RDMA/hns: Fix endian issue in trace events

Avoid using __le32 directly in trace events to fix sparse complains:

drivers/infiniband/hw/hns/./hns_roce_trace.h:48:1: sparse: sparse: cast to restricted __le32
drivers/infiniband/hw/hns/./hns_roce_trace.h:48:1: sparse: sparse: restricted __le32 degrades to integer
drivers/infiniband/hw/hns/./hns_roce_trace.h:48:1: sparse: sparse: restricted __le32 degrades to integer
drivers/infiniband/hw/hns/./hns_roce_trace.h:90:1: sparse: sparse: cast to restricted __le32
drivers/infiniband/hw/hns/./hns_roce_trace.h:90:1: sparse: sparse: restricted __le32 degrades to integer
drivers/infiniband/hw/hns/./hns_roce_trace.h:90:1: sparse: sparse: restricted __le32 degrades to integer
drivers/infiniband/hw/hns/./hns_roce_trace.h:173:1: sparse: sparse: cast to restricted __le32
drivers/infiniband/hw/hns/./hns_roce_trace.h:173:1: sparse: sparse: restricted __le32 degrades to integer
drivers/infiniband/hw/hns/./hns_roce_trace.h:173:1: sparse: sparse: restricted __le32 degrades to integer

Fixes: 6c98c8670806 ("RDMA/hns: Add trace for WQE dumping")
Fixes: 1e63e2f96613 ("RDMA/hns: Add trace for AEQE dumping")
Fixes: 6bd18dabf1c9 ("RDMA/hns: Add trace for CMDQ dumping")
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202505170327.TNOpreil-lkp@intel.com/
Signed-off-by: Junxian Huang <huangjunxian6@hisilicon.com>
Link: https://patch.msgid.link/20250523023433.2171003-1-huangjunxian6@hisilicon.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/mlx5: Avoid flexible array warning

The following warning is reported by sparse tool:
drivers/infiniband/hw/mlx5/fs.c:1664:26: warning: array of flexible
structures

Avoid it by simply splitting array into two separate structs.

Link: https://patch.msgid.link/7b891b96a9fc053d01284c184d25ae98d35db2d4.1747827041.git.leon@kernel.org
Reviewed-by: Zhu Yanjun <yanjun.zhu@linux.dev>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>

IB/cm: Remove dead code and adjust naming

Drop ib_send_cm_mra parameters which are always constant. Remove branch
which is never taken. Adjust name to ib_prepare_cm_mra, which better
reflects its functionality - no MRA is actually sent. Adjust name of
related tracepoints. Push setting of the constant service timeout to
cm.c and drop IB_CM_MRA_FLAG_DELAY.

Signed-off-by: Vlad Dumitrescu <vdumitrescu@nvidia.com>
Reviewed-by: Sean Hefty <shefty@nvidia.com>
Link: https://patch.msgid.link/cdd2a237acf2b495c19ce02e4b1c42c41c6751c2.1747827207.git.leon@kernel.org
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/core: Avoid hmm_dma_map_alloc() for virtual DMA devices

Drivers such as rxe, which use virtual DMA, must not call into the DMA
mapping core since they lack physical DMA capabilities. Otherwise, a NULL
pointer dereference is observed as shown below. This patch ensures the RDMA
core handles virtual and physical DMA paths appropriately.

This fixes the following kernel oops:

 BUG: kernel NULL pointer dereference, address: 00000000000002fc
 #PF: supervisor read access in kernel mode
 #PF: error_code(0x0000) - not-present page
 PGD 1028eb067 P4D 1028eb067 PUD 105da0067 PMD 0
 Oops: Oops: 0000 [#1] SMP NOPTI
 CPU: 3 UID: 1000 PID: 1854 Comm: python3 Tainted: G        W           6.15.0-rc1+ #11 PREEMPT(voluntary)
 Tainted: [W]=WARN
 Hardware name: Trigkey Key N/Key N, BIOS KEYN101 09/02/2024
 RIP: 0010:hmm_dma_map_alloc+0x25/0x100
 Code: 90 90 90 90 90 0f 1f 44 00 00 55 48 89 e5 41 57 41 56 49 89 d6 49 c1 e6 0c 41 55 41 54 53 49 39 ce 0f 82 c6 00 00 00 49 89 fc <f6> 87 fc 02 00 00 20 0f 84 af 00 00 00 49 89 f5 48 89 d3 49 89 cf
 RSP: 0018:ffffd3d3420eb830 EFLAGS: 00010246
 RAX: 0000000000001000 RBX: ffff8b727c7f7400 RCX: 0000000000001000
 RDX: 0000000000000001 RSI: ffff8b727c7f74b0 RDI: 0000000000000000
 RBP: ffffd3d3420eb858 R08: 0000000000000000 R09: 0000000000000000
 R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000000
 R13: 00007262a622a000 R14: 0000000000001000 R15: ffff8b727c7f74b0
 FS:  00007262a62a1080(0000) GS:ffff8b762ac3e000(0000) knlGS:0000000000000000
 CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
 CR2: 00000000000002fc CR3: 000000010a1f0004 CR4: 0000000000f72ef0
 PKRU: 55555554
 Call Trace:
  <TASK>
  ib_init_umem_odp+0xb6/0x110 [ib_uverbs]
  ib_umem_odp_get+0xf0/0x150 [ib_uverbs]
  rxe_odp_mr_init_user+0x71/0x170 [rdma_rxe]
  rxe_reg_user_mr+0x217/0x2e0 [rdma_rxe]
  ib_uverbs_reg_mr+0x19e/0x2e0 [ib_uverbs]
  ib_uverbs_handler_UVERBS_METHOD_INVOKE_WRITE+0xd9/0x150 [ib_uverbs]
  ib_uverbs_cmd_verbs+0xd19/0xee0 [ib_uverbs]
  ? mmap_region+0x63/0xd0
  ? __pfx_ib_uverbs_handler_UVERBS_METHOD_INVOKE_WRITE+0x10/0x10 [ib_uverbs]
  ib_uverbs_ioctl+0xba/0x130 [ib_uverbs]
  __x64_sys_ioctl+0xa4/0xe0
  x64_sys_call+0x1178/0x2660
  do_syscall_64+0x7e/0x170
  ? syscall_exit_to_user_mode+0x4e/0x250
  ? do_syscall_64+0x8a/0x170
  ? do_syscall_64+0x8a/0x170
  ? syscall_exit_to_user_mode+0x4e/0x250
  ? do_syscall_64+0x8a/0x170
  ? syscall_exit_to_user_mode+0x4e/0x250
  ? do_syscall_64+0x8a/0x170
  ? do_user_addr_fault+0x1d2/0x8d0
  ? irqentry_exit_to_user_mode+0x43/0x250
  ? irqentry_exit+0x43/0x50
  ? exc_page_fault+0x93/0x1d0
  entry_SYSCALL_64_after_hwframe+0x76/0x7e
 RIP: 0033:0x7262a6124ded
 Code: 04 25 28 00 00 00 48 89 45 c8 31 c0 48 8d 45 10 c7 45 b0 10 00 00 00 48 89 45 b8 48 8d 45 d0 48 89 45 c0 b8 10 00 00 00 0f 05 <89> c2 3d 00 f0 ff ff 77 1a 48 8b 45 c8 64 48 2b 04 25 28 00 00 00
 RSP: 002b:00007fffd08c3960 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
 RAX: ffffffffffffffda RBX: 00007fffd08c39f0 RCX: 00007262a6124ded
 RDX: 00007fffd08c3a10 RSI: 00000000c0181b01 RDI: 0000000000000007
 RBP: 00007fffd08c39b0 R08: 0000000014107820 R09: 00007fffd08c3b44
 R10: 000000000000000c R11: 0000000000000246 R12: 00007fffd08c3b44
 R13: 000000000000000c R14: 00007fffd08c3b58 R15: 0000000014107960
  </TASK>

Fixes: 1efe8c0670d6 ("RDMA/core: Convert UMEM ODP DMA mapping to caching IOVA and page linkage")
Closes: https://lore.kernel.org/all/3e8f343f-7d66-4f7a-9f08-3910623e322f@gmail.com/
Signed-off-by: Daisuke Matsuda <dskmtsd@gmail.com>
Link: https://patch.msgid.link/20250524144328.4361-1-dskmtsd@gmail.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/rxe: Break endless pagefault loop for RO pages

RO pages has "perm" equal to 0, that caused to the situation
where such pages were marked as needed to have fault and caused
to infinite loop.

Fixes: eedd5b1276e7 ("RDMA/umem: Store ODP access mask information in PFN")
Reported-by: Daisuke Matsuda <dskmtsd@gmail.com>
Closes: https://lore.kernel.org/all/3016329a-4edd-4550-862f-b298a1b79a39@gmail.com/
Link: https://patch.msgid.link/096fab178d48ed86942ee22eafe9be98e29092aa.1747913377.git.leonro@nvidia.com
Tested-by: Daisuke Matsuda <dskmtsd@gmail.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>

RDMA/bnxt_re: Fix return code of bnxt_re_configure_cc

Driver currently supports modifying GEN0_EXT0 CC parameters
through debugfs hook.

Fixed to return -EOPNOTSUPP instead of -EINVAL in bnxt_re_configure_cc()
when the user tries to modify any other CC parameters.

Fixes: 656dff55da19 ("RDMA/bnxt_re: Congestion control settings using debugfs hook")
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Link: https://patch.msgid.link/20250520035910.1061918-4-kalesh-anakkur.purayil@broadcom.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/bnxt_re: Fix missing error handling for tx_queue

bnxt_re_fill_gen0_ext0() did not return an error when
attempting to modify CMDQ_MODIFY_ROCE_CC_MODIFY_MASK_TX_QUEUE,
leading to silent failures.

Fixed this by returning -EOPNOTSUPP for tx_queue modifications and
ensuring proper error propagation in bnxt_re_configure_cc().

Fixes: 656dff55da19 ("RDMA/bnxt_re: Congestion control settings using debugfs hook")
Signed-off-by: Gautam R A <gautam-r.a@broadcom.com>
Link: https://patch.msgid.link/20250520035910.1061918-3-kalesh-anakkur.purayil@broadcom.com
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/bnxt_re: Fix incorrect display of inactivity_cp in debugfs output

The inactivity_cp parameter in debugfs was not being read or
written correctly, resulting in "Invalid argument" errors.

Fixed this by ensuring proper mapping of inactivity_cp in
both the map_cc_config_offset_gen0_ext0 and
bnxt_re_fill_gen0_ext0() functions.

Fixes: 656dff55da19 ("RDMA/bnxt_re: Congestion control settings using debugfs hook")
Signed-off-by: Gautam R A <gautam-r.a@broadcom.com>
Link: https://patch.msgid.link/20250520035910.1061918-2-kalesh-anakkur.purayil@broadcom.com
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/mlx5: Add support for 200Gbps per lane speeds

Add support for 200Gbps per lane speeds speed when querying PTYS and
report it back correctly when needed.

Signed-off-by: Patrisious Haddad <phaddad@nvidia.com>
Reviewed-by: Maor Gottlieb <maorg@nvidia.com>
Link: https://patch.msgid.link/b842d2f523e9b82e221378c444ebd5860d612959.1747134197.git.leon@kernel.org
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/mlx5: Remove the redundant MLX5_IB_STAGE_UAR stage

The MLX5_IB_STAGE_UAR stage in the RDMA driver is redundant and should
be removed.

