[PATCH v2 0/5] Add multimode invoke request IOCTl support

[PATCH v2 0/5] Add multimode invoke request IOCTl support

[Ekansh Gupta]

Add changes to support multimode invocation ioctl request. This
ioctl call facilitates multiple types of requests from user including
CRC check, performance counters, shared context bank usage, etc.
This series also carries patch to save and restore interrupted
context.

Ekansh Gupta (5):
  misc: fastrpc: Add CRC support for remote buffers
  misc: fastrpc: Capture kernel and DSP performance counters
  misc: fastrpc: Add support to save and restore interrupted
  misc: fastrpc: Add support to save and restore interrupted
  misc: fastrpc: Add support to allocate shared context bank

 drivers/misc/fastrpc.c      | 355 +++++++++++++++++++++++++++++++++++++-------
 include/uapi/misc/fastrpc.h |  26 ++++
 2 files changed, 325 insertions(+), 56 deletions(-)

-- 
2.7.4

[PATCH v2 1/5] misc: fastrpc: Add CRC support for remote buffers

[Ekansh Gupta]

CRC check for input and output argument helps in ensuring data
consistency over a remote call. If user intends to enable CRC check,
first local user CRC is calculated at user end and a CRC buffer is
passed to DSP to capture remote CRC values. DSP is expected to
write to the remote CRC buffer which is then compared at user level
with the local CRC values.

Signed-off-by: Ekansh Gupta <quic_ekangupt@quicinc.com>
---
 drivers/misc/fastrpc.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c
index 04eaf6c..337ec1f 100644
--- a/drivers/misc/fastrpc.c
+++ b/drivers/misc/fastrpc.c
@@ -610,6 +610,7 @@ static struct fastrpc_invoke_ctx *fastrpc_context_alloc(
 	/* Released in fastrpc_context_put() */
 	fastrpc_channel_ctx_get(cctx);
 
+	ctx->crc = (u32 *)(uintptr_t)invoke->crc;
 	ctx->sc = sc;
 	ctx->retval = -1;
 	ctx->pid = current->pid;
@@ -1063,6 +1064,7 @@ static int fastrpc_put_args(struct fastrpc_invoke_ctx *ctx,
 	struct fastrpc_invoke_buf *list;
 	struct fastrpc_phy_page *pages;
 	u64 *fdlist;
+	u32 *crclist;
 	int i, inbufs, outbufs, handles;
 
 	inbufs = REMOTE_SCALARS_INBUFS(ctx->sc);
@@ -1070,7 +1072,8 @@ static int fastrpc_put_args(struct fastrpc_invoke_ctx *ctx,
 	handles = REMOTE_SCALARS_INHANDLES(ctx->sc) + REMOTE_SCALARS_OUTHANDLES(ctx->sc);
 	list = fastrpc_invoke_buf_start(rpra, ctx->nscalars);
 	pages = fastrpc_phy_page_start(list, ctx->nscalars);
-	fdlist = (uint64_t *)(pages + inbufs + outbufs + handles);
+	fdlist = (u64 *)(pages + inbufs + outbufs + handles);
+	crclist = (u32 *)(fdlist + FASTRPC_MAX_FDLIST);
 
 	for (i = inbufs; i < ctx->nbufs; ++i) {
 		if (!ctx->maps[i]) {
@@ -1094,6 +1097,10 @@ static int fastrpc_put_args(struct fastrpc_invoke_ctx *ctx,
 			fastrpc_map_put(mmap);
 	}
 
+	if (ctx->crc && crclist && rpra) {
+		if (copy_to_user((void __user *)ctx->crc, crclist, FASTRPC_MAX_CRCLIST * sizeof(u32)))
+			return -EFAULT;
+	}
 	return 0;
 }
 
@@ -1706,6 +1713,7 @@ static int fastrpc_multimode_invoke(struct fastrpc_user *fl, char __user *argp)
 
 	switch (invoke.req) {
 	case FASTRPC_INVOKE:
+	case FASTRPC_INVOKE_ENHANCED:
 		/* nscalars is truncated here to max supported value */
 		if (copy_from_user(&einv, (void __user *)(uintptr_t)invoke.invparam,
 				   invoke.size))
-- 
2.7.4

Re: [PATCH v2 1/5] misc: fastrpc: Add CRC support for remote buffers

[Srinivas Kandagatla]

Hi Ekansh,
Thankyou for the patch,

Looks like this series depends on Multi Mode series, can you include 
them together.

