reference to `__bswapdi2'
mips-linux-ld: decompress.c:(.text+0x8c78): undefined reference to `__bswapdi2'
mips-linux-ld: decompress.c:(.text+0x8d04): undefined reference to `__bswapdi2'
mips-linux-ld: arch/mips/boot/compressed/decompress.o:decompress.c:(.text+0xa010): more undefined references to `__bswapdi2' follow

Fixes: 0652035a5794 ("asm-generic: unaligned: remove byteshift helpers")
Fixes: cddc40f5617e ("mips: always link byteswap helpers into decompressor")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Reported-by: kernel test robot <lkp@intel.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: linux-mips@vger.kernel.org
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

mips: BCM63XX: ensure that CPU_SUPPORTS_32BIT_KERNEL is set

[ Upstream commit 5eeaafc8d69373c095e461bdb39e5c9b62228ac5 ]

Several header files need info on CONFIG_32BIT or CONFIG_64BIT,
but kconfig symbol BCM63XX does not provide that info. This leads
to many build errors, e.g.:

   arch/mips/include/asm/page.h:196:13: error: use of undeclared identifier 'CAC_BASE'
           return x - PAGE_OFFSET + PHYS_OFFSET;
   arch/mips/include/asm/mach-generic/spaces.h:91:23: note: expanded from macro 'PAGE_OFFSET'
   #define PAGE_OFFSET             (CAC_BASE + PHYS_OFFSET)
   arch/mips/include/asm/io.h:134:28: error: use of undeclared identifier 'CAC_BASE'
           return (void *)(address + PAGE_OFFSET - PHYS_OFFSET);
   arch/mips/include/asm/mach-generic/spaces.h:91:23: note: expanded from macro 'PAGE_OFFSET'
   #define PAGE_OFFSET             (CAC_BASE + PHYS_OFFSET)

arch/mips/include/asm/uaccess.h:82:10: error: use of undeclared identifier '__UA_LIMIT'
           return (__UA_LIMIT & (addr | (addr + size) | __ua_size(size))) == 0;

Selecting the SYS_HAS_CPU_BMIPS* symbols causes SYS_HAS_CPU_BMIPS to be
set, which then selects CPU_SUPPORT_32BIT_KERNEL, which causes
CONFIG_32BIT to be set. (a bit more indirect than v1 [RFC].)

Fixes: e7300d04bd08 ("MIPS: BCM63xx: Add support for the Broadcom BCM63xx family of SOCs.")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Reported-by: kernel test robot <lkp@intel.com>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Florian Fainelli <f.fainelli@gmail.com>
Cc: bcm-kernel-feedback-list@broadcom.com
Cc: linux-mips@vger.kernel.org
Cc: Paul Burton <paulburton@kernel.org>
Cc: Maxime Bizon <mbizon@freebox.fr>
Cc: Ralf Baechle <ralf@linux-mips.org>
Suggested-by: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

KVM: arm64: Fix host stage-2 finalization

[ Upstream commit 50a8d3315960c74095c59e204db44abd937d4b5d ]

We currently walk the hypervisor stage-1 page-table towards the end of
hyp init in nVHE protected mode and adjust the host page ownership
attributes in its stage-2 in order to get a consistent state from both
point of views. The walk is done on the entire hyp VA space, and expects
to only ever find page-level mappings. While this expectation is
reasonable in the half of hyp VA space that maps memory with a fixed
offset (see the loop in pkvm_create_mappings_locked()), it can be
incorrect in the other half where nothing prevents the usage of block
mappings. For instance, on systems where memory is physically aligned at
an address that happens to maps to a PMD aligned VA in the hyp_vmemmap,
kvm_pgtable_hyp_map() will install block mappings when backing the
hyp_vmemmap, which will later cause finalize_host_mappings() to fail.
Furthermore, it should be noted that all pages backing the hyp_vmemmap
are also mapped in the 'fixed offset range' of the hypervisor, which
implies that finalize_host_mappings() will walk both aliases and update
the host stage-2 attributes twice. The order in which this happens is
unpredictable, though, since the hyp VA layout is highly dependent on
the position of the idmap page, hence resulting in a fragile mess at
best.