Responsibility for initializing the device's UAR pointer
(mdev->priv.uar) lies with mlx5_core, which already sets it during the
mlx5_load() process.

At present, the RDMA UAR stage overwrites this pointer, which was
correctly initialized by mlx5_core, creating the risk of inconsistency.

Ownership and management of the UAR pointer should remain exclusively
within mlx5_core.

In the current upstream code, we luckily receive the same pointer, since
mlx5_get_uars_page() still finds available BF registers for that UAR,
allowing it to be shared.

However, future changes in mlx5_core may expose this flaw.
For instance, if mlx5_alloc_bfreg() is invoked twice before the RDMA UAR
stage runs, the RDMA driver may overwrite the UAR allocated by
mlx5_core.

This could lead to real bugs. For example, if mlx5_ib is unloaded
(rmmod), it might free the UAR, leaving mlx5_core with a dangling
reference to an invalid UAR.

Signed-off-by: Yishai Hadas <yishaih@nvidia.com>
Reviewed-by: Fan Li <fanl@nvidia.com>
Link: https://patch.msgid.link/feaa84ec6f20468b4935c439923e9266122a93d0.1747134130.git.leon@kernel.org
Signed-off-by: Leon Romanovsky <leon@kernel.org>

Merge branch 'for-next' of https://git./linux/kernel/git/tnguy/linux into wip/leon-for-next

Tony Nguyen says:

====================
Prepare for Intel IPU E2000 (GEN3)

This is the first part in introducing RDMA support for idpf.

----------------------------------------------------------------
Tatyana Nikolova says:

To align with review comments, the patch series introducing RDMA
RoCEv2 support for the Intel Infrastructure Processing Unit (IPU)
E2000 line of products is going to be submitted in three parts:

1. Modify ice to use specific and common IIDC definitions and
   pass a core device info to irdma.

2. Add RDMA support to idpf and modify idpf to use specific and
   common IIDC definitions and pass a core device info to irdma.

3. Add RDMA RoCEv2 support for the E2000 products, referred to as
   GEN3 to irdma.

This first part is a 5 patch series based on the original
"iidc/ice/irdma: Update IDC to support multiple consumers" patch
to allow for multiple CORE PCI drivers, using the auxbus.

Patches:
1) Move header file to new name for clarity and replace ice
   specific DSCP define with a kernel equivalent one in irdma
2) Unify naming convention
3) Separate header file into common and driver specific info
4) Replace ice specific DSCP define with a kernel equivalent
   one in ice
5) Implement core device info struct and update drivers to use it
----------------------------------------------------------------

v1: https://lore.kernel.org/20250505212037.2092288-1-anthony.l.nguyen@intel.com

IWL reviews:
[v5] https://lore.kernel.org/20250416021549.606-1-tatyana.e.nikolova@intel.com
[v4] https://lore.kernel.org/20250225050428.2166-1-tatyana.e.nikolova@intel.com
[v3] https://lore.kernel.org/20250207194931.1569-1-tatyana.e.nikolova@intel.com
[v2] https://lore.kernel.org/20240824031924.421-1-tatyana.e.nikolova@intel.com
[v1] https://lore.kernel.org/20240724233917.704-1-tatyana.e.nikolova@intel.com

* 'for-next' of https://git.kernel.org/pub/scm/linux/kernel/git/tnguy/linux:
  iidc/ice/irdma: Update IDC to support multiple consumers
  ice: Replace ice specific DSCP mapping num with a kernel define
  iidc/ice/irdma: Break iidc.h into two headers
  iidc/ice/irdma: Rename to iidc_* convention
  iidc/ice/irdma: Rename IDC header file

====================
Link: https://patch.msgid.link/20250509200712.2911060-1-anthony.l.nguyen@intel.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/iwcm: Fix use-after-free of work objects after cm_id destruction

The commit 59c68ac31e15 ("iw_cm: free cm_id resources on the last
deref") simplified cm_id resource management by freeing cm_id once all
references to the cm_id were removed. The references are removed either
upon completion of iw_cm event handlers or when the application destroys
the cm_id. This commit introduced the use-after-free condition where
cm_id_private object could still be in use by event handler works during
the destruction of cm_id. The commit aee2424246f9 ("RDMA/iwcm: Fix a
use-after-free related to destroying CM IDs") addressed this use-after-
free by flushing all pending works at the cm_id destruction.

However, still another use-after-free possibility remained. It happens
with the work objects allocated for each cm_id_priv within
alloc_work_entries() during cm_id creation, and subsequently freed in
dealloc_work_entries() once all references to the cm_id are removed.
If the cm_id's last reference is decremented in the event handler work,
the work object for the work itself gets removed, and causes the use-
after-free BUG below:

  BUG: KASAN: slab-use-after-free in __pwq_activate_work+0x1ff/0x250
  Read of size 8 at addr ffff88811f9cf800 by task kworker/u16:1/147091

  CPU: 2 UID: 0 PID: 147091 Comm: kworker/u16:1 Not tainted 6.15.0-rc2+ #27 PREEMPT(voluntary)
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.3-3.fc41 04/01/2014
  Workqueue:  0x0 (iw_cm_wq)
  Call Trace:
   <TASK>
   dump_stack_lvl+0x6a/0x90
   print_report+0x174/0x554
   ? __virt_addr_valid+0x208/0x430
   ? __pwq_activate_work+0x1ff/0x250
   kasan_report+0xae/0x170
   ? __pwq_activate_work+0x1ff/0x250
   __pwq_activate_work+0x1ff/0x250
   pwq_dec_nr_in_flight+0x8c5/0xfb0
   process_one_work+0xc11/0x1460
   ? __pfx_process_one_work+0x10/0x10
   ? assign_work+0x16c/0x240
   worker_thread+0x5ef/0xfd0
   ? __pfx_worker_thread+0x10/0x10
   kthread+0x3b0/0x770
   ? __pfx_kthread+0x10/0x10
   ? rcu_is_watching+0x11/0xb0
   ? _raw_spin_unlock_irq+0x24/0x50
   ? rcu_is_watching+0x11/0xb0
   ? __pfx_kthread+0x10/0x10
   ret_from_fork+0x30/0x70
   ? __pfx_kthread+0x10/0x10
   ret_from_fork_asm+0x1a/0x30
   </TASK>

  Allocated by task 147416:
   kasan_save_stack+0x2c/0x50
   kasan_save_track+0x10/0x30
   __kasan_kmalloc+0xa6/0xb0
   alloc_work_entries+0xa9/0x260 [iw_cm]
   iw_cm_connect+0x23/0x4a0 [iw_cm]
   rdma_connect_locked+0xbfd/0x1920 [rdma_cm]
   nvme_rdma_cm_handler+0x8e5/0x1b60 [nvme_rdma]
   cma_cm_event_handler+0xae/0x320 [rdma_cm]
   cma_work_handler+0x106/0x1b0 [rdma_cm]
   process_one_work+0x84f/0x1460
   worker_thread+0x5ef/0xfd0
   kthread+0x3b0/0x770
   ret_from_fork+0x30/0x70
   ret_from_fork_asm+0x1a/0x30

  Freed by task 147091:
   kasan_save_stack+0x2c/0x50
   kasan_save_track+0x10/0x30
   kasan_save_free_info+0x37/0x60
   __kasan_slab_free+0x4b/0x70
   kfree+0x13a/0x4b0
   dealloc_work_entries+0x125/0x1f0 [iw_cm]
   iwcm_deref_id+0x6f/0xa0 [iw_cm]
   cm_work_handler+0x136/0x1ba0 [iw_cm]
   process_one_work+0x84f/0x1460
   worker_thread+0x5ef/0xfd0
   kthread+0x3b0/0x770
   ret_from_fork+0x30/0x70
   ret_from_fork_asm+0x1a/0x30

  Last potentially related work creation:
   kasan_save_stack+0x2c/0x50
   kasan_record_aux_stack+0xa3/0xb0
   __queue_work+0x2ff/0x1390
   queue_work_on+0x67/0xc0
   cm_event_handler+0x46a/0x820 [iw_cm]
   siw_cm_upcall+0x330/0x650 [siw]
   siw_cm_work_handler+0x6b9/0x2b20 [siw]
   process_one_work+0x84f/0x1460
   worker_thread+0x5ef/0xfd0
   kthread+0x3b0/0x770
   ret_from_fork+0x30/0x70
   ret_from_fork_asm+0x1a/0x30

This BUG is reproducible by repeating the blktests test case nvme/061
for the rdma transport and the siw driver.

To avoid the use-after-free of cm_id_private work objects, ensure that
the last reference to the cm_id is decremented not in the event handler
works, but in the cm_id destruction context. For that purpose, move
iwcm_deref_id() call from destroy_cm_id() to the callers of
destroy_cm_id(). In iw_destroy_cm_id(), call iwcm_deref_id() after
flushing the pending works.

During the fix work, I noticed that iw_destroy_cm_id() is called from
cm_work_handler() and process_event() context. However, the comment of
iw_destroy_cm_id() notes that the function "cannot be called by the
event thread". Drop the false comment.

Closes: https://lore.kernel.org/linux-rdma/r5676e754sv35aq7cdsqrlnvyhiq5zktteaurl7vmfih35efko@z6lay7uypy3c/
Fixes: 59c68ac31e15 ("iw_cm: free cm_id resources on the last deref")
Cc: stable@vger.kernel.org
Signed-off-by: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com>
Link: https://patch.msgid.link/20250510101036.1756439-1-shinichiro.kawasaki@wdc.com
Reviewed-by: Zhu Yanjun <yanjun.zhu@linux.dev>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

net: mana: Add support for auxiliary device servicing events

Handle soc servicing events which require the rdma auxiliary device resources to
be cleaned up during a suspend, and re-initialized during a resume.

Signed-off-by: Shiraz Saleem <shirazsaleem@microsoft.com>
Signed-off-by: Konstantin Taranov <kotaranov@microsoft.com>
Link: https://patch.msgid.link/1746633545-17653-5-git-send-email-kotaranov@linux.microsoft.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/mana_ib: unify mana_ib functions to support any gdma device

Use the installed gdma_device instead of hard-coded device
in requests to the HW.

Signed-off-by: Konstantin Taranov <kotaranov@microsoft.com>
Link: https://patch.msgid.link/1746633545-17653-4-git-send-email-kotaranov@linux.microsoft.com
Reviewed-by: Long Li <longli@microsoft.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/mana_ib: Add support of mana_ib for RNIC and ETH nic

Allow mana_ib to be created over ethernet gdma device and
over rnic gdma device. The HW has two devices with different
capabilities and different use-cases. Initialize required
resources depending on the used gdma device.

Signed-off-by: Konstantin Taranov <kotaranov@microsoft.com>
Link: https://patch.msgid.link/1746633545-17653-3-git-send-email-kotaranov@linux.microsoft.com
Reviewed-by: Long Li <longli@microsoft.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

net: mana: Probe rdma device in mana driver

Initialize gdma device for rdma inside mana module.
For each gdma device, initialize an auxiliary ib device.

