From: Mario Limonciello <superm1@kernel.org>
To: Lizhi Hou <lizhi.hou@amd.com>,
ogabbay@kernel.org, quic_jhugo@quicinc.com,
maciej.falkowski@linux.intel.com,
dri-devel@lists.freedesktop.org
Cc: linux-kernel@vger.kernel.org, max.zhen@amd.com, sonal.santan@amd.com
Subject: Re: [PATCH V1] accel/amdxdna: Support preemption requests
Date: Tue, 4 Nov 2025 13:30:17 -0600 [thread overview]
Message-ID: <851cd32b-e64e-4e56-bf49-7c8b3336815f@kernel.org> (raw)
In-Reply-To: <8c9b2d14-96c4-d917-226e-b1422d1390c8@amd.com>
On 11/4/25 1:28 PM, Lizhi Hou wrote:
>
> On 11/4/25 10:58, Mario Limonciello wrote:
>> On 11/4/25 12:53 PM, Lizhi Hou wrote:
>>> The driver checks the firmware version during initialization.If
>>> preemption
>>> is supported, the driver configures preemption accordingly and handles
>>> userspace preemption requests. Otherwise, the driver returns an error
>>> for
>>> userspace preemption requests.
>>>
>>> Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
>>> ---
>>> drivers/accel/amdxdna/aie2_message.c | 95 +++++++++++++++++++++++++
>>> drivers/accel/amdxdna/aie2_msg_priv.h | 3 +
>>> drivers/accel/amdxdna/aie2_pci.c | 63 ++++++++++++++++
>>> drivers/accel/amdxdna/aie2_pci.h | 8 +++
>>> drivers/accel/amdxdna/amdxdna_ctx.h | 17 +++++
>>> drivers/accel/amdxdna/amdxdna_pci_drv.c | 3 +-
>>> drivers/accel/amdxdna/npu4_regs.c | 4 ++
>>> include/uapi/drm/amdxdna_accel.h | 16 ++++-
>>> 8 files changed, 207 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/drivers/accel/amdxdna/aie2_message.c b/drivers/accel/
>>> amdxdna/aie2_message.c
>>> index 69cdce9ff208..d493bb1c3360 100644
>>> --- a/drivers/accel/amdxdna/aie2_message.c
>>> +++ b/drivers/accel/amdxdna/aie2_message.c
>>> @@ -210,6 +210,14 @@ int aie2_create_context(struct amdxdna_dev_hdl
>>> *ndev, struct amdxdna_hwctx *hwct
>>> hwctx->fw_ctx_id = resp.context_id;
>>> WARN_ONCE(hwctx->fw_ctx_id == -1, "Unexpected context id");
>>> + if (ndev->force_preempt_enabled) {
>>> + ret = aie2_runtime_cfg(ndev, AIE2_RT_CFG_FORCE_PREEMPT,
>>> &hwctx->fw_ctx_id);
>>> + if (ret) {
>>> + XDNA_ERR(xdna, "failed to enable force preempt %d", ret);
>>> + return ret;
>>> + }
>>> + }
>>> +
>>> cq_pair = &resp.cq_pair[0];
>>> x2i.mb_head_ptr_reg = AIE2_MBOX_OFF(ndev, cq_pair-
>>> >x2i_q.head_addr);
>>> x2i.mb_tail_ptr_reg = AIE2_MBOX_OFF(ndev, cq_pair-
>>> >x2i_q.tail_addr);
>>> @@ -601,6 +609,11 @@ aie2_cmdlist_fill_dpu(struct amdxdna_gem_obj
>>> *cmd_bo, void *slot, size_t *size)
>>> return 0;
>>> }
>>> +static int aie2_cmdlist_unsupp(struct amdxdna_gem_obj *cmd_bo,
>>> void *slot, size_t *size)
>>> +{
>>> + return -EOPNOTSUPP;
>>> +}
>>> +
>>> static u32 aie2_get_chain_msg_op(u32 cmd_op)
>>> {
>>> switch (cmd_op) {
