Re: [PATCH v8 3/4] misc: fastrpc: Allocate entire reserved memory for Audio PD in probe

[Jianping Li <jianping.li@oss.qualcomm.com>]

On 7/2/2026 4:37 AM, Srinivas Kandagatla wrote:
>
> On 6/9/26 3:59 AM, Jianping Li wrote:
>> Allocating and freeing Audio PD memory from userspace is unsafe because
>> the kernel cannot reliably determine when the DSP has finished using the
>> memory. Userspace may free buffers while they are still in use by the DSP,
>> and remote free requests cannot be safely trusted.
>>
>> Additionally, the current implementation allows userspace to repeatedly
>> grow the Audio PD heap, but does not support shrinking it. This can lead
>> to unbounded memory usage over time, effectively causing a memory leak.
>>
>> Fix this by allocating the entire Audio PD reserved-memory region during
>> rpmsg probe and tying its lifetime to the rpmsg channel. This removes
>> userspace-controlled alloc/free and ensures that memory is reclaimed only
>> when the DSP process is torn down.
>>
>> Add explicit validation for remote_heap presence and size before sending
>> the memory to DSP, and fail early if the reserved-memory region is
>> missing or incomplete.
>>
>> Fixes: 0871561055e66 ("misc: fastrpc: Add support for audiopd")
>> Cc: stable@kernel.org
>> Signed-off-by: Jianping Li <jianping.li@oss.qualcomm.com>
>> ---
>>   drivers/misc/fastrpc.c | 112 ++++++++++++++++++++++-------------------
>>   1 file changed, 59 insertions(+), 53 deletions(-)
>>
>> diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c
>> index 517884000331..1942e74535e5 100644
>> --- a/drivers/misc/fastrpc.c
>> +++ b/drivers/misc/fastrpc.c
>> @@ -276,6 +276,8 @@ struct fastrpc_channel_ctx {
>>   	struct kref refcount;
>>   	/* Flag if dsp attributes are cached */
>>   	bool valid_attributes;
>> +	/* Flag if audio PD init mem was allocated */
>> +	bool audio_init_mem;
>>   	u32 dsp_attributes[FASTRPC_MAX_DSP_ATTRIBUTES];
>>   	struct fastrpc_device *secure_fdevice;
>>   	struct fastrpc_device *fdevice;
>> @@ -1341,15 +1343,24 @@ static int fastrpc_init_create_static_process(struct fastrpc_user *fl,
>>   	struct fastrpc_init_create_static init;
>>   	struct fastrpc_invoke_args *args;
>>   	struct fastrpc_phy_page pages[1];
>> +	struct fastrpc_channel_ctx *cctx = fl->cctx;
>>   	char *name;
>>   	int err;
>> -	bool scm_done = false;
>>   	struct {
>>   		int client_id;
>>   		u32 namelen;
>>   		u32 pageslen;
>>   	} inbuf;
>>   	u32 sc;
>> +	unsigned long flags;
>> +
>> +	if (!cctx->remote_heap || !cctx->remote_heap->dma_addr ||
>> +	    !cctx->remote_heap->size) {
>> +		err = -ENOMEM;
>> +		dev_err(fl->sctx->dev,
>> +			"remote heap memory region is not added\n");
>> +		return err;
>> +	}
>>   
>>   	args = kzalloc_objs(*args, FASTRPC_CREATE_STATIC_PROCESS_NARGS);
>>   	if (!args)
>> @@ -1373,31 +1384,6 @@ static int fastrpc_init_create_static_process(struct fastrpc_user *fl,
>>   	inbuf.client_id = fl->client_id;
>>   	inbuf.namelen = init.namelen;
>>   	inbuf.pageslen = 0;
>> -	if (!fl->cctx->remote_heap) {
>> -		err = fastrpc_remote_heap_alloc(fl, fl->sctx->dev, init.memlen,
>> -						&fl->cctx->remote_heap);
>> -		if (err)
>> -			goto err_name;
>> -
>> -		/* Map if we have any heap VMIDs associated with this ADSP Static Process. */
>> -		if (fl->cctx->vmcount) {
>> -			u64 src_perms = BIT(QCOM_SCM_VMID_HLOS);
>> -
>> -			err = qcom_scm_assign_mem(fl->cctx->remote_heap->dma_addr,
>> -							(u64)fl->cctx->remote_heap->size,
>> -							&src_perms,
>> -							fl->cctx->vmperms, fl->cctx->vmcount);
>> -			if (err) {
>> -				dev_err(fl->sctx->dev,
>> -					"Failed to assign memory with dma_addr %pad size 0x%llx err %d\n",
>> -					&fl->cctx->remote_heap->dma_addr,
>> -					fl->cctx->remote_heap->size, err);
>> -				goto err_map;
>> -			}
>> -			scm_done = true;
>> -			inbuf.pageslen = 1;
>> -		}
>> -	}
>>   
>>   	fl->pd = USER_PD;
>>   
>> @@ -1409,8 +1395,17 @@ static int fastrpc_init_create_static_process(struct fastrpc_user *fl,
>>   	args[1].length = inbuf.namelen;
>>   	args[1].fd = -1;
>>   
>> -	pages[0].addr = fl->cctx->remote_heap->dma_addr;
>> -	pages[0].size = fl->cctx->remote_heap->size;
>> +	spin_lock_irqsave(&cctx->lock, flags);
>> +	if (!cctx->audio_init_mem) {
>> +		pages[0].addr = cctx->remote_heap->dma_addr;
>> +		pages[0].size = cctx->remote_heap->size;
>> +		cctx->audio_init_mem = true;
>> +		inbuf.pageslen = 1;
>> +	} else {
>> +		pages[0].addr = 0;
>> +		pages[0].size = 0;
> What is the expected behavoiur in this case?