Signed-off-by: Konstantin Taranov <kotaranov@microsoft.com>
Link: https://patch.msgid.link/1746633545-17653-2-git-send-email-kotaranov@linux.microsoft.com
Reviewed-by: Long Li <longli@microsoft.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/siw: replace redundant ternary operator with just rv

The use of the ternary operator on rv is redundant, rv is
either the initialized value of 0 or a negative error return
code, so it can never be greater than zero, and hence the
zero assignment in ternary operator is redundant. Just return
rv instead.

Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Link: https://patch.msgid.link/20250507131834.253823-1-colin.i.king@gmail.com
Acked-by: Bernard Metzler <bmt@zurich.ibm.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/umem: Separate implicit ODP initialization from explicit ODP

Create separate functions for the implicit ODP initialization
which is different from the explicit ODP initialization.

Tested-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>

RDMA/core: Convert UMEM ODP DMA mapping to caching IOVA and page linkage

Reuse newly added DMA API to cache IOVA and only link/unlink pages
in fast path for UMEM ODP flow.

Tested-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>

RDMA/umem: Store ODP access mask information in PFN

As a preparation to remove dma_list, store access mask in PFN pointer
and not in dma_addr_t.

Tested-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>

mm/hmm: provide generic DMA managing logic

HMM callers use PFN list to populate range while calling
to hmm_range_fault(), the conversion from PFN to DMA address
is done by the callers with help of another DMA list. However,
it is wasteful on any modern platform and by doing the right
logic, that DMA list can be avoided.

Provide generic logic to manage these lists and gave an interface
to map/unmap PFNs to DMA addresses, without requiring from the callers
to be an experts in DMA core API.

Tested-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>

mm/hmm: let users to tag specific PFN with DMA mapped bit

Introduce new sticky flag (HMM_PFN_DMA_MAPPED), which isn't overwritten
by HMM range fault. Such flag allows users to tag specific PFNs with
information if this specific PFN was already DMA mapped.

Tested-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>

Provide a new two step DMA mapping API

Currently the only efficient way to map a complex memory description through
the DMA API is by using the scatter list APIs. The SG APIs are unique in that
they efficiently combine the two fundamental operations of sizing and allocating
a large IOVA window from the IOMMU and processing all the per-address
swiotlb/flushing/p2p/map details.

This uniqueness has been a long standing pain point as the scatter list API
is mandatory, but expensive to use. It prevents any kind of optimization or
feature improvement (such as avoiding struct page for P2P) due to the
impossibility of improving the scatter list.

Several approaches have been explored to expand the DMA API with additional
scatterlist-like structures (BIO, rlist), instead split up the DMA API
to allow callers to bring their own data structure.

The API is split up into parts:
 - Allocate IOVA space:
    To do any pre-allocation required. This is done based on the caller
    supplying some details about how much IOMMU address space it would need
    in worst case.
 - Map and unmap relevant structures to pre-allocated IOVA space:
    Perform the actual mapping into the pre-allocated IOVA. This is very
    similar to dma_map_page().

Thanks

Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/hns: Fix build error of hns_roce_trace

Add include path to find hns_roce_trace.h to fix the following
build error:

In file included from drivers/infiniband/hw/hns/hns_roce_trace.h:213,
                 from drivers/infiniband/hw/hns/hns_roce_hw_v2.c:53:
./include/trace/define_trace.h:110:42: fatal error: ./hns_roce_trace.h: No such file or directory
  110 | #include TRACE_INCLUDE(TRACE_INCLUDE_FILE)
      |                                          ^
compilation terminated.

Fixes: 02007e3ddc07 ("RDMA/hns: Add trace for flush CQE")
Reported-by: Paul E. McKenney <paulmck@kernel.org>
Closes: https://lore.kernel.org/linux-next/b7dd4dda-37d8-47e4-8d78-b6585be21cfd@paulmck-laptop/T/#t
Tested-by: Paul E. McKenney <paulmck@kernel.org>
Signed-off-by: Junxian Huang <huangjunxian6@hisilicon.com>
Link: https://patch.msgid.link/20250507033903.2879433-1-huangjunxian6@hisilicon.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

iidc/ice/irdma: Update IDC to support multiple consumers

In preparation of supporting more than a single core PCI driver
for RDMA, move ice specific structs like qset_params, qos_info
and qos_params from iidc_rdma.h to iidc_rdma_ice.h.

Previously, the ice driver was just exporting its entire PF struct
to the auxiliary driver, but since each core driver will have its own
different PF struct, implement a universal struct that all core drivers
can provide to the auxiliary driver through the probe call.

Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Signed-off-by: Dave Ertman <david.m.ertman@intel.com>
Co-developed-by: Mustafa Ismail <mustafa.ismail@intel.com>
Signed-off-by: Mustafa Ismail <mustafa.ismail@intel.com>
Co-developed-by: Shiraz Saleem <shiraz.saleem@intel.com>
Signed-off-by: Shiraz Saleem <shiraz.saleem@intel.com>
Co-developed-by: Tatyana Nikolova <tatyana.e.nikolova@intel.com>
Signed-off-by: Tatyana Nikolova <tatyana.e.nikolova@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

RDMA/siw: Remove unused siw_mem_add

siw_mem_add() was added in 2019 by commit 2251334dcac9 ("rdma/siw:
application buffer management") but has remained unused.

Remove it.

Link: https://patch.msgid.link/r/20250505210226.88994-1-linux@treblig.org
Signed-off-by: Dr. David Alan Gilbert <linux@treblig.org>
Acked-by: Bernard Metzler <bmt@zurich.ibm.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

IB/hfi1: Remove unused sc_drop and sdma_all_idle

sc_drop() and sdma_all_idle() were both added in 2015's commit
7724105686e7 ("IB/hfi1: add driver files") but have remained unused.

Remove them.

Link: https://patch.msgid.link/r/20250505205419.88131-1-linux@treblig.org
Signed-off-by: Dr. David Alan Gilbert <linux@treblig.org>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

docs: core-api: document the IOVA-based API

Add an explanation of the newly added IOVA-based mapping API.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Tested-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>

dma-mapping: add a dma_need_unmap helper

Add helper that allows a driver to skip calling dma_unmap_*
if the DMA layer can guarantee that they are no-nops.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Tested-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>

dma-mapping: Implement link/unlink ranges API

Introduce new DMA APIs to perform DMA linkage of buffers
in layers higher than DMA.

In proposed API, the callers will perform the following steps.
In map path:
if (dma_can_use_iova(...))
    dma_iova_alloc()
    for (page in range)
       dma_iova_link_next(...)
    dma_iova_sync(...)
else
     /* Fallback to legacy map pages */
             for (all pages)
       dma_map_page(...)

In unmap path:
if (dma_can_use_iova(...))
     dma_iova_destroy()
else
     for (all pages)
dma_unmap_page(...)

Reviewed-by: Christoph Hellwig <hch@lst.de>
Tested-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>

iommu/dma: Factor out a iommu_dma_map_swiotlb helper

Split the iommu logic from iommu_dma_map_page into a separate helper.
This not only keeps the code neatly separated, but will also allow for
reuse in another caller.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Tested-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>

dma-mapping: Provide an interface to allow allocate IOVA

The existing .map_pages() callback provides both allocating of IOVA
and linking DMA pages. That combination works great for most of the
callers who use it in control paths, but is less effective in fast
paths where there may be multiple calls to map_page().

These advanced callers already manage their data in some sort of
database and can perform IOVA allocation in advance, leaving range
linkage operation to be in fast path.

Provide an interface to allocate/deallocate IOVA and next patch
link/unlink DMA ranges to that specific IOVA.

In the new API a DMA mapping transaction is identified by a
struct dma_iova_state, which holds some recomputed information
for the transaction which does not change for each page being
mapped, so add a check if IOVA can be used for the specific
transaction.

The API is exported from dma-iommu as it is the only implementation
supported, the namespace is clearly different from iommu_* functions
which are not allowed to be used. This code layout allows us to save
function call per API call used in datapath as well as a lot of boilerplate
code.

Reviewed-by: Christoph Hellwig <hch@lst.de>
Tested-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>

iommu: add kernel-doc for iommu_unmap_fast

Add kernel-doc section for iommu_unmap_fast to document existing
limitation of underlying functions which can't split individual ranges.

Suggested-by: Jason Gunthorpe <jgg@nvidia.com>
Acked-by: Will Deacon <will@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Tested-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>

iommu: generalize the batched sync after map interface

For the upcoming IOVA-based DMA API we want to batch the
ops->iotlb_sync_map() call after mapping multiple IOVAs from
dma-iommu without having a scatterlist. Improve the API.

Add a wrapper for the map_sync as iommu_sync_map() so that callers
don't need to poke into the methods directly.

Formalize __iommu_map() into iommu_map_nosync() which requires the
caller to call iommu_sync_map() after all maps are completed.

Refactor the existing sanity checks from all the different layers
into iommu_map_nosync().

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Will Deacon <will@kernel.org>
Tested-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>

dma-mapping: move the PCI P2PDMA mapping helpers to pci-p2pdma.h

To support the upcoming non-scatterlist mapping helpers, we need to go
back to have them called outside of the DMA API.  Thus move them out of
dma-map-ops.h, which is only for DMA API implementations to pci-p2pdma.h,
which is for driver use.

Note that the core helper is still not exported as the mapping is
expected to be done only by very highlevel subsystem code at least for
now.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Logan Gunthorpe <logang@deltatee.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Tested-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>

PCI/P2PDMA: Refactor the p2pdma mapping helpers

The current scheme with a single helper to determine the P2P status
and map a scatterlist segment force users to always use the map_sg
helper to DMA map, which we're trying to get away from because they
are very cache inefficient.

Refactor the code so that there is a single helper that checks the P2P
state for a page, including the result that it is not a P2P page to
simplify the callers, and a second one to perform the address translation
for a bus mapped P2P transfer that does not depend on the scatterlist
structure.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Logan Gunthorpe <logang@deltatee.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Tested-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>

RDMA/mlx5: Fix error flow upon firmware failure for RQ destruction

Upon RQ destruction if the firmware command fails which is the
last resource to be destroyed some SW resources were already cleaned
regardless of the failure.

Now properly rollback the object to its original state upon such failure.