>>> @@ -621,6 +634,8 @@ static struct aie2_exec_msg_ops
>>> legacy_exec_message_ops = {
>>> .init_chain_req = aie2_init_exec_chain_req,
>>> .fill_cf_slot = aie2_cmdlist_fill_cf,
>>> .fill_dpu_slot = aie2_cmdlist_fill_dpu,
>>> + .fill_preempt_slot = aie2_cmdlist_unsupp,
>>> + .fill_elf_slot = aie2_cmdlist_unsupp,
>>> .get_chain_msg_op = aie2_get_chain_msg_op,
>>> };
>>> @@ -680,6 +695,74 @@ aie2_cmdlist_fill_npu_dpu(struct
>>> amdxdna_gem_obj *cmd_bo, void *slot, size_t *si
>>> return 0;
>>> }
>>> +static int
>>> +aie2_cmdlist_fill_npu_preempt(struct amdxdna_gem_obj *cmd_bo, void
>>> *slot, size_t *size)
>>> +{
>>> + struct cmd_chain_slot_npu *npu_slot = slot;
>>> + struct amdxdna_cmd_preempt_data *pd;
>>> + u32 cmd_len;
>>> + u32 arg_sz;
>>> +
>>> + pd = amdxdna_cmd_get_payload(cmd_bo, &cmd_len);
>>> + arg_sz = cmd_len - sizeof(*pd);
>>> + if (cmd_len < sizeof(*pd) || arg_sz > MAX_NPU_ARGS_SIZE)
>>> + return -EINVAL;
>>> +
>>> + if (*size < sizeof(*npu_slot) + arg_sz)
>>> + return -EINVAL;
>>> +
>>> + npu_slot->cu_idx = amdxdna_cmd_get_cu_idx(cmd_bo);
>>> + if (npu_slot->cu_idx == INVALID_CU_IDX)
>>> + return -EINVAL;
>>> +
>>> + memset(npu_slot, 0, sizeof(*npu_slot));
>>> + npu_slot->type = EXEC_NPU_TYPE_PREEMPT;
>>> + npu_slot->inst_buf_addr = pd->inst_buf;
>>> + npu_slot->save_buf_addr = pd->save_buf;
>>> + npu_slot->restore_buf_addr = pd->restore_buf;
>>> + npu_slot->inst_size = pd->inst_size;
>>> + npu_slot->save_size = pd->save_size;
>>> + npu_slot->restore_size = pd->restore_size;
>>> + npu_slot->inst_prop_cnt = pd->inst_prop_cnt;
>>> + npu_slot->arg_cnt = arg_sz / sizeof(u32);
>>> + memcpy(npu_slot->args, pd->prop_args, arg_sz);
>>
>> Am I following this right? I would think this should be:
>>
>> memcpy(npu_slot->args, pd->prop_args, npu_slot->arg_cnt);
>
> npu_slot->arg_cnt is the number of u32. So arg_sz is used for memcpy here.
>
>
Got it thanks. No other concerns.
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
next prev parent reply other threads:[~2025-11-04 19:30 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-11-04 18:53 [PATCH V1] accel/amdxdna: Support preemption requests Lizhi Hou
2025-11-04 18:58 ` Mario Limonciello
2025-11-04 19:28 ` Lizhi Hou
2025-11-04 19:30 ` Mario Limonciello [this message]
2025-11-05 17:05 ` Lizhi Hou
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=851cd32b-e64e-4e56-bf49-7c8b3336815f@kernel.org \
--to=superm1@kernel.org \
--cc=dri-devel@lists.freedesktop.org \
--cc=linux-kernel@vger.kernel.org \
--cc=lizhi.hou@amd.com \
--cc=maciej.falkowski@linux.intel.com \
--cc=max.zhen@amd.com \
--cc=ogabbay@kernel.org \
--cc=quic_jhugo@quicinc.com \
--cc=sonal.santan@amd.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.