In order to fix all of this, let's restrict the finalization walk to
only cover memory regions in the 'fixed-offset range' of the hyp VA
space and nothing else. This not only fixes a correctness issue, but
will also result in a slighlty faster hyp initialization overall.

Fixes: 2c50166c62ba ("KVM: arm64: Mark host bss and rodata section as shared")
Signed-off-by: Quentin Perret <qperret@google.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20211108154636.393384-1-qperret@google.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

clk: qcom: gcc-msm8996: Drop (again) gcc_aggre1_pnoc_ahb_clk

[ Upstream commit 05cf3ec00d460b50088d421fb878a0f83f57e262 ]

The gcc_aggre1_pnoc_ahb_clk is crucial for the proper MSM8996/APQ8096
functioning. If it gets disabled, several subsytems will stop working
(including eMMC/SDCC and USB). There are no in-kernel users of this
clock, so it is much simpler to remove from the kernel.

The clock was first removed in the commit 9e60de1cf270 ("clk: qcom:
Remove gcc_aggre1_pnoc_ahb_clk from msm8996") by Stephen Boyd, but got
added back in the commit b567752144e3 ("clk: qcom: Add some missing gcc
clks for msm8996") by Rajendra Nayak.

Let's remove it again in hope that nobody adds it back.

Reported-by: Vladimir Zapolskiy <vladimir.zapolskiy@linaro.org>
Cc: Rajendra Nayak <rnayak@codeaurora.org>
Cc: Konrad Dybcio <konrad.dybcio@somainline.org>
Fixes: b567752144e3 ("clk: qcom: Add some missing gcc clks for msm8996")
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Link: https://lore.kernel.org/r/20211104011155.2209654-1-dmitry.baryshkov@linaro.org
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

clk/ast2600: Fix soc revision for AHB

[ Upstream commit f45c5b1c27293f834682e89003f88b3512329ab4 ]

Move the soc revision parsing to the initial probe, saving the driver
from parsing the register multiple times.

Use this variable to select the correct divisor table for the AHB clock.
Before this fix the A2 would have used the A0 table.

Fixes: 2d491066ccd4 ("clk: ast2600: Fix AHB clock divider for A1")
Signed-off-by: Joel Stanley <joel@jms.id.au>
Link: https://lore.kernel.org/r/20210922235449.213631-1-joel@jms.id.au
Reviewed-by: Andrew Jeffery <andrew@aj.id.au>
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

clk: ingenic: Fix bugs with divided dividers

[ Upstream commit ed84ef1cd7eddf933d4ffce2caa8161d6f947245 ]

Two fixes in one:

- In the "impose hardware constraints" block, the "logical" divider
  value (aka. not translated to the hardware) was clamped to fit in the
  register area, but this totally ignored the fact that the divider
  value can itself have a fixed divider.

- The code that made sure that the divider value returned by the
  function was a multiple of its own fixed divider could result in a
  wrong value being calculated, because it was rounded down instead of
  rounded up.

Fixes: 4afe2d1a6ed5 ("clk: ingenic: Allow divider value to be divided")
Co-developed-by: Artur Rojek <contact@artur-rojek.eu>
Signed-off-by: Artur Rojek <contact@artur-rojek.eu>
Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Link: https://lore.kernel.org/r/20211001172033.122329-1-paul@crapouillou.net
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

f2fs: fix incorrect return value in f2fs_sanity_check_ckpt()

[ Upstream commit ca98d72141dd81f42893a9a43d7ededab3355fba ]

As Pavel Machek reported in [1]

This code looks quite confused: part of function returns 1 on
corruption, part returns -errno. The problem is not stable-specific.