In order to avoid a use-after free in case someone tries to destroy the
object again, which results in the following kernel trace:
refcount_t: underflow; use-after-free.
WARNING: CPU: 0 PID: 37589 at lib/refcount.c:28 refcount_warn_saturate+0xf4/0x148
Modules linked in: rdma_ucm(OE) rdma_cm(OE) iw_cm(OE) ib_ipoib(OE) ib_cm(OE) ib_umad(OE) mlx5_ib(OE) rfkill mlx5_core(OE) mlxdevm(OE) ib_uverbs(OE) ib_core(OE) psample mlxfw(OE) mlx_compat(OE) macsec tls pci_hyperv_intf sunrpc vfat fat virtio_net net_failover failover fuse loop nfnetlink vsock_loopback vmw_vsock_virtio_transport_common vmw_vsock_vmci_transport vmw_vmci vsock xfs crct10dif_ce ghash_ce sha2_ce sha256_arm64 sha1_ce virtio_console virtio_gpu virtio_blk virtio_dma_buf virtio_mmio dm_mirror dm_region_hash dm_log dm_mod xpmem(OE)
CPU: 0 UID: 0 PID: 37589 Comm: python3 Kdump: loaded Tainted: G           OE     -------  ---  6.12.0-54.el10.aarch64 #1
Tainted: [O]=OOT_MODULE, [E]=UNSIGNED_MODULE
Hardware name: QEMU KVM Virtual Machine, BIOS 0.0.0 02/06/2015
pstate: 60400005 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
pc : refcount_warn_saturate+0xf4/0x148
lr : refcount_warn_saturate+0xf4/0x148
sp : ffff80008b81b7e0
x29: ffff80008b81b7e0 x28: ffff000133d51600 x27: 0000000000000001
x26: 0000000000000000 x25: 00000000ffffffea x24: ffff00010ae80f00
x23: ffff00010ae80f80 x22: ffff0000c66e5d08 x21: 0000000000000000
x20: ffff0000c66e0000 x19: ffff00010ae80340 x18: 0000000000000006
x17: 0000000000000000 x16: 0000000000000020 x15: ffff80008b81b37f
x14: 0000000000000000 x13: 2e656572662d7265 x12: ffff80008283ef78
x11: ffff80008257efd0 x10: ffff80008283efd0 x9 : ffff80008021ed90
x8 : 0000000000000001 x7 : 00000000000bffe8 x6 : c0000000ffff7fff
x5 : ffff0001fb8e3408 x4 : 0000000000000000 x3 : ffff800179993000
x2 : 0000000000000000 x1 : 0000000000000000 x0 : ffff000133d51600
Call trace:
 refcount_warn_saturate+0xf4/0x148
 mlx5_core_put_rsc+0x88/0xa0 [mlx5_ib]
 mlx5_core_destroy_rq_tracked+0x64/0x98 [mlx5_ib]
 mlx5_ib_destroy_wq+0x34/0x80 [mlx5_ib]
 ib_destroy_wq_user+0x30/0xc0 [ib_core]
 uverbs_free_wq+0x28/0x58 [ib_uverbs]
 destroy_hw_idr_uobject+0x34/0x78 [ib_uverbs]
 uverbs_destroy_uobject+0x48/0x240 [ib_uverbs]
 __uverbs_cleanup_ufile+0xd4/0x1a8 [ib_uverbs]
 uverbs_destroy_ufile_hw+0x48/0x120 [ib_uverbs]
 ib_uverbs_close+0x2c/0x100 [ib_uverbs]
 __fput+0xd8/0x2f0
 __fput_sync+0x50/0x70
 __arm64_sys_close+0x40/0x90
 invoke_syscall.constprop.0+0x74/0xd0
 do_el0_svc+0x48/0xe8
 el0_svc+0x44/0x1d0
 el0t_64_sync_handler+0x120/0x130
 el0t_64_sync+0x1a4/0x1a8

Fixes: e2013b212f9f ("net/mlx5_core: Add RQ and SQ event handling")
Signed-off-by: Patrisious Haddad <phaddad@nvidia.com>
Link: https://patch.msgid.link/3181433ccdd695c63560eeeb3f0c990961732101.1745839855.git.leon@kernel.org
Signed-off-by: Leon Romanovsky <leon@kernel.org>

IB/cm: Drop lockdep assert and WARN when freeing old msg

The send completion handler can run after cm_id has advanced to another
message.  The cm_id lock is not needed in this case, but a recent change
re-used cm_free_priv_msg(), which asserts that the lock is held and
WARNs if the cm_id's currently outstanding msg is different than the one
being freed.

Fixes: 1e5159219076 ("IB/cm: Do not hold reference on cm_id unless needed")
Signed-off-by: Vlad Dumitrescu <vdumitrescu@nvidia.com>
Reviewed-by: Sean Hefty <shefty@nvidia.com>
Link: https://patch.msgid.link/0c364c29142f72b7875fdeba51f3c9bd6ca863ee.1745839788.git.leon@kernel.org
Signed-off-by: Leon Romanovsky <leon@kernel.org>

ice: Replace ice specific DSCP mapping num with a kernel define

Replace ice driver specific DSCP mapping number defines
ICE_DSCP_NUM_VAL and IIDC_MAX_DSCP_MAPPING with
an equivalent kernel define DSCP_MAX.

Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Signed-off-by: Tatyana Nikolova <tatyana.e.nikolova@intel.com>
Signed-off-by: Dave Ertman <david.m.ertman@intel.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

iidc/ice/irdma: Break iidc.h into two headers

In preparation of supporting more than a single core PCI driver
for RDMA, break the iidc_rdma.h header file into two more focused
headers.

Only the elements universal to all Intel drivers will remain in
the generic iidc_rdma.h header. Move the ice specific information
to an ice specific header file named iidc_rdma_ice.h.

Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Signed-off-by: Dave Ertman <david.m.ertman@intel.com>
Signed-off-by: Tatyana Nikolova <tatyana.e.nikolova@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

iidc/ice/irdma: Rename to iidc_* convention

In preparation of supporting more than a single core PCI driver
for RDMA, homogenize naming to iidc_rdma_* and IIDC_RDMA_*
form.

Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Signed-off-by: Dave Ertman <david.m.ertman@intel.com>
Signed-off-by: Tatyana Nikolova <tatyana.e.nikolova@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

iidc/ice/irdma: Rename IDC header file

To prepare for the IDC upgrade to support different CORE
PCI drivers, rename header file from iidc.h to iidc_rdma.h
since this files functionality is specifically for RDMA support.

Use net/dscp.h include in irdma osdep.h and DSCP_MAX type.h,
instead of iidc header and define.

Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Signed-off-by: Dave Ertman <david.m.ertman@intel.com>
Signed-off-by: Tatyana Nikolova <tatyana.e.nikolova@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

IB/hfi1: Adjust fd->entry_to_rb allocation type

In preparation for making the kmalloc family of allocators type aware,
we need to make sure that the returned type from the allocation matches
the type of the variable being assigned. (Before, the allocator would
always return "void *", which can be implicitly cast to any pointer type.)

The assigned type is "struct tid_rb_node **", but the return type will be
"struct rb_node **". These are the same allocation size (pointer size),
but the types do not match. Adjust the allocation type to match the
assignment.

Signed-off-by: Kees Cook <kees@kernel.org>
Link: https://patch.msgid.link/20250426061247.work.261-kees@kernel.org
Signed-off-by: Leon Romanovsky <leon@kernel.org>

IB/mthca: Adjust buddy->bits allocation type

In preparation for making the kmalloc family of allocators type aware,
we need to make sure that the returned type from the allocation matches
the type of the variable being assigned. (Before, the allocator would
always return "void *", which can be implicitly cast to any pointer type.)

The assigned type is "unsigned long **", but the returned type will be
"long **". These are the same allocation size (pointer size), but the
types do not match. Adjust the allocation type to match the assignment.

Signed-off-by: Kees Cook <kees@kernel.org>
Link: https://patch.msgid.link/20250426061208.work.000-kees@kernel.org
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/hns: Add trace for CMDQ dumping

Add trace for CMDQ dumping.

Output example:
$ cat /sys/kernel/debug/tracing/trace
  tracer: nop

  entries-in-buffer/entries-written: 2/2   #P:128

                                 _-----=> irqs-off/BH-disabled
                               / _----=> need-resched
                               | / _---=> hardirq/softirq
                               || / _--=> preempt-depth
                               ||| / _-=> migrate-disable
                               |||| /     delay
            TASK-PID     CPU#  |||||  TIMESTAMP  FUNCTION
               | |         |   |||||     |         |
  kworker/u512:1-14003   [089] b..1. 50737.238304: hns_cmdq_req: 0000:bd:00.0 cmdq opcode:0x8500, flag:0x1, retval:0x0, data:{0x2,0x0,0x0,0xffff0000,0x32323232,0x0}

  kworker/u512:1-14003   [089] b..1. 50737.238316: hns_cmdq_resp: 0000:bd:00.0 cmdq opcode:0x8500, flag:0x2, retval:0x0, data:{0x2,0x0,0x0,0xffff0000,0x32323232,0x0}

Signed-off-by: Junxian Huang <huangjunxian6@hisilicon.com>
Link: https://patch.msgid.link/20250421132750.1363348-7-huangjunxian6@hisilicon.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/hns: Include hnae3.h in hns_roce_hw_v2.h

hns_roce_hw_v2.h has a direct dependency on hnae3.h due to the
inline function hns_roce_write64(), but it doesn't include this
header currently. This leads to that files including
hns_roce_hw_v2.h must also include hnae3.h to avoid compilation
errors, even if they themselves don't really rely on hnae3.h.
This doesn't make sense, hns_roce_hw_v2.h should include hnae3.h
directly.

Fixes: d3743fa94ccd ("RDMA/hns: Fix the chip hanging caused by sending doorbell during reset")
Signed-off-by: Junxian Huang <huangjunxian6@hisilicon.com>
Link: https://patch.msgid.link/20250421132750.1363348-6-huangjunxian6@hisilicon.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/hns: Add trace for MR/MTR attribute dumping

Add trace for MR/MTR attribute dumping.

Output example:
$ cat /sys/kernel/debug/tracing/trace
  tracer: nop

  entries-in-buffer/entries-written: 2/2   #P:128

                                 _-----=> irqs-off/BH-disabled
                               / _----=> need-resched
                               | / _---=> hardirq/softirq
                               || / _--=> preempt-depth
                               ||| / _-=> migrate-disable
                               |||| /     delay
            TASK-PID     CPU#  |||||  TIMESTAMP  FUNCTION
               | |         |   |||||     |         |
      ib_send_bw-14751   [111] .....  8763.823038: hns_buf_attr: rg cnt:1,
pg_sft:0xc, mtt_only:no, rg 0 (sz:131072, hop:2), rg 1 (sz:0, hop:0),
rg 2 (sz:0, hop:0)

      ib_send_bw-14751   [111] .....  8763.823118: hns_mr:
iova:0xffffb2968000, size:131072, key:512, pd:1, pbl_hop:1, npages:4,
type:0, status:0

Signed-off-by: Junxian Huang <huangjunxian6@hisilicon.com>
Link: https://patch.msgid.link/20250421132750.1363348-5-huangjunxian6@hisilicon.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/hns: Add trace for AEQE dumping

Add trace for AEQE dumping.

Output example:
$ cat /sys/kernel/debug/tracing/trace
  tracer: nop

  entries-in-buffer/entries-written: 2/2   #P:128

                                 _-----=> irqs-off/BH-disabled
                               / _----=> need-resched
                               | / _---=> hardirq/softirq
                               || / _--=> preempt-depth
                               ||| / _-=> migrate-disable
                               |||| /     delay
            TASK-PID     CPU#  |||||  TIMESTAMP  FUNCTION
               | |         |   |||||     |         |
  <idle>-0       [120] d.h1.  7995.835587: hns_ae_info: event 19 aeqe:
{0x80006013,0x0,0x0,0x10d2c,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x0}

Signed-off-by: Junxian Huang <huangjunxian6@hisilicon.com>
Link: https://patch.msgid.link/20250421132750.1363348-4-huangjunxian6@hisilicon.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/hns: Add trace for WQE dumping

Add trace for WQE dumping, including SQ, RQ and SRQ.