--srini

On 20/09/2023 11:31, Ekansh Gupta wrote:
> CRC check for input and output argument helps in ensuring data
> consistency over a remote call. If user intends to enable CRC check,
> first local user CRC is calculated at user end and a CRC buffer is
> passed to DSP to capture remote CRC values. DSP is expected to
> write to the remote CRC buffer which is then compared at user level
> with the local CRC values.
> 
> Signed-off-by: Ekansh Gupta <quic_ekangupt@quicinc.com>
> ---
>   drivers/misc/fastrpc.c | 10 +++++++++-
>   1 file changed, 9 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c
> index 04eaf6c..337ec1f 100644
> --- a/drivers/misc/fastrpc.c
> +++ b/drivers/misc/fastrpc.c
> @@ -610,6 +610,7 @@ static struct fastrpc_invoke_ctx *fastrpc_context_alloc(
>   	/* Released in fastrpc_context_put() */
>   	fastrpc_channel_ctx_get(cctx);
>   
> +	ctx->crc = (u32 *)(uintptr_t)invoke->crc;
>   	ctx->sc = sc;
>   	ctx->retval = -1;
>   	ctx->pid = current->pid;
> @@ -1063,6 +1064,7 @@ static int fastrpc_put_args(struct fastrpc_invoke_ctx *ctx,
>   	struct fastrpc_invoke_buf *list;
>   	struct fastrpc_phy_page *pages;
>   	u64 *fdlist;
> +	u32 *crclist;
>   	int i, inbufs, outbufs, handles;
>   
>   	inbufs = REMOTE_SCALARS_INBUFS(ctx->sc);
> @@ -1070,7 +1072,8 @@ static int fastrpc_put_args(struct fastrpc_invoke_ctx *ctx,
>   	handles = REMOTE_SCALARS_INHANDLES(ctx->sc) + REMOTE_SCALARS_OUTHANDLES(ctx->sc);
>   	list = fastrpc_invoke_buf_start(rpra, ctx->nscalars);
>   	pages = fastrpc_phy_page_start(list, ctx->nscalars);
> -	fdlist = (uint64_t *)(pages + inbufs + outbufs + handles);
> +	fdlist = (u64 *)(pages + inbufs + outbufs + handles);
> +	crclist = (u32 *)(fdlist + FASTRPC_MAX_FDLIST);
>   
>   	for (i = inbufs; i < ctx->nbufs; ++i) {
>   		if (!ctx->maps[i]) {
> @@ -1094,6 +1097,10 @@ static int fastrpc_put_args(struct fastrpc_invoke_ctx *ctx,
>   			fastrpc_map_put(mmap);
>   	}
>   
> +	if (ctx->crc && crclist && rpra) {
> +		if (copy_to_user((void __user *)ctx->crc, crclist, FASTRPC_MAX_CRCLIST * sizeof(u32)))
> +			return -EFAULT;
> +	}
>   	return 0;
>   }
>   
> @@ -1706,6 +1713,7 @@ static int fastrpc_multimode_invoke(struct fastrpc_user *fl, char __user *argp)
>   
>   	switch (invoke.req) {
>   	case FASTRPC_INVOKE:
> +	case FASTRPC_INVOKE_ENHANCED:
>   		/* nscalars is truncated here to max supported value */
>   		if (copy_from_user(&einv, (void __user *)(uintptr_t)invoke.invparam,
>   				   invoke.size))

[PATCH v2 2/5] misc: fastrpc: Capture kernel and DSP performance counters

[Ekansh Gupta]

Add support to capture kernel performance counters for different
kernel level operations. These counters collects the information
for remote call and copies the information to a buffer shared
by user.

Collection of DSP performance counters is also added as part of
this change. DSP updates the performance information in the
metadata which is then copied to a buffer passed by the user.

Signed-off-by: Ekansh Gupta <quic_ekangupt@quicinc.com>
---
 drivers/misc/fastrpc.c      | 140 +++++++++++++++++++++++++++++++++++++++++---
 include/uapi/misc/fastrpc.h |  14 +++++
 2 files changed, 146 insertions(+), 8 deletions(-)

diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c
index 337ec1f..f8c1c381 100644
--- a/drivers/misc/fastrpc.c
+++ b/drivers/misc/fastrpc.c
@@ -18,6 +18,7 @@
 #include <linux/rpmsg.h>
 #include <linux/scatterlist.h>
 #include <linux/slab.h>
+#include <linux/delay.h>
 #include <linux/firmware/qcom/qcom_scm.h>
 #include <uapi/misc/fastrpc.h>
 #include <linux/of_reserved_mem.h>
@@ -32,6 +33,8 @@
 #define FASTRPC_ALIGN		128
 #define FASTRPC_MAX_FDLIST	16
 #define FASTRPC_MAX_CRCLIST	64
+#define FASTRPC_KERNEL_PERF_LIST (PERF_KEY_MAX)
+#define FASTRPC_DSP_PERF_LIST 12
 #define FASTRPC_PHYS(p)	((p) & 0xffffffff)
 #define FASTRPC_CTX_MAX (256)
 #define FASTRPC_INIT_HANDLE	1
@@ -104,6 +107,27 @@
 