[1] https://lkml.org/lkml/2021/9/19/207

Let's fix to make 'insane cp_payload case' to return 1 rater than
EFSCORRUPTED, so that return value can be kept consistent for all
error cases, it can avoid confusion of code logic.

Fixes: 65ddf6564843 ("f2fs: fix to do sanity check for sb/cp fields correctly")
Reported-by: Pavel Machek <pavel@denx.de>
Reviewed-by: Pavel Machek <pavel@denx.de>
Signed-off-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

f2fs: compress: disallow disabling compress on non-empty compressed file

[ Upstream commit 02d58cd253d7536c412993573fc6b3b4454960eb ]

Compresse file and normal file has differ in i_addr addressing,
specifically addrs per inode/block. So, we will face data loss, if we
disable the compression flag on non-empty files. Therefore we should
disallow not only enabling but disabling the compression flag on
non-empty files.

Fixes: 4c8ff7095bef ("f2fs: support data compression")
Signed-off-by: Sungjong Seo <sj1557.seo@samsung.com>
Signed-off-by: Hyeong-Jun Kim <hj514.kim@samsung.com>
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

sh: define __BIG_ENDIAN for math-emu

[ Upstream commit b929926f01f2d14635345d22eafcf60feed1085e ]

Fix this by defining both ENDIAN macros in
<asm/sfp-machine.h> so that they can be utilized in
<math-emu/soft-fp.h> according to the latter's comment:
/* Allow sfp-machine to have its own byte order definitions. */

(This is what is done in arch/nds32/include/asm/sfp-machine.h.)

This placates these build warnings:

In file included from ../arch/sh/math-emu/math.c:23:
.../include/math-emu/single.h:50:21: warning: "__BIG_ENDIAN" is not defined, evaluates to 0 [-Wundef]
   50 | #if __BYTE_ORDER == __BIG_ENDIAN
In file included from ../arch/sh/math-emu/math.c:24:
.../include/math-emu/double.h:59:21: warning: "__BIG_ENDIAN" is not defined, evaluates to 0 [-Wundef]
   59 | #if __BYTE_ORDER == __BIG_ENDIAN

Fixes: 4b565680d163 ("sh: math-emu support")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Cc: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Tested-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>
Signed-off-by: Rich Felker <dalias@libc.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

sh: math-emu: drop unused functions

[ Upstream commit e25c252a9b033523c626f039d4b9a304f12f6775 ]

Delete ieee_fpe_handler() since it is not used. After that is done,
delete denormal_to_double() since it is not used:

.../arch/sh/math-emu/math.c:505:12: error: 'ieee_fpe_handler' defined but not used [-Werror=unused-function]
  505 | static int ieee_fpe_handler(struct pt_regs *regs)

.../arch/sh/math-emu/math.c:477:13: error: 'denormal_to_double' defined but not used [-Werror=unused-function]
  477 | static void denormal_to_double(struct sh_fpu_soft_struct *fpu, int n)

Fixes: 7caf62de25554da3 ("sh: remove unused do_fpu_error")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Takashi YOSHII <takasi-y@ops.dti.ne.jp>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Cc: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Rich Felker <dalias@libc.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

sh: fix kconfig unmet dependency warning for FRAME_POINTER

[ Upstream commit fda1bc533094a7db68b11e7503d2c6c73993d12a ]

FRAME_POINTER depends on DEBUG_KERNEL so DWARF_UNWINDER should
depend on DEBUG_KERNEL before selecting FRAME_POINTER.