Output example:
$ cat /sys/kernel/debug/tracing/trace
  tracer: nop

  entries-in-buffer/entries-written: 2/2   #P:128

                                 _-----=> irqs-off/BH-disabled
                               / _----=> need-resched
                               | / _---=> hardirq/softirq
       || / _--=> preempt-depth
                               ||| / _-=> migrate-disable
                               |||| /     delay
            TASK-PID     CPU#  |||||  TIMESTAMP  FUNCTION
               | |         |   |||||     |         |
  roce_test_main-22730   [074] d..1. 16133.898282: hns_sq_wqe: SQ 0xc wqe
(0x0/0xffff0820a6076060): {0x180,0x639c,0x0,0x1000000,0x0,0x0,0x0,0x0,
0x639c,0x300,0xf7e38000,0x0,0x0,0x0,0x0,0x0}

Signed-off-by: Junxian Huang <huangjunxian6@hisilicon.com>
Link: https://patch.msgid.link/20250421132750.1363348-3-huangjunxian6@hisilicon.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/hns: Add trace for flush CQE

Add trace to print the producer index of QP when triggering flush CQE.

Output example:
$ cat /sys/kernel/debug/tracing/trace
  tracer: nop

  entries-in-buffer/entries-written: 2/2   #P:128

                                 _-----=> irqs-off/BH-disabled
                                / _----=> need-resched
                               | / _---=> hardirq/softirq
                               || / _--=> preempt-depth
                               ||| / _-=> migrate-disable
                               |||| /     delay
            TASK-PID     CPU#  |||||  TIMESTAMP  FUNCTION
               | |         |   |||||     |         |
      ib_send_bw-11474   [075] d..1.  2393.434738: hns_sq_flush_cqe: SQ 0x2 flush head 0xb5c7.
      ib_send_bw-11474   [075] d..1.  2393.434739: hns_rq_flush_cqe: RQ 0x2 flush head 0.

Signed-off-by: Junxian Huang <huangjunxian6@hisilicon.com>
Link: https://patch.msgid.link/20250421132750.1363348-2-huangjunxian6@hisilicon.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/core: Move ODP capability definitions to uapi

The bits are used from both kernel space and userland, so they should be
placed in UAPI.

Signed-off-by: Daisuke Matsuda <matsuda-daisuke@fujitsu.com>
Link: https://patch.msgid.link/20250418051345.1022339-2-matsuda-daisuke@fujitsu.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/rxe: Remove 32-bit architecture support

Major linux distibutions have phased out support for 32-bit machines. Since
rxe is primarily used for development and testing, the benefit of
maintaining 32-bit support is minimal. This change simplifies ATOMIC WRITE
implementations and improves maintainability of the driver.

Signed-off-by: Daisuke Matsuda <matsuda-daisuke@fujitsu.com>
Link: https://patch.msgid.link/20250421025101.3588139-1-matsuda-daisuke@fujitsu.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/rxe: Remove unused rxe_run_task

rxe_run_task() has been unused since 2024's
commit 23bc06af547f ("RDMA/rxe: Don't call direct between tasks")

Remove it.

Signed-off-by: Dr. David Alan Gilbert <linux@treblig.org>
Link: https://patch.msgid.link/20250419132725.199785-1-linux@treblig.org
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/rxe: Fix "trying to register non-static key in rxe_qp_do_cleanup" bug

Call Trace:
 <TASK>
 __dump_stack lib/dump_stack.c:94 [inline]
 dump_stack_lvl+0x116/0x1f0 lib/dump_stack.c:120
 assign_lock_key kernel/locking/lockdep.c:986 [inline]
 register_lock_class+0x4a3/0x4c0 kernel/locking/lockdep.c:1300
 __lock_acquire+0x99/0x1ba0 kernel/locking/lockdep.c:5110
 lock_acquire kernel/locking/lockdep.c:5866 [inline]
 lock_acquire+0x179/0x350 kernel/locking/lockdep.c:5823
 __timer_delete_sync+0x152/0x1b0 kernel/time/timer.c:1644
 rxe_qp_do_cleanup+0x5c3/0x7e0 drivers/infiniband/sw/rxe/rxe_qp.c:815
 execute_in_process_context+0x3a/0x160 kernel/workqueue.c:4596
 __rxe_cleanup+0x267/0x3c0 drivers/infiniband/sw/rxe/rxe_pool.c:232
 rxe_create_qp+0x3f7/0x5f0 drivers/infiniband/sw/rxe/rxe_verbs.c:604
 create_qp+0x62d/0xa80 drivers/infiniband/core/verbs.c:1250
 ib_create_qp_kernel+0x9f/0x310 drivers/infiniband/core/verbs.c:1361
 ib_create_qp include/rdma/ib_verbs.h:3803 [inline]
 rdma_create_qp+0x10c/0x340 drivers/infiniband/core/cma.c:1144
 rds_ib_setup_qp+0xc86/0x19a0 net/rds/ib_cm.c:600
 rds_ib_cm_initiate_connect+0x1e8/0x3d0 net/rds/ib_cm.c:944
 rds_rdma_cm_event_handler_cmn+0x61f/0x8c0 net/rds/rdma_transport.c:109
 cma_cm_event_handler+0x94/0x300 drivers/infiniband/core/cma.c:2184
 cma_work_handler+0x15b/0x230 drivers/infiniband/core/cma.c:3042
 process_one_work+0x9cc/0x1b70 kernel/workqueue.c:3238
 process_scheduled_works kernel/workqueue.c:3319 [inline]
 worker_thread+0x6c8/0xf10 kernel/workqueue.c:3400
 kthread+0x3c2/0x780 kernel/kthread.c:464
 ret_from_fork+0x45/0x80 arch/x86/kernel/process.c:153
 ret_from_fork_asm+0x1a/0x30 arch/x86/entry/entry_64.S:245
 </TASK>

The root cause is as below:

In the function rxe_create_qp, the function rxe_qp_from_init is called
to create qp, if this function rxe_qp_from_init fails, rxe_cleanup will
be called to handle all the allocated resources, including the timers:
retrans_timer and rnr_nak_timer.

The function rxe_qp_from_init calls the function rxe_qp_init_req to
initialize the timers: retrans_timer and rnr_nak_timer.

But these timers are initialized in the end of rxe_qp_init_req.
If some errors occur before the initialization of these timers, this
problem will occur.

The solution is to check whether these timers are initialized or not.
If these timers are not initialized, ignore these timers.

Fixes: 8700e3e7c485 ("Soft RoCE driver")
Reported-by: syzbot+4edb496c3cad6e953a31@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=4edb496c3cad6e953a31
Signed-off-by: Zhu Yanjun <yanjun.zhu@linux.dev>
Link: https://patch.msgid.link/20250419080741.1515231-1-yanjun.zhu@linux.dev
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/cma: Remove unused rdma_res_to_id

The last use of rdma_res_to_id() was removed in 2020 by
commi t211cd9459fda ("RDMA: Add dedicated CM_ID resource tracker function")

Remove it.

Signed-off-by: Dr. David Alan Gilbert <linux@treblig.org>
Link: https://patch.msgid.link/20250418165848.241305-1-linux@treblig.org
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/mana_ib: Add support of 4M, 1G, and 2G pages

Check PF capability flag whether the 4M, 1G, and 2G pages are
supported. Add these pages sizes to mana_ib, if supported.

Define possible page sizes in enum gdma_page_type and
remove unused enum atb_page_size.

Signed-off-by: Konstantin Taranov <kotaranov@microsoft.com>
Link: https://patch.msgid.link/1744621234-26114-4-git-send-email-kotaranov@linux.microsoft.com
Reviewed-by: Long Li <longli@microsoft.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/mana_ib: support of the zero based MRs

Add IB_ZERO_BASED to the valid flags and use
the corresponding MR creation request for the zero
based memory.

Signed-off-by: Konstantin Taranov <kotaranov@microsoft.com>
Link: https://patch.msgid.link/1744621234-26114-3-git-send-email-kotaranov@linux.microsoft.com
Reviewed-by: Long Li <longli@microsoft.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/mana_ib: Access remote atomic for MRs

Add IB_ACCESS_REMOTE_ATOMIC to the valid flags for MRs and use
the corresponding flag bit during MR creation in the HW.

Signed-off-by: Konstantin Taranov <kotaranov@microsoft.com>
Link: https://patch.msgid.link/1744621234-26114-2-git-send-email-kotaranov@linux.microsoft.com
Reviewed-by: Long Li <longli@microsoft.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/hns: initialize db in update_srq_db()

On x86_64 with gcc version 13.3.0, I compile
drivers/infiniband/hw/hns/hns_roce_hw_v2.c with:

  make defconfig
  ./scripts/kconfig/merge_config.sh .config <(
    echo CONFIG_COMPILE_TEST=y
    echo CONFIG_HNS3=m
    echo CONFIG_INFINIBAND=m
    echo CONFIG_INFINIBAND_HNS_HIP08=m
  )
  make KCFLAGS="-fno-inline-small-functions -fno-inline-functions-called-once" \
    drivers/infiniband/hw/hns/hns_roce_hw_v2.o

Then I get a compile error:

    CALL    scripts/checksyscalls.sh
    DESCEND objtool
    INSTALL libsubcmd_headers
    CC [M]  drivers/infiniband/hw/hns/hns_roce_hw_v2.o
  In file included from drivers/infiniband/hw/hns/hns_roce_hw_v2.c:47:
  drivers/infiniband/hw/hns/hns_roce_hw_v2.c: In function 'update_srq_db':
  drivers/infiniband/hw/hns/hns_roce_common.h:74:17: error: 'db' is used uninitialized [-Werror=uninitialized]
     74 |                 *((__le32 *)_ptr + (field_h) / 32) &=                          \
        |                 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  drivers/infiniband/hw/hns/hns_roce_common.h:90:17: note: in expansion of macro '_hr_reg_clear'
     90 |                 _hr_reg_clear(ptr, field_type, field_h, field_l);              \
        |                 ^~~~~~~~~~~~~
  drivers/infiniband/hw/hns/hns_roce_common.h:95:39: note: in expansion of macro '_hr_reg_write'
     95 | #define hr_reg_write(ptr, field, val) _hr_reg_write(ptr, field, val)
        |                                       ^~~~~~~~~~~~~
  drivers/infiniband/hw/hns/hns_roce_hw_v2.c:948:9: note: in expansion of macro 'hr_reg_write'
    948 |         hr_reg_write(&db, DB_TAG, srq->srqn);
        |         ^~~~~~~~~~~~
  drivers/infiniband/hw/hns/hns_roce_hw_v2.c:946:31: note: 'db' declared here
    946 |         struct hns_roce_v2_db db;
        |                               ^~
  cc1: all warnings being treated as errors

Signed-off-by: Chen Linxuan <chenlinxuan@uniontech.com>
Co-developed-by: Winston Wen <wentao@uniontech.com>
Signed-off-by: Winston Wen <wentao@uniontech.com>
Link: https://patch.msgid.link/FF922C77946229B6+20250411105459.90782-5-chenlinxuan@uniontech.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/rxe: Fix mismatched type declarations

Some functions return int values while they are defined as enum resp_states
variables. This patch resolves the mismatches in rxe.

Signed-off-by: Daisuke Matsuda <matsuda-daisuke@fujitsu.com>
Link: https://patch.msgid.link/20250409102701.1275265-1-matsuda-daisuke@fujitsu.com
Reviewed-by: Zhu Yanjun <yanjun.zhu@linux.dev>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA: Don't use %pK through printk

In the past %pK was preferable to %p as it would not leak raw pointer
values into the kernel log.
Since commit ad67b74d2469 ("printk: hash addresses printed with %p")
the regular %p has been improved to avoid this issue.
Furthermore, restricted pointers ("%pK") were never meant to be used
through printk(). They can still unintentionally leak raw pointers or
acquire sleeping looks in atomic contexts.