 #define miscdev_to_fdevice(d) container_of(d, struct fastrpc_device, miscdev)
 
+#define PERF_END ((void)0)
+
+#define PERF(enb, cnt, ff) \
+	{\
+		struct timespec64 startT = {0};\
+		uint64_t *counter = cnt;\
+		if (enb && counter) {\
+			ktime_get_real_ts64(&startT);\
+		} \
+		ff ;\
+		if (enb && counter) {\
+			*counter += getnstimediff(&startT);\
+		} \
+	}
+
+#define GET_COUNTER(perf_ptr, offset)  \
+	(perf_ptr != NULL ?\
+		(((offset >= 0) && (offset < PERF_KEY_MAX)) ?\
+			(uint64_t *)(perf_ptr + offset)\
+				: (uint64_t *)NULL) : (uint64_t *)NULL)
+
 static const char *domains[FASTRPC_DEV_MAX] = { "adsp", "mdsp",
 						"sdsp", "cdsp"};
 struct fastrpc_phy_page {
@@ -227,6 +251,19 @@ struct fastrpc_map {
 	struct kref refcount;
 };
 
+struct fastrpc_perf {
+	u64 count;
+	u64 flush;
+	u64 map;
+	u64 copy;
+	u64 link;
+	u64 getargs;
+	u64 putargs;
+	u64 invargs;
+	u64 invoke;
+	u64 tid;
+};
+
 struct fastrpc_invoke_ctx {
 	int nscalars;
 	int nbufs;
@@ -235,6 +272,8 @@ struct fastrpc_invoke_ctx {
 	int tgid;
 	u32 sc;
 	u32 *crc;
+	u64 *perf_kernel;
+	u64 *perf_dsp;
 	u64 ctxid;
 	u64 msg_sz;
 	struct kref refcount;
@@ -249,6 +288,7 @@ struct fastrpc_invoke_ctx {
 	struct fastrpc_invoke_args *args;
 	struct fastrpc_buf_overlap *olaps;
 	struct fastrpc_channel_ctx *cctx;
+	struct fastrpc_perf *perf;
 };
 
 struct fastrpc_session_ctx {
@@ -298,6 +338,7 @@ struct fastrpc_user {
 	struct fastrpc_session_ctx *sctx;
 	struct fastrpc_buf *init_mem;
 
+	u32 profile;
 	int tgid;
 	int pd;
 	bool is_secure_dev;
@@ -307,6 +348,17 @@ struct fastrpc_user {
 	struct mutex mutex;
 };
 
+static inline int64_t getnstimediff(struct timespec64 *start)
+{
+	int64_t ns;
+	struct timespec64 ts, b;
+
+	ktime_get_real_ts64(&ts);
+	b = timespec64_sub(ts, *start);
+	ns = timespec64_to_ns(&b);
+	return ns;
+}
+
 static void fastrpc_free_map(struct kref *ref)
 {
 	struct fastrpc_map *map;
@@ -492,6 +544,9 @@ static void fastrpc_context_free(struct kref *ref)
 	if (ctx->buf)
 		fastrpc_buf_free(ctx->buf);
 
+	if (ctx->fl->profile)
+		kfree(ctx->perf);
+
 	spin_lock_irqsave(&cctx->lock, flags);
 	idr_remove(&cctx->ctx_idr, ctx->ctxid >> 4);
 	spin_unlock_irqrestore(&cctx->lock, flags);
@@ -611,6 +666,14 @@ static struct fastrpc_invoke_ctx *fastrpc_context_alloc(
 	fastrpc_channel_ctx_get(cctx);
 