WARNING: unmet direct dependencies detected for FRAME_POINTER
  Depends on [n]: DEBUG_KERNEL [=n] && (M68K || UML || SUPERH [=y]) || ARCH_WANT_FRAME_POINTERS [=n]
  Selected by [y]:
  - DWARF_UNWINDER [=y]

Fixes: bd353861c735 ("sh: dwarf unwinder support.")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Matt Fleming <matt@console-pimps.org>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Cc: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>
Cc: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Tested-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>
Signed-off-by: Rich Felker <dalias@libc.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

f2fs: fix wrong condition to trigger background checkpoint correctly

[ Upstream commit cd6d697a6e2013a0a85f8b261b16c8cfd50c1f5f ]

In f2fs_balance_fs_bg(), it needs to check both NAT_ENTRIES and INO_ENTRIES
memory usage to decide whether we should skip background checkpoint, otherwise
we may always skip checking INO_ENTRIES memory usage, so that INO_ENTRIES may
potentially cause high memory footprint.

Fixes: 493720a48543 ("f2fs: fix to avoid REQ_TIME and CP_TIME collision")
Signed-off-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

f2fs: fix to use WHINT_MODE

[ Upstream commit 011e0868e0cf1237675b22e36fffa958fb08f46e ]

Since active_logs can be set to 2 or 4 or NR_CURSEG_PERSIST_TYPE(6),
it cannot be set to NR_CURSEG_TYPE(8).
That is, whint_mode is always off.

Therefore, the condition is changed from NR_CURSEG_TYPE to NR_CURSEG_PERSIST_TYPE.

Cc: Chao Yu <chao@kernel.org>
Fixes: d0b9e42ab615 (f2fs: introduce inmem curseg)
Reported-by: tanghuan <tanghuan@vivo.com>
Signed-off-by: Keoseong Park <keosung.park@samsung.com>
Signed-off-by: Fengnan Chang <changfengnan@vivo.com>
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

f2fs: fix up f2fs_lookup tracepoints

[ Upstream commit 70a9ac36ffd807ac506ed0b849f3e8ce3c6623f2 ]

Fix up a misuse that the filename pointer isn't always valid in
the ring buffer, and we should copy the content instead.

Fixes: 0c5e36db17f5 ("f2fs: trace f2fs_lookup")
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

maple: fix wrong return value of maple_bus_init().

[ Upstream commit bde82ee391fa6d3ad054313c4aa7b726d32515ce ]

If KMEM_CACHE or maple_alloc_dev failed, the maple_bus_init() will return 0
rather than error, because the retval is not changed after KMEM_CACHE or
maple_alloc_dev failed.

Fixes: 17be2d2b1c33 ("sh: Add maple bus support for the SEGA Dreamcast.")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Lu Wei <luwei32@huawei.com>
Acked-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>
Signed-off-by: Rich Felker <dalias@libc.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

sh: check return code of request_irq

[ Upstream commit 0e38225c92c7964482a8bb6b3e37fde4319e965c ]

request_irq is marked __must_check, but the call in shx3_prepare_cpus
has a void return type, so it can't propagate failure to the caller.
Follow cues from hexagon and just print an error.

Fixes: c7936b9abcf5 ("sh: smp: Hook in to the generic IPI handler for SH-X3 SMP.")
Cc: Miguel Ojeda <miguel.ojeda.sandonis@gmail.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Reported-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Tested-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>
Reviewed-by: Miguel Ojeda <ojeda@kernel.org>
Signed-off-by: Rich Felker <dalias@libc.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

powerpc/8xx: Fix Oops with STRICT_KERNEL_RWX without DEBUG_RODATA_TEST

[ Upstream commit c12ab8dbc492b992e1ea717db933cee568780c47 ]

Until now, all tests involving CONFIG_STRICT_KERNEL_RWX were done with
DEBUG_RODATA_TEST to check the result. But now that
CONFIG_STRICT_KERNEL_RWX is selected by default, it came without
CONFIG_DEBUG_RODATA_TEST and led to the following Oops