You're right that this behaviour is subtle and deserves an in-code
comment. I'll add one in v9, right above the spin_lock, along the
lines of:
        /*
	* Audio PD is a static PD and retains the remote heap
	* information across daemon restarts. Therefore only
	* the first attach should provide heap information to
	* DSP. Subsequent attaches reuse the previously
	* initialized memory pool.
	*/

>
>> +	}
>> +	spin_unlock_irqrestore(&cctx->lock, flags);
>>   
>>   	args[2].ptr = (u64)(uintptr_t) pages;
>>   	args[2].length = sizeof(*pages);
>> @@ -1428,27 +1423,7 @@ static int fastrpc_init_create_static_process(struct fastrpc_user *fl,
>>   
>>   	return 0;
>>   err_invoke:
>> -	if (fl->cctx->vmcount && scm_done) {
>> -		u64 src_perms = 0;
>> -		struct qcom_scm_vmperm dst_perms;
>> -		u32 i;
>> -
>> -		for (i = 0; i < fl->cctx->vmcount; i++)
>> -			src_perms |= BIT(fl->cctx->vmperms[i].vmid);
>> -
>> -		dst_perms.vmid = QCOM_SCM_VMID_HLOS;
>> -		dst_perms.perm = QCOM_SCM_PERM_RWX;
>> -		err = qcom_scm_assign_mem(fl->cctx->remote_heap->dma_addr,
>> -						(u64)fl->cctx->remote_heap->size,
>> -						&src_perms, &dst_perms, 1);
>> -		if (err)
>> -			dev_err(fl->sctx->dev, "Failed to assign memory dma_addr %pad size 0x%llx err %d\n",
>> -				&fl->cctx->remote_heap->dma_addr, fl->cctx->remote_heap->size, err);
>> -	}
>> -err_map:
>> -	fastrpc_buf_free(fl->cctx->remote_heap);
>> -	fl->cctx->remote_heap = NULL;
>> -err_name:
>> +	cctx->audio_init_mem = false;
> this can race.

Thanks for the review. I'd like to clarify the intent here — please
let me know if I'm missing something.

The lock is not really there to protect 'audio_init_mem' itself,
but to make sure the following is only sent to the DSP once:

     pages[0].addr = fl->cctx->remote_heap->phys;
     pages[0].size = fl->cctx->remote_heap->size;

The flow I had in mind is:

   - Thread A enters the critical section, sees audio_init_mem == false,
     sets it to true, and proceeds to send the pages to the DSP.
   - If Thread A's invoke fails, we reset audio_init_mem back to false
     in err_invoke so that a later attempt can retry sending the region.
   - Once any thread successfully sends the region, audio_init_mem stays
     true forever and the false branch is never taken again.