Switch to the regular pointer formatting which is safer and
easier to reason about.

Signed-off-by: Thomas Weißschuh <thomas.weissschuh@linutronix.de>
Link: https://patch.msgid.link/20250407-restricted-pointers-infiniband-v1-1-22b20504b84d@linutronix.de
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/rxe: Enable ODP in ATOMIC WRITE operation

Add rxe_odp_do_atomic_write() so that ODP specific steps are applied to
ATOMIC WRITE requests.

Signed-off-by: Daisuke Matsuda <matsuda-daisuke@fujitsu.com>
Link: https://patch.msgid.link/20250324075649.3313968-3-matsuda-daisuke@fujitsu.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>

RDMA/rxe: Enable ODP in RDMA FLUSH operation

For persistent memories, add rxe_odp_flush_pmem_iova() so that ODP specific
steps are executed. Otherwise, no additional consideration is required.

Signed-off-by: Daisuke Matsuda <matsuda-daisuke@fujitsu.com>
Link: https://patch.msgid.link/20250324075649.3313968-2-matsuda-daisuke@fujitsu.com
Reviewed-by: Li Zhijian <lizhijian@fujitsu.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>

IB/cm: use rwlock for MAD agent lock

In workloads where there are many processes establishing connections using
RDMA CM in parallel (large scale MPI), there can be heavy contention for
mad_agent_lock in cm_alloc_msg.

This contention can occur while inside of a spin_lock_irq region, leading
to interrupts being disabled for extended durations on many
cores. Furthermore, it leads to the serialization of rdma_create_ah calls,
which has negative performance impacts for NICs which are capable of
processing multiple address handle creations in parallel.

The end result is the machine becoming unresponsive, hung task warnings,
netdev TX timeouts, etc.

Since the lock appears to be only for protection from cm_remove_one, it
can be changed to a rwlock to resolve these issues.

Reproducer:

Server:
  for i in $(seq 1 512); do
    ucmatose -c 32 -p $((i + 5000)) &
  done

Client:
  for i in $(seq 1 512); do
    ucmatose -c 32 -p $((i + 5000)) -s 10.2.0.52 &
  done

Fixes: 76039ac9095f ("IB/cm: Protect cm_dev, cm_ports and mad_agent with kref and lock")
Link: https://patch.msgid.link/r/20250220175612.2763122-1-jmoroni@google.com
Signed-off-by: Jacob Moroni <jmoroni@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Zhu Yanjun <yanjun.zhu@linux.dev>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA/hns: Remove unused parameters

Remove unused parameters.

Link: https://patch.msgid.link/r/20250327114724.3454268-2-huangjunxian6@hisilicon.com
Signed-off-by: Chengchang Tang <tangchengchang@huawei.com>
Signed-off-by: Junxian Huang <huangjunxian6@hisilicon.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

IB/hfi1: Avoid -Wflex-array-member-not-at-end warning

-Wflex-array-member-not-at-end was introduced in GCC-14, and we are
getting ready to enable it, globally.

Remove unused flex-array member `class_data` from
`struct opa_mad_notice_attr`.

Fix the following warning:

drivers/infiniband/hw/hfi1/mad.c:23:36: warning: structure containing a flexible array member is not at the end of another structure [-Wflex-array-member-not-at-end]

Link: https://patch.msgid.link/r/Z-wiYkll8Vo3ME3P@kspp
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA/core: Convert to use ERR_CAST()

As opposed to open-code, using the ERR_CAST macro clearly indicates that
this is a pointer to an error value and a type conversion was performed.

Link: https://patch.msgid.link/r/20250401211015750qxOfU9XZ8QgKizM1Lcyq2@zte.com.cn
Signed-off-by: Li Haoran <li.haoran7@zte.com.cn>
Signed-off-by: Shao Mingyin <shao.mingyin@zte.com.cn>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA/uverbs: Convert to use ERR_CAST()

As opposed to open-code, using the ERR_CAST macro clearly indicates that
this is a pointer to an error value and a type conversion was performed.

Link: https://patch.msgid.link/r/202504012109233981_YPVbd4wQzmAzP3tA5IG@zte.com.cn
Signed-off-by: Li Haoran <li.haoran7@zte.com.cn>
Signed-off-by: Shao Mingyin <shao.mingyin@zte.com.cn>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA/core: Convert to use ERR_CAST()

As opposed to open-code, using the ERR_CAST macro clearly indicates that
this is a pointer to an error value and a type conversion was performed.

Link: https://patch.msgid.link/r/20250401210840146_IyrV3zlejzz3eAnDmMSB@zte.com.cn
Signed-off-by: Li Haoran <li.haoran7@zte.com.cn>
Signed-off-by: Shao Mingyin <shao.mingyin@zte.com.cn>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA: Replace msecs_to_jiffies with secs_to_jiffies for timeout

In drivers/infiniband/hw/mlx4/mcg.c and drivers/infiniband/hw/mlx5/mr.c,
`msecs_to_jiffies` is used to convert milliseconds to jiffies.
For constant milliseconds, using `msecs_to_jiffies` introduces additional
computational overhead. For example, it is unnecessary to check
if m > jiffies_to_msecs(MAX_JIFFY_OFFSET) or (int)m < 0 for constants,
while using `secs_to_jiffies` can avoid these extra calculations.

Link: https://patch.msgid.link/r/20250326221955611qu6Ix3Pt5WgKvhL6sTySX@zte.com.cn
Signed-off-by: Peng Jiang <jiang.peng9@zte.com.cn>
Signed-off-by: Ye Xingchen <ye.xingchen@zte.com.cn>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

RDMA/mlx5: convert timeouts to secs_to_jiffies()

Commit b35108a51cf7 ("jiffies: Define secs_to_jiffies()") introduced
secs_to_jiffies().  As the value here is a multiple of 1000, use
secs_to_jiffies() instead of msecs_to_jiffies to avoid the multiplication.

This is converted using scripts/coccinelle/misc/secs_to_jiffies.cocci with
the following Coccinelle rules:

@depends on patch@
expression E;
@@

-msecs_to_jiffies(E * 1000)
+secs_to_jiffies(E)

-msecs_to_jiffies(E * MSEC_PER_SEC)
+secs_to_jiffies(E)

Link: https://patch.msgid.link/r/20250219-rdma-secs-to-jiffies-v1-2-b506746561a9@linux.microsoft.com
Signed-off-by: Easwar Hariharan <eahariha@linux.microsoft.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

Linux 6.15-rc1

tools/include: make uapi/linux/types.h usable from assembly

The "real" linux/types.h UAPI header gracefully degrades to a NOOP when
included from assembly code.

Mirror this behaviour in the tools/ variant.

Test for __ASSEMBLER__ over __ASSEMBLY__ as the former is provided by the
toolchain automatically.

Reported-by: Mark Brown <broonie@kernel.org>
Closes: https://lore.kernel.org/lkml/af553c62-ca2f-4956-932c-dd6e3a126f58@sirena.org.uk/
Fixes: c9fbaa879508 ("selftests: vDSO: parse_vdso: Use UAPI headers instead of libc headers")
Signed-off-by: Thomas Weißschuh <thomas.weissschuh@linutronix.de>
Link: https://patch.msgid.link/20250321-uapi-consistency-v1-1-439070118dc0@linutronix.de
Signed-off-by: Mark Brown <broonie@kernel.org>
Reviewed-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Merge tag 'turbostat-2025.05.06' of git://git./linux/kernel/git/lenb/linux

Pull turbostat updates from Len Brown:

 - support up to 8192 processors

 - add cpuidle governor debug telemetry, disabled by default

 - update default output to exclude cpuidle invocation counts

 - bug fixes

* tag 'turbostat-2025.05.06' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux:
  tools/power turbostat: v2025.05.06
  tools/power turbostat: disable "cpuidle" invocation counters, by default
  tools/power turbostat: re-factor sysfs code
  tools/power turbostat: Restore GFX sysfs fflush() call
  tools/power turbostat: Document GNR UncMHz domain convention
  tools/power turbostat: report CoreThr per measurement interval
  tools/power turbostat: Increase CPU_SUBSET_MAXCPUS to 8192
  tools/power turbostat: Add idle governor statistics reporting
  tools/power turbostat: Fix names matching
  tools/power turbostat: Allow Zero return value for some RAPL registers
  tools/power turbostat: Clustered Uncore MHz counters should honor show/hide options

Merge tag 'soundwire-6.15-rc1-fixes' of git://git./linux/kernel/git/vkoul/soundwire

Pull soundwire fix from Vinod Koul:

 - add missing config symbol CONFIG_SND_HDA_EXT_CORE required for asoc
   driver CONFIG_SND_SOF_SOF_HDA_SDW_BPT

* tag 'soundwire-6.15-rc1-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vkoul/soundwire:
  ASoC: SOF: Intel: Let SND_SOF_SOF_HDA_SDW_BPT select SND_HDA_EXT_CORE

tools/power turbostat: v2025.05.06

Support up to 8192 processors
Add cpuidle governor debug telemetry, disabled by default
Update default output to exclude cpuidle invocation counts
Bug fixes

Signed-off-by: Len Brown <len.brown@intel.com>

tools/power turbostat: disable "cpuidle" invocation counters, by default

Create "pct_idle" counter group, the sofware notion of residency
so it can now be singled out, independent of other counter groups.

Create "cpuidle" group, the cpuidle invocation counts.
Disable "cpuidle", by default.

Create "swidle" = "cpuidle" + "pct_idle".
Undocument "sysfs", the old name for "swidle", but keep it working
for backwards compatibilty.

Create "hwidle", all the HW idle counters

Modify "idle", enabled by default
"idle" = "hwidle" + "pct_idle" (and now excludes "cpuidle")

Signed-off-by: Len Brown <len.brown@intel.com>

Merge tag 'perf-urgent-2025-04-06' of git://git./linux/kernel/git/tip/tip

Pull perf event fix from Ingo Molnar:
 "Fix a perf events time accounting bug"

* tag 'perf-urgent-2025-04-06' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf/core: Fix child_total_time_enabled accounting bug at task exit

Merge tag 'sched-urgent-2025-04-06' of git://git./linux/kernel/git/tip/tip

Pull scheduler fixes from Ingo Molnar:

 - Fix a nonsensical Kconfig combination

 - Remove an unnecessary rseq-notification

* tag 'sched-urgent-2025-04-06' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  rseq: Eliminate useless task_work on execve
  sched/isolation: Make CONFIG_CPU_ISOLATION depend on CONFIG_SMP

Disable SLUB_TINY for build testing

... and don't error out so hard on missing module descriptions.

Before commit 6c6c1fc09de3 ("modpost: require a MODULE_DESCRIPTION()")
we used to warn about missing module descriptions, but only when
building with extra warnigns (ie 'W=1').

After that commit the warning became an unconditional hard error.

And it turns out not all modules have been converted despite the claims
to the contrary.  As reported by Damian Tometzki, the slub KUnit test
didn't have a module description, and apparently nobody ever really
noticed.