[    6.830908] Freeing unused kernel image (initmem) memory: 352K
[    6.840077] BUG: Unable to handle kernel data access on write at 0xc1285200
[    6.846836] Faulting instruction address: 0xc0004b6c
[    6.851745] Oops: Kernel access of bad area, sig: 11 [#1]
[    6.857075] BE PAGE_SIZE=16K PREEMPT CMPC885
[    6.861348] SAF3000 DIE NOTIFICATION
[    6.864830] CPU: 0 PID: 1 Comm: swapper Not tainted 5.15.0-rc5-s3k-dev-02255-g2747d7b7916f #451
[    6.873429] NIP:  c0004b6c LR: c0004b60 CTR: 00000000
[    6.878419] REGS: c902be60 TRAP: 0300   Not tainted  (5.15.0-rc5-s3k-dev-02255-g2747d7b7916f)
[    6.886852] MSR:  00009032 <EE,ME,IR,DR,RI>  CR: 53000335  XER: 8000ff40
[    6.893564] DAR: c1285200 DSISR: 82000000
[    6.893564] GPR00: 0c000000 c902bf20 c20f4000 08000000 00000001 04001f00 c1800000 00000035
[    6.893564] GPR08: ff0001ff c1280000 00000002 c0004b60 00001000 00000000 c0004b1c 00000000
[    6.893564] GPR16: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
[    6.893564] GPR24: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 c1060000
[    6.932034] NIP [c0004b6c] kernel_init+0x50/0x138
[    6.936682] LR [c0004b60] kernel_init+0x44/0x138
[    6.941245] Call Trace:
[    6.943653] [c902bf20] [c0004b60] kernel_init+0x44/0x138 (unreliable)
[    6.950022] [c902bf30] [c001122c] ret_from_kernel_thread+0x5c/0x64
[    6.956135] Instruction dump:
[    6.959060] 48ffc521 48045469 4800d8cd 3d20c086 89295fa0 2c090000 41820058 480796c9
[    6.966890] 4800e48d 3d20c128 39400002 3fe0c106 <91495200> 3bff8000 4806fa1d 481f7d75
[    6.974902] ---[ end trace 1e397bacba4aa610 ]---

0xc1285200 corresponds to 'system_state' global var that the kernel is trying to set to
SYSTEM_RUNNING. This var is above the RO/RW limit so it shouldn't Oops.

It oopses because the dirty bit is missing.

Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/3d5800b0bbcd7b19761b98f50421358667b45331.1635520232.git.christophe.leroy@csgroup.eu
Signed-off-by: Sasha Levin <sashal@kernel.org>

powerpc/dcr: Use cmplwi instead of 3-argument cmpli

[ Upstream commit fef071be57dc43679a32d5b0e6ee176d6f12e9f2 ]

In dcr-low.S we use cmpli with three arguments, instead of four
arguments as defined in the ISA:

cmpli cr0,r3,1024

This appears to be a PPC440-ism, looking at the "PPC440x5 CPU Core
User’s Manual" it shows cmpli having no L field, but implied to be 0 due
to the core being 32-bit. It mentions that the ISA defines four
arguments and recommends using cmplwi.

It also corresponds to the old POWER instruction set, which had no L
field there, a reserved bit instead.

dcr-low.S is only built 32-bit, because it is only built when
DCR_NATIVE=y, which is only selected by 40x and 44x. Looking at the
generated code (with gcc/gas) we see cmplwi as expected.

Although gas is happy with the 3-argument version when building for
32-bit, the LLVM assembler is not and errors out with:

  arch/powerpc/sysdev/dcr-low.S:27:10: error: invalid operand for instruction
   cmpli 0,%r3,1024; ...
           ^

Switch to the cmplwi extended opcode, which avoids any confusion when
reading the ISA, fixes the issue with the LLVM assembler, and also means
the code could be built 64-bit in future (though that's very unlikely).