So the reset on failure is only meant to allow a retry, not to be
racy with a concurrent successful path (there can't be one, since
only one thread is inside the critical section at a time). Do you
see a specific interleaving I've overlooked? If so, I'm happy to
switch to a different scheme.

>
>>   	kfree(name);
>>   err:
>>   	kfree(args);
>> @@ -2415,12 +2390,23 @@ static int fastrpc_rpmsg_probe(struct rpmsg_device *rpdev)
>>   		}
>>   	}
>>   
>> -	if (domain_id == SDSP_DOMAIN_ID) {
>> +	if (domain_id == SDSP_DOMAIN_ID || domain_id == ADSP_DOMAIN_ID) {
>>   		struct resource res;
>>   		u64 src_perms;
>>   
>>   		err = of_reserved_mem_region_to_resource(rdev->of_node, 0, &res);
>>   		if (!err) {
>> +			if (domain_id == ADSP_DOMAIN_ID) {
>> +				data->remote_heap =
>> +					kzalloc_obj(*data->remote_heap, GFP_KERNEL);
>> +				if (!data->remote_heap) {
>> +					err = -ENOMEM;
>> +					goto err_free_data;
>> +				}
>> +
>> +				data->remote_heap->dma_addr = res.start;
>> +				data->remote_heap->size = resource_size(&res);
>> +			}
>>   			src_perms = BIT(QCOM_SCM_VMID_HLOS);
>>   
>>   			err = qcom_scm_assign_mem(res.start, resource_size(&res), &src_perms,
>> @@ -2428,7 +2414,6 @@ static int fastrpc_rpmsg_probe(struct rpmsg_device *rpdev)
>>   			if (err)
>>   				goto err_free_data;
>>   		}
>> -
>>   	}
>>   
>>   	secure_dsp = !(of_property_read_bool(rdev->of_node, "qcom,non-secure-domain"));
>> @@ -2487,6 +2472,7 @@ static int fastrpc_rpmsg_probe(struct rpmsg_device *rpdev)
>>   		misc_deregister(&data->secure_fdevice->miscdev);
>>   
>>   err_free_data:
>> +	kfree(data->remote_heap);
>>   	kfree(data);
>>   	return err;
>>   }
>> @@ -2509,6 +2495,7 @@ static void fastrpc_rpmsg_remove(struct rpmsg_device *rpdev)
>>   	struct fastrpc_buf *buf, *b;
>>   	struct fastrpc_user *user;
>>   	unsigned long flags;
>> +	int err;
>>   
>>   	/* No invocations past this point */
>>   	spin_lock_irqsave(&cctx->lock, flags);
>> @@ -2526,8 +2513,27 @@ static void fastrpc_rpmsg_remove(struct rpmsg_device *rpdev)
>>   	list_for_each_entry_safe(buf, b, &cctx->invoke_interrupted_mmaps, node)
>>   		list_del(&buf->node);
>>   
>> -	if (cctx->remote_heap)
>> -		fastrpc_buf_free(cctx->remote_heap);
>> +	if (cctx->remote_heap && cctx->vmcount) {
>> +		u64 src_perms = 0;
>> +		struct qcom_scm_vmperm dst_perms;
>> +
>> +		for (u32 i = 0; i < cctx->vmcount; i++)
>> +			src_perms |= BIT(cctx->vmperms[i].vmid);
>> +
>> +		dst_perms.vmid = QCOM_SCM_VMID_HLOS;
>> +		dst_perms.perm = QCOM_SCM_PERM_RWX;
>> +
>> +		err = qcom_scm_assign_mem(cctx->remote_heap->dma_addr,
>> +					  cctx->remote_heap->size, &src_perms,
>> +					  &dst_perms, 1);
>> +		if (err)
>> +			dev_err(&rpdev->dev,
>> +				"Failed to assign memory back to HLOS: dma_addr %pad size %#llx err %d\n",
>> +				&cctx->remote_heap->dma_addr, cctx->remote_heap->size, err);
>> +	}
>> +
>> +	kfree(cctx->remote_heap);
>> +	cctx->remote_heap = NULL;
>>   
>>   	of_platform_depopulate(&rpdev->dev);
>>