The reason nobody noticed seems to be that the slub KUnit tests get
disabled by SLUB_TINY, which also ends up disabling a lot of other code,
both in tests and in slub itself.  And so anybody doing full build tests
didn't actually see this failre.

So let's disable SLUB_TINY for build-only tests, since it clearly ends
up limiting build coverage.  Also turn the missing module descriptions
error back into a warning, but let's keep it around for non-'W=1'
builds.

Reported-by: Damian Tometzki <damian@riscv-rocks.de>
Link: https://lore.kernel.org/all/01070196099fd059-e8463438-7b1b-4ec8-816d-173874be9966-000000@eu-central-1.amazonses.com/
Cc: Masahiro Yamada <masahiroy@kernel.org>
Cc: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
Fixes: 6c6c1fc09de3 ("modpost: require a MODULE_DESCRIPTION()")
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

tools/power turbostat: re-factor sysfs code

Probe cpuidle "sysfs" residency and counts separately,
since soon we will make one disabled on, and the
other disabled off.

Clarify that some BIC (build-in-counters) are actually "groups".
since we're about to re-name some of those groups.

no functional change.

Signed-off-by: Len Brown <len.brown@intel.com>

tools/power turbostat: Restore GFX sysfs fflush() call

Do fflush() to discard the buffered data, before each read of the
graphics sysfs knobs.

Fixes: ba99a4fc8c24 ("tools/power turbostat: Remove unnecessary fflush() call")
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

tools/power turbostat: Document GNR UncMHz domain convention

Document that on Intel Granite Rapids Systems,
Uncore domains 0-2 are CPU domains, and
uncore domains 3-4 are IO domains.

Signed-off-by: Len Brown <len.brown@intel.com>

tools/power turbostat: report CoreThr per measurement interval

The CoreThr column displays total thermal throttling events
since boot time.

Change it to report events during the measurement interval.

This is more useful for showing a user the current conditions.
Total events since boot time are still available to the user via
/sys/devices/system/cpu/cpu*/thermal_throttle/*

Document CoreThr on turbostat.8

Fixes: eae97e053fe30 ("turbostat: Support thermal throttle count print")
Reported-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Cc: Chen Yu <yu.c.chen@intel.com>

tools/power turbostat: Increase CPU_SUBSET_MAXCPUS to 8192

On systems with >= 1024 cpus (in my case 1152), turbostat fails with the error output:
"turbostat: /sys/fs/cgroup/cpuset.cpus.effective: cpu str malformat 0-1151"

A similar error appears with the use of turbostat --cpu when the inputted cpu
range contains a cpu number >= 1024:
# turbostat -c 1100-1151
"--cpu 1100-1151" malformed
...

Both errors are caused by parse_cpu_str() reaching its limit of CPU_SUBSET_MAXCPUS.

It's a good idea to limit the maximum cpu number being parsed, but 1024 is too low.
For a small increase in compute and allocated memory, increasing CPU_SUBSET_MAXCPUS
brings support for parsing cpu numbers >= 1024.

Increase CPU_SUBSET_MAXCPUS to 8192, a common setting for CONFIG_NR_CPUS on x86_64.

Signed-off-by: Justin Ernst <justin.ernst@hpe.com>
Signed-off-by: Len Brown <len.brown@intel.com>

Merge tag 'timers-cleanups-2025-04-06' of git://git./linux/kernel/git/tip/tip

Pull timer cleanups from Thomas Gleixner:
 "A set of final cleanups for the timer subsystem:

   - Convert all del_timer[_sync]() instances over to the new
     timer_delete[_sync]() API and remove the legacy wrappers.

     Conversion was done with coccinelle plus some manual fixups as
     coccinelle chokes on scoped_guard().

   - The final cleanup of the hrtimer_init() to hrtimer_setup()
     conversion.

     This has been delayed to the end of the merge window, so that all
     patches which have been merged through other trees are in mainline
     and all new users are catched.

  Doing this right before rc1 ensures that new code which is merged post
  rc1 is not introducing new instances of the original functionality"

* tag 'timers-cleanups-2025-04-06' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  tracing/timers: Rename the hrtimer_init event to hrtimer_setup
  hrtimers: Rename debug_init_on_stack() to debug_setup_on_stack()
  hrtimers: Rename debug_init() to debug_setup()
  hrtimers: Rename __hrtimer_init_sleeper() to __hrtimer_setup_sleeper()
  hrtimers: Remove unnecessary NULL check in hrtimer_start_range_ns()
  hrtimers: Make callback function pointer private
  hrtimers: Merge __hrtimer_init() into __hrtimer_setup()
  hrtimers: Switch to use __htimer_setup()
  hrtimers: Delete hrtimer_init()
  treewide: Convert new and leftover hrtimer_init() users
  treewide: Switch/rename to timer_delete[_sync]()

Merge tag 'irq-urgent-2025-04-06' of git://git./linux/kernel/git/tip/tip

Pull more irq updates from Thomas Gleixner:
 "A set of updates for the interrupt subsystem:

   - A treewide cleanup for the irq_domain code, which makes the naming
     consistent and gets rid of the original oddity of naming domains
     'host'.

     This is a trivial mechanical change and is done late to ensure that
     all instances have been catched and new code merged post rc1 wont
     reintroduce new instances.

   - A trivial consistency fix in the migration code

     The recent introduction of irq_force_complete_move() in the core
     code, causes a problem for the nostalgia crowd who maintains ia64
     out of tree.

     The code assumes that hierarchical interrupt domains are enabled
     and dereferences irq_data::parent_data unconditionally. That works
     in mainline because both architectures which enable that code have
     hierarchical domains enabled. Though it breaks the ia64 build,
     which enables the functionality, but does not have hierarchical
     domains.

     While it's not really a problem for mainline today, this
     unconditional dereference is inconsistent and trivially fixable by
     using the existing helper function irqd_get_parent_data(), which
     has the appropriate #ifdeffery in place"

* tag 'irq-urgent-2025-04-06' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  genirq/migration: Use irqd_get_parent_data() in irq_force_complete_move()
  irqdomain: Stop using 'host' for domain
  irqdomain: Rename irq_get_default_host() to irq_get_default_domain()
  irqdomain: Rename irq_set_default_host() to irq_set_default_domain()

Merge tag 'timers-urgent-2025-04-06' of git://git./linux/kernel/git/tip/tip

Pull timer fix from Thomas Gleixner:
 "A revert to fix a adjtimex() regression:

  The recent change to prevent that time goes backwards for the coarse
  time getters due to immediate multiplier adjustments via adjtimex(),
  changed the way how the timekeeping core treats that.

  That change result in a regression on the adjtimex() side, which is
  user space visible:

   1) The forwarding of the base time moves the update out of the
      original period and establishes a new one. That's changing the
      behaviour of the [PF]LL control, which user space expects to be
      applied periodically.

   2) The clearing of the accumulated NTP error due to #1, changes the
      behaviour as well.

  An attempt to delay the multiplier/frequency update to the next tick
  did not solve the problem as userspace expects that the multiplier or
  frequency updates are in effect, when the syscall returns.

  There is a different solution for the coarse time problem available,
  so revert the offending commit to restore the existing adjtimex()
  behaviour"

* tag 'timers-urgent-2025-04-06' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  Revert "timekeeping: Fix possible inconsistencies in _COARSE clockids"

Merge tag 'sh-for-v6.15-tag1' of git://git./linux/kernel/git/glaubitz/sh-linux

Pull sh updates from John Paul Adrian Glaubitz:
 "One important fix and one small configuration update.

  The first patch by Artur Rojek fixes an issue with the J2 firmware
  loader not being able to find the location of the device tree blob due
  to insufficient alignment of the .bss section which rendered J2 boards
  unbootable.

  The second patch by Johan Korsnes updates the defconfigs on sh to drop
  the CONFIG_NET_CLS_TCINDEX configuration option which became obsolete
  after 8c710f75256b ("net/sched: Retire tcindex classifier").

  Summary:

   - sh: defconfig: Drop obsolete CONFIG_NET_CLS_TCINDEX

   - sh: Align .bss section padding to 8-byte boundary"

* tag 'sh-for-v6.15-tag1' of git://git.kernel.org/pub/scm/linux/kernel/git/glaubitz/sh-linux:
  sh: defconfig: Drop obsolete CONFIG_NET_CLS_TCINDEX
  sh: Align .bss section padding to 8-byte boundary

Merge tag 'kbuild-v6.15' of git://git./linux/kernel/git/masahiroy/linux-kbuild

Pull Kbuild updates from Masahiro Yamada:

 - Improve performance in gendwarfksyms

 - Remove deprecated EXTRA_*FLAGS and KBUILD_ENABLE_EXTRA_GCC_CHECKS

 - Support CONFIG_HEADERS_INSTALL for ARCH=um

 - Use more relative paths to sources files for better reproducibility

 - Support the loong64 Debian architecture

 - Add Kbuild bash completion

 - Introduce intermediate vmlinux.unstripped for architectures that need
   static relocations to be stripped from the final vmlinux

 - Fix versioning in Debian packages for -rc releases

 - Treat missing MODULE_DESCRIPTION() as an error

 - Convert Nios2 Makefiles to use the generic rule for built-in DTB

 - Add debuginfo support to the RPM package

* tag 'kbuild-v6.15' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild: (40 commits)
  kbuild: rpm-pkg: build a debuginfo RPM
  kconfig: merge_config: use an empty file as initfile
  nios2: migrate to the generic rule for built-in DTB
  rust: kbuild: skip `--remap-path-prefix` for `rustdoc`
  kbuild: pacman-pkg: hardcode module installation path
  kbuild: deb-pkg: don't set KBUILD_BUILD_VERSION unconditionally
  modpost: require a MODULE_DESCRIPTION()
  kbuild: make all file references relative to source root
  x86: drop unnecessary prefix map configuration
  kbuild: deb-pkg: add comment about future removal of KDEB_COMPRESS
  kbuild: Add a help message for "headers"
  kbuild: deb-pkg: remove "version" variable in mkdebian
  kbuild: deb-pkg: fix versioning for -rc releases
  Documentation/kbuild: Fix indentation in modules.rst example
  x86: Get rid of Makefile.postlink
  kbuild: Create intermediate vmlinux build with relocations preserved
  kbuild: Introduce Kconfig symbol for linking vmlinux with relocations
  kbuild: link-vmlinux.sh: Make output file name configurable
  kbuild: do not generate .tmp_vmlinux*.map when CONFIG_VMLINUX_MAP=y
  Revert "kheaders: Ignore silly-rename files"
  ...