Reported-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
BugLink: https://github.com/ClangBuiltLinux/linux/issues/1419
Link: https://lore.kernel.org/r/20211014024424.528848-1-mpe@ellerman.id.au
Signed-off-by: Sasha Levin <sashal@kernel.org>

iommu/dart: Initialize DART_STREAMS_ENABLE

[ Upstream commit 5a009fc1364170b240a4d351b345e69bb3728b3e ]

DART has an additional global register to control which streams are
isolated. This register is a bit redundant since DART_TCR can already
be used to control isolation and is usually initialized to DART_STREAM_ALL
by the time we get control. Some DARTs (namely the one used for the audio
controller) however have some streams disabled initially. Make sure those
work by initializing DART_STREAMS_ENABLE during reset.

Reported-by: Martin Povišer <povik@protonmail.com>
Signed-off-by: Sven Peter <sven@svenpeter.dev>
Reviewed-by: Hector Martin <marcan@marcan.st>
Link: https://lore.kernel.org/r/20211019162253.45919-1-sven@svenpeter.dev
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

clk: at91: sama7g5: remove prescaler part of master clock

[ Upstream commit facb87ad75603813bc3b1314f5a87377f020fcb8 ]

On SAMA7G5 the prescaler part of master clock has been implemented as a
changeable one. Everytime the prescaler is changed the PMC_SR.MCKRDY bit
must be polled. Value 1 for PMC_SR.MCKRDY means the prescaler update is
done. Driver polls for this bit until it becomes 1. On SAMA7G5 it has
been discovered that in some conditions the PMC_SR.MCKRDY is not rising
but the rate it provides it's stable. The workaround is to add a timeout
when polling for PMC_SR.MCKRDY. At the moment, for SAMA7G5, the prescaler
will be removed from Linux clock tree as all the frequencies for CPU could
be obtained from PLL and also there will be less overhead when changing
frequency via DVFS.

Signed-off-by: Claudiu Beznea <claudiu.beznea@microchip.com>
Link: https://lore.kernel.org/r/20211011112719.3951784-14-claudiu.beznea@microchip.com
Acked-by: Nicolas Ferre <nicolas.ferre@microchip.com>
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ALSA: usb-audio: fix null pointer dereference on pointer cs_desc

[ Upstream commit b97053df0f04747c3c1e021ecbe99db675342954 ]

The pointer cs_desc return from snd_usb_find_clock_source could
be null, so there is a potential null pointer dereference issue.
Fix this by adding a null check before dereference.

Signed-off-by: Chengfeng Ye <cyeaa@connect.ust.hk>
Link: https://lore.kernel.org/r/20211024111736.11342-1-cyeaa@connect.ust.hk
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ALSA: gus: fix null pointer dereference on pointer block

[ Upstream commit a0d21bb3279476c777434c40d969ea88ca64f9aa ]

The pointer block return from snd_gf1_dma_next_block could be
null, so there is a potential null pointer dereference issue.
Fix this by adding a null check before dereference.

Signed-off-by: Chengfeng Ye <cyeaa@connect.ust.hk>
Link: https://lore.kernel.org/r/20211024104611.9919-1-cyeaa@connect.ust.hk
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

arm64: dts: qcom: Fix node name of rpm-msg-ram device nodes

[ Upstream commit 179811bebc7b91e0f9d0adee9bfa3d2af9c43869 ]

According to the new DT schema for qcom,rpm-msg-ram the node name
should be sram@. memory@ is reserved for definition of physical RAM
(usable by Linux).

This fixes the following dtbs_check error on various device trees:
memory@60000: 'device_type' is a required property
        From schema: dtschema/schemas/memory.yaml

Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Link: https://lore.kernel.org/r/20211018110009.30837-1-stephan@gerhold.net
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: dts: qcom: fix memory and mdio nodes naming for RB3011

[ Upstream commit 14a1f6c9d8017ffbf388e82e1a1f023196d98612 ]

Fixes warnings regarding to memory and mdio nodes and
apply new naming following dt-schema.

