smu_cmn_sen...g_with_param(9)

✨ smu_cmn_send_smc_msg_with_param ✨

okay so bestie, this one is from the Kernel API Manual and honestly it's kind of a baddie — it's the function that literally slides a message into the SMU and waits for it to text back 💌

---

🎀 SYNOPSIS 🎀

int smu_cmn_send_smc_msg_with_param (struct smu_context *smu , enum
smu_message_type msg , uint32_t param , uint32_t *read_arg );

---

💅 ARGUMENTS 💅

• smu — pointer to an SMU context, aka the whole vibe check of where we're sending from 🫶
• msg — the message to send, bestie. this is the actual thing we're shipping off 📨
• param — the parameter to send to the SMU, tagging along for the ride ✨
• read_arg — pointer to u32 used to return a value from the SMU back to the caller, like a little reply note passed back to you 💌

---

🌸 DESCRIPTION 🌸

So here's the tea: this function sends the message msg with parameter param to the SMU, waits for the command to be completed (yes, it waits, it's a patient icon), and then hands a value from the SMU back to you through the read_arg pointer. periodt. 💅

It returns 0 on success, and -errno when something goes wrong while sending the message or receiving the reply. But — and this is the important part, ngl — if there's a PCI bus recovery happening, or the destination is a virtual GPU that doesn't allow this message type, the message is just straight up dropped and success is returned anyway 🙈. If you want the details of what that -errno actually means, go look at smu_msg_v1_decode_response.

If we weren't able to send the message to the SMU at all, we also print the error to the standard log. That's our little receipts moment 🔥

Command completion status gets printed only if the -errno is -EREMOTEIO — which is the signal that the SMU returned back an undefined/unknown/unspecified result 😭. All the other cases are well-defined, so they don't get printed; instead they're handed back to the client so it can decide what to do next. main character energy: the client gets to make the call 👑

And the return value read_arg is read back regardless, because we want to give the client as much info as possible. On error it would most likely be param, but like — we can't just assume that, bestie 🫠. This also eliminates more conditionals, which is honestly iconic of it.

---

🦋 SEE ALSO 🦋

• Kernel file ./drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c
• smu_cmn_wait_for_response(9)
• smu_cmn_send_smc_msg_with_params_ext(9)
• smu_msg_v1_send_msg(9)
• smu_msg_send_async_locked(9)
• smu_cmn_disable_all_features_with_exception(9)
• smu_cmn_get_pptable_from_firmware(9)

---

⭐ ok bye bestie, this page came to you from ./drivers/gpu/drm/amd/pm/swsmu, dated October 2026, and you're reading the Kernel API Manual 💖 go forth and send your messages, you've got this 🧚💜