Merge tag 'drm-next-2025-04-05' of https://gitlab.freedesktop.org/drm/kernel

Pull drm fixes from Dave Airlie:
 "Weekly fixes, mostly from the end of last week, this week was very
  quiet, maybe you scared everyone away. It's mostly amdgpu, and xe,
  with some i915, adp and bridge bits, since I think this is overly
  quiet I'd expect rc2 to be a bit more lively.

  bridge:
   - tda998x: Select CONFIG_DRM_KMS_HELPER

  amdgpu:
   - Guard against potential division by 0 in fan code
   - Zero RPM support for SMU 14.0.2
   - Properly handle SI and CIK support being disabled
   - PSR fixes
   - DML2 fixes
   - DP Link training fix
   - Vblank fixes
   - RAS fixes
   - Partitioning fix
   - SDMA fix
   - SMU 13.0.x fixes
   - Rom fetching fix
   - MES fixes
   - Queue reset fix

  xe:
   - Fix NULL pointer dereference on error path
   - Add missing HW workaround for BMG
   - Fix survivability mode not triggering
   - Fix build warning when DRM_FBDEV_EMULATION is not set

  i915:
   - Bounds check for scalers in DSC prefill latency computation
   - Fix build by adding a missing include

  adp:
   - Fix error handling in plane setup"

  # -----BEGIN PGP SIGNATURE-----

* tag 'drm-next-2025-04-05' of https://gitlab.freedesktop.org/drm/kernel: (34 commits)
  drm/i2c: tda998x: select CONFIG_DRM_KMS_HELPER
  drm/amdgpu/gfx12: fix num_mec
  drm/amdgpu/gfx11: fix num_mec
  drm/amd/pm: Add gpu_metrics_v1_8
  drm/amdgpu: Prefer shadow rom when available
  drm/amd/pm: Update smu metrics table for smu_v13_0_6
  drm/amd/pm: Remove host limit metrics support
  Remove unnecessary firmware version check for gc v9_4_2
  drm/amdgpu: stop unmapping MQD for kernel queues v3
  Revert "drm/amdgpu/sdma_v4_4_2: update VM flush implementation for SDMA"
  drm/amdgpu: Parse all deferred errors with UMC aca handle
  drm/amdgpu: Update ta ras block
  drm/amdgpu: Add NPS2 to DPX compatible mode
  drm/amdgpu: Use correct gfx deferred error count
  drm/amd/display: Actually do immediate vblank disable
  drm/amd/display: prevent hang on link training fail
  Revert "drm/amd/display: dml2 soc dscclk use DPM table clk setting"
  drm/amd/display: Increase vblank offdelay for PSR panels
  drm/amd: Handle being compiled without SI or CIK support better
  drm/amd/pm: Add zero RPM enabled OD setting support for SMU14.0.2
  ...

kbuild: rpm-pkg: build a debuginfo RPM

The rpm-pkg make target currently suffers from a few issues related to
debuginfo:
1. debuginfo for things built into the kernel (vmlinux) is not available
   in any RPM produced by make rpm-pkg. This makes using tools like
   systemtap against a make rpm-pkg kernel impossible.
2. debug source for the kernel is not available. This means that
   commands like 'disas /s' in gdb, which display source intermixed with
   assembly, can only print file names/line numbers which then must be
   painstakingly resolved to actual source in a separate editor.
3. debuginfo for modules is available, but it remains bundled with the
   .ko files that contain module code, in the main kernel RPM. This is a
   waste of space for users who do not need to debug the kernel (i.e.
   most users).

Address all of these issues by additionally building a debuginfo RPM
when the kernel configuration allows for it, in line with standard
patterns followed by RPM distributors. With these changes:
1. systemtap now works (when these changes are backported to 6.11, since
   systemtap lags a bit behind in compatibility), as verified by the
   following simple test script:

   # stap -e 'probe kernel.function("do_sys_open").call { printf("%s\n", $$parms); }'
   dfd=0xffffffffffffff9c filename=0x7fe18800b160 flags=0x88800 mode=0x0
   ...

2. disas /s works correctly in gdb, with source and disassembly
   interspersed:

   # gdb vmlinux --batch -ex 'disas /s blk_op_str'
   Dump of assembler code for function blk_op_str:
   block/blk-core.c:
   125     {
      0xffffffff814c8740 <+0>:     endbr64

   127
   128             if (op < ARRAY_SIZE(blk_op_name) && blk_op_name[op])
      0xffffffff814c8744 <+4>:     mov    $0xffffffff824a7378,%rax
      0xffffffff814c874b <+11>:    cmp    $0x23,%edi
      0xffffffff814c874e <+14>:    ja     0xffffffff814c8768 <blk_op_str+40>
      0xffffffff814c8750 <+16>:    mov    %edi,%edi

   126             const char *op_str = "UNKNOWN";
      0xffffffff814c8752 <+18>:    mov    $0xffffffff824a7378,%rdx

   127
   128             if (op < ARRAY_SIZE(blk_op_name) && blk_op_name[op])
      0xffffffff814c8759 <+25>:    mov    -0x7dfa0160(,%rdi,8),%rax

   126             const char *op_str = "UNKNOWN";
      0xffffffff814c8761 <+33>:    test   %rax,%rax
      0xffffffff814c8764 <+36>:    cmove  %rdx,%rax

   129                     op_str = blk_op_name[op];
   130
   131             return op_str;
   132     }
      0xffffffff814c8768 <+40>:    jmp    0xffffffff81d01360 <__x86_return_thunk>
   End of assembler dump.

3. The size of the main kernel package goes down substantially,
   especially if many modules are built (quite typical). Here is a
   comparison of installed size of the kernel package (configured with
   allmodconfig, dwarf4 debuginfo, and module compression turned off)
   before and after this patch:

   # rpm -qi kernel-6.13* | grep -E '^(Version|Size)'
   Version     : 6.13.0postpatch+
   Size        : 1382874089
   Version     : 6.13.0prepatch+
   Size        : 17870795887

   This is a ~92% size reduction.

Note that a debuginfo package can only be produced if the following
configs are set:
- CONFIG_DEBUG_INFO=y
- CONFIG_MODULE_COMPRESS=n
- CONFIG_DEBUG_INFO_SPLIT=n

The first of these is obvious - we can't produce debuginfo if the build
does not generate it. The second two requirements can in principle be
removed, but doing so is difficult with the current approach, which uses
a generic rpmbuild script find-debuginfo.sh that processes all packaged
executables. If we want to remove those requirements the best path
forward is likely to add some debuginfo extraction/installation logic to
the modules_install target (controllable by flags). That way, it's
easier to operate on modules before they're compressed, and the logic
can be reused by all packaging targets.

Signed-off-by: Uday Shankar <ushankar@purestorage.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

kconfig: merge_config: use an empty file as initfile

The scripts/kconfig/merge_config.sh script requires an existing
$INITFILE (or the $1 argument) as a base file for merging Kconfig
fragments. However, an empty $INITFILE can serve as an initial starting
point, later referenced by the KCONFIG_ALLCONFIG Makefile variable
if -m is not used. This variable can point to any configuration file
containing preset config symbols (the merged output) as stated in
Documentation/kbuild/kconfig.rst. When -m is used $INITFILE will
contain just the merge output requiring the user to run make (i.e.
KCONFIG_ALLCONFIG=<$INITFILE> make <allnoconfig/alldefconfig> or make
olddefconfig).

Instead of failing when `$INITFILE` is missing, create an empty file and
use it as the starting point for merges.

Signed-off-by: Daniel Gomez <da.gomez@samsung.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

nios2: migrate to the generic rule for built-in DTB

Commit 654102df2ac2 ("kbuild: add generic support for built-in boot
DTBs") introduced generic support for built-in DTBs.

Select GENERIC_BUILTIN_DTB when built-in DTB support is enabled.

To keep consistency across architectures, this commit also renames
CONFIG_NIOS2_DTB_SOURCE_BOOL to CONFIG_BUILTIN_DTB, and
CONFIG_NIOS2_DTB_SOURCE to CONFIG_BUILTIN_DTB_NAME.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

sh: defconfig: Drop obsolete CONFIG_NET_CLS_TCINDEX

This option was removed from Kconfig in 8c710f75256b ("net/sched:
Retire tcindex classifier") but from the defconfigs.

Fixes: 8c710f75256b ("net/sched: Retire tcindex classifier")
Signed-off-by: Johan Korsnes <johan.korsnes@gmail.com>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Cc: Rich Felker <dalias@libc.org>
Cc: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>
Signed-off-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>

sh: Align .bss section padding to 8-byte boundary

J2-based devices expect to find a device tree blob at the end of the
.bss section. As of a77725a9a3c5 ("scripts/dtc: Update to upstream
version v1.6.1-19-g0a3a9d3449c8"), libfdt enforces 8-byte alignment
for the DTB, causing J2 devices to fail early in sh_fdt_init().

As the J2 loader firmware calculates the DTB location based on the kernel
image .bss section size rather than the __bss_stop symbol offset, the
required alignment can't be enforced with BSS_SECTION(0, PAGE_SIZE, 8).

To fix this, inline a modified version of the above macro which grows
.bss by the required size. While this change affects all existing SH
boards, it should be benign on platforms which don't need this alignment.

Signed-off-by: Artur Rojek <contact@artur-rojek.eu>
Reviewed-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>
Tested-by: Rob Landley <rob@landley.net>
Signed-off-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>

Merge tag 'input-for-v6.15-rc0' of git://git./linux/kernel/git/dtor/input

Pull input updates from Dmitry Torokhov:

 - a brand new driver for touchpads and touchbars in newer Apple devices

 - support for Berlin-A series in goodix-berlin touchscreen driver

 - improvements to matrix_keypad driver to better handle GPIOs toggling

 - assorted small cleanups in other input drivers

* tag 'input-for-v6.15-rc0' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: goodix_berlin - add support for Berlin-A series
  dt-bindings: input: goodix,gt9916: Document gt9897 compatible
  dt-bindings: input: matrix_keypad - add wakeup-source property
  dt-bindings: input: matrix_keypad - add missing property
  Input: pm8941-pwrkey - fix dev_dbg() output in pm8941_pwrkey_irq()
  Input: synaptics - hide unused smbus_pnp_ids[] array
  Input: apple_z2 - fix potential confusion in Kconfig
  Input: matrix_keypad - use fsleep for delays after activating columns
  Input: matrix_keypad - add settle time after enabling all columns
  dt-bindings: input: matrix_keypad: add settle time after enabling all columns
  dt-bindings: input: matrix_keypad: convert to YAML
  dt-bindings: input: Correct indentation and style in DTS example
  MAINTAINERS: Add entries for Apple Z2 touchscreen driver
  Input: apple_z2 - add a driver for Apple Z2 touchscreens
  dt-bindings: input: touchscreen: Add Z2 controller
  Input: Switch to use hrtimer_setup()
  Input: drop vb2_ops_wait_prepare/finish

tracing/timers: Rename the hrtimer_init event to hrtimer_setup

The function hrtimer_init() doesn't exist anymore. It was replaced by
hrtimer_setup().

Thus, rename the hrtimer_init trace event to hrtimer_setup to keep it
consistent.

Signed-off-by: Nam Cao <namcao@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/all/cba84c3d853c5258aa3a262363a6eac08e2c7afc.1738746927.git.namcao@linutronix.de

hrtimers: Rename debug_init_on_stack() to debug_setup_on_stack()

All the hrtimer_init*() functions have been renamed to hrtimer_setup*().
Rename debug_init_on_stack() to debug_setup_on_stack() as well, to keep the
names consistent.

Signed-off-by: Nam Cao <namcao@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/all/073cf6162779a2f5b12624677d4c49ee7eccc1ed.1738746927.git.namcao@linutronix.de

hrtimers: Rename debug_init() to debug_setup()

All the hrtimer_init*() functions have been renamed to hrtimer_setup*().
Rename debug_init() to debug_setup() as well, to keep the names consistent.

Signed-off-by: Nam Cao <namcao@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/all/4b730c1f79648b16a1c5413f928fdc2e138dfc43.1738746927.git.namcao@linutronix.de