Signed-off-by: David Heidelberg <david@ixit.cz>
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Link: https://lore.kernel.org/r/20211020214741.261509-1-david@ixit.cz
Signed-off-by: Sasha Levin <sashal@kernel.org>

powerpc/5200: dts: fix memory node unit name

[ Upstream commit aed2886a5e9ffc8269a4220bff1e9e030d3d2eb1 ]

Fixes build warnings:
Warning (unit_address_vs_reg): /memory: node has a reg or ranges property, but no unit name

Signed-off-by: Anatolij Gustschin <agust@denx.de>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20211013220532.24759-4-agust@denx.de
Signed-off-by: Sasha Levin <sashal@kernel.org>

memory: tegra20-emc: Add runtime dependency on devfreq governor module

[ Upstream commit 14b43c20c283de36131da0cb44f3170b9ffa7630 ]

Tegra20 EMC driver uses simple devfreq governor. Add simple devfreq
governor to the list of the Tegra20 EMC driver module softdeps to allow
userspace initramfs tools like dracut to automatically pull the devfreq
module into ramfs image together with the EMC module.

Reported-by: Nicolas Chauvet <kwizart@gmail.com>
Suggested-by: Nicolas Chauvet <kwizart@gmail.com>
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Link: https://lore.kernel.org/r/20211019231524.888-1-digetx@gmail.com
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@canonical.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

scsi: lpfc: Allow fabric node recovery if recovery is in progress before devloss

[ Upstream commit af984c87293b19dccbd0b16afc57c5c9a4a279c7 ]

A link bounce to a slow fabric may observe FDISC response delays lasting
longer than devloss tmo.  Current logic decrements the final fabric node
kref during a devloss tmo event.  This results in a NULL ptr dereference
crash if the FDISC completes for that fabric node after devloss tmo.

Fix by adding the NLP_IN_RECOV_POST_DEV_LOSS flag, which is set when
devloss tmo triggers and we've noticed that fabric node recovery has
already started or finished in between the time lpfc_dev_loss_tmo_callbk
queues lpfc_dev_loss_tmo_handler.  If fabric node recovery succeeds, then
the driver reverses the devloss tmo marked kref put with a kref get.  If
fabric node recovery fails, then the final kref put relies on the ELS
timing out or the REG_LOGIN cmpl routine.

Link: https://lore.kernel.org/r/20211020211417.88754-8-jsmart2021@gmail.com
Co-developed-by: Justin Tee <justin.tee@broadcom.com>
Signed-off-by: Justin Tee <justin.tee@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

scsi: lpfc: Fix link down processing to address NULL pointer dereference

[ Upstream commit 1854f53ccd88ad4e7568ddfafafffe71f1ceb0a6 ]

If an FC link down transition while PLOGIs are outstanding to fabric well
known addresses, outstanding ABTS requests may result in a NULL pointer
dereference. Driver unload requests may hang with repeated "2878" log
messages.

The Link down processing results in ABTS requests for outstanding ELS
requests. The Abort WQEs are sent for the ELSs before the driver had set
the link state to down. Thus the driver is sending the Abort with the
expectation that an ABTS will be sent on the wire. The Abort request is
stalled waiting for the link to come up. In some conditions the driver may
auto-complete the ELSs thus if the link does come up, the Abort completions
may reference an invalid structure.

Fix by ensuring that Abort set the flag to avoid link traffic if issued due
to conditions where the link failed.

Link: https://lore.kernel.org/r/20211020211417.88754-7-jsmart2021@gmail.com
Co-developed-by: Justin Tee <justin.tee@broadcom.com>
Signed-off-by: Justin Tee <justin.tee@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

scsi: lpfc: Fix use-after-free in lpfc_unreg_rpi() routine

[ Upstream commit 79b20beccea3a3938a8500acef4e6b9d7c66142f ]

An error is detected with the following report when unloading the driver:
  "KASAN: use-after-free in lpfc_unreg_rpi+0x1b1b"

The NLP_REG_LOGIN_SEND nlp_flag is set in lpfc_reg_fab_ctrl_node(), but the
flag is not cleared upon completion of the login.

This allows a second call to lpfc_unreg_rpi() to proceed with nlp_rpi set
to LPFC_RPI_ALLOW_ERROR.  This results in a use after free access when used
as an rpi_ids array index.

Fix by clearing the NLP_REG_LOGIN_SEND nlp_flag in
lpfc_mbx_cmpl_fc_reg_login().

Link: https://lore.kernel.org/r/20211020211417.88754-5-jsmart2021@gmail.com
Co-developed-by: Justin Tee <justin.tee@broadcom.com>
Signed-off-by: Justin Tee <justin.tee@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

RDMA/core: Use kvzalloc when allocating the struct ib_port

[ Upstream commit 911a81c9c7092bfd75432ce79b2ef879127ea065 ]

The 'struct attribute' flex array contains some struct lock_class_key's
which become big when lockdep is turned on. Big enough that some drivers
will not load when CONFIG_PROVE_LOCKING=y because they cannot allocate
enough memory:

 WARNING: CPU: 36 PID: 8 at mm/page_alloc.c:5350 __alloc_pages+0x27e/0x3e0
  Call Trace:
   kmalloc_order+0x2a/0xb0
   kmalloc_order_trace+0x19/0xf0
   __kmalloc+0x231/0x270
   ib_setup_port_attrs+0xd8/0x870 [ib_core]
   ib_register_device+0x419/0x4e0 [ib_core]
   bnxt_re_task+0x208/0x2d0 [bnxt_re]

Link: https://lore.kernel.org/r/20211019002656.17745-1-wangyugui@e16-tech.com
Signed-off-by: wangyugui <wangyugui@e16-tech.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

iio: imu: st_lsm6dsx: Avoid potential array overflow in st_lsm6dsx_set_odr()

[ Upstream commit 94be878c882d8d784ff44c639bf55f3b029f85af ]

The length of hw->settings->odr_table is 2 and ref_sensor->id is an enum
variable whose value is between 0 and 5.
However, the value ST_LSM6DSX_ID_MAX (i.e. 5) is not caught properly in
 switch (sensor->id) {

If ref_sensor->id is ST_LSM6DSX_ID_MAX, an array overflow will ocurrs in
function st_lsm6dsx_check_odr():
  odr_table = &sensor->hw->settings->odr_table[sensor->id];

and in function st_lsm6dsx_set_odr():
  reg = &hw->settings->odr_table[ref_sensor->id].reg;

To avoid this array overflow, handle ST_LSM6DSX_ID_GYRO explicitly and
return -EINVAL for the default case.

The enum value ST_LSM6DSX_ID_MAX is only present as an easy way to check
the limit and as such is never used, however this is not locally obvious.

Reported-by: TOTE Robot <oslab@tsinghua.edu.cn>
Signed-off-by: Teng Qi <starmiku1207184332@gmail.com>
Acked-by: Lorenzo Bianconi <lorenzo@kernel.org>
Link: https://lore.kernel.org/r/20211011114003.976221-1-starmiku1207184332@gmail.com
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

scsi: target: Fix alua_tg_pt_gps_count tracking

[ Upstream commit 1283c0d1a32bb924324481586b5d6e8e76f676ba ]

We can't free the tg_pt_gp in core_alua_set_tg_pt_gp_id() because it's
still accessed via configfs. Its release must go through the normal
configfs/refcount process.

The max alua_tg_pt_gps_count check should probably have been done in
core_alua_allocate_tg_pt_gp(), but with the current code userspace could
have created 0x0000ffff + 1 groups, but only set the id for 0x0000ffff.
Then it could have deleted a group with an ID set, and then set the ID for
that extra group and it would work ok.

It's unlikely, but just in case this patch continues to allow that type of
behavior, and just fixes the kfree() while in use bug.

Link: https://lore.kernel.org/r/20210930020422.92578-4-michael.christie@oracle.com
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>