 	ctx->crc = (u32 *)(uintptr_t)invoke->crc;
+	ctx->perf_dsp = (u64 *)(uintptr_t)invoke->perf_dsp;
+	ctx->perf_kernel = (u64 *)(uintptr_t)invoke->perf_kernel;
+	if (ctx->fl->profile) {
+		ctx->perf = kzalloc(sizeof(*(ctx->perf)), GFP_KERNEL);
+		if (!ctx->perf)
+			return ERR_PTR(-ENOMEM);
+		ctx->perf->tid = ctx->fl->tgid;
+	}
 	ctx->sc = sc;
 	ctx->retval = -1;
 	ctx->pid = current->pid;
@@ -872,7 +935,8 @@ static int fastrpc_get_meta_size(struct fastrpc_invoke_ctx *ctx)
 		sizeof(struct fastrpc_invoke_buf) +
 		sizeof(struct fastrpc_phy_page)) * ctx->nscalars +
 		sizeof(u64) * FASTRPC_MAX_FDLIST +
-		sizeof(u32) * FASTRPC_MAX_CRCLIST;
+		sizeof(u32) * FASTRPC_MAX_CRCLIST +
+		sizeof(u32) + sizeof(u64) * FASTRPC_DSP_PERF_LIST;
 
 	return size;
 }
@@ -939,16 +1003,22 @@ static int fastrpc_get_args(u32 kernel, struct fastrpc_invoke_ctx *ctx)
 	int inbufs, i, oix, err = 0;
 	u64 len, rlen, pkt_size;
 	u64 pg_start, pg_end;
+	u64 *perf_counter = NULL;
 	uintptr_t args;
 	int metalen;
 
+	if (ctx->fl->profile)
+		perf_counter = (u64 *)ctx->perf + PERF_COUNT;
+
 	inbufs = REMOTE_SCALARS_INBUFS(ctx->sc);
 	metalen = fastrpc_get_meta_size(ctx);
 	pkt_size = fastrpc_get_payload_size(ctx, metalen);
 
+	PERF(ctx->fl->profile, GET_COUNTER(perf_counter, PERF_MAP),
 	err = fastrpc_create_maps(ctx);
 	if (err)
 		return err;
+	PERF_END);
 
 	ctx->msg_sz = pkt_size;
 
@@ -980,6 +1050,7 @@ static int fastrpc_get_args(u32 kernel, struct fastrpc_invoke_ctx *ctx)
 		if (ctx->maps[i]) {
 			struct vm_area_struct *vma = NULL;
 
+			PERF(ctx->fl->profile, GET_COUNTER(perf_counter, PERF_MAP),
 			rpra[i].buf.pv = (u64) ctx->args[i].ptr;
 			pages[i].addr = ctx->maps[i]->phys;
 
@@ -994,9 +1065,9 @@ static int fastrpc_get_args(u32 kernel, struct fastrpc_invoke_ctx *ctx)
 			pg_end = ((ctx->args[i].ptr + len - 1) & PAGE_MASK) >>
 				  PAGE_SHIFT;
 			pages[i].size = (pg_end - pg_start + 1) * PAGE_SIZE;
-
+			PERF_END);
 		} else {
-
+			PERF(ctx->fl->profile, GET_COUNTER(perf_counter, PERF_COPY),
 			if (ctx->olaps[oix].offset == 0) {
 				rlen -= ALIGN(args, FASTRPC_ALIGN) - args;
 				args = ALIGN(args, FASTRPC_ALIGN);
@@ -1018,12 +1089,14 @@ static int fastrpc_get_args(u32 kernel, struct fastrpc_invoke_ctx *ctx)
 			pages[i].size = (pg_end - pg_start + 1) * PAGE_SIZE;
 			args = args + mlen;
 			rlen -= mlen;
+			PERF_END);
 		}
 
 		if (i < inbufs && !ctx->maps[i]) {
 			void *dst = (void *)(uintptr_t)rpra[i].buf.pv;
 			void *src = (void *)(uintptr_t)ctx->args[i].ptr;
 
+			PERF(ctx->fl->profile, GET_COUNTER(perf_counter, PERF_COPY),
 			if (!kernel) {
 				if (copy_from_user(dst, (void __user *)src,
 						   len)) {
@@ -1033,6 +1106,7 @@ static int fastrpc_get_args(u32 kernel, struct fastrpc_invoke_ctx *ctx)
 			} else {
 				memcpy(dst, src, len);
 			}
+			PERF_END);
 		}
 	}
 
@@ -1063,9 +1137,9 @@ static int fastrpc_put_args(struct fastrpc_invoke_ctx *ctx,
 	struct fastrpc_map *mmap = NULL;
 	struct fastrpc_invoke_buf *list;
 	struct fastrpc_phy_page *pages;
-	u64 *fdlist;
-	u32 *crclist;
-	int i, inbufs, outbufs, handles;
+	u64 *fdlist, *perf_dsp_list;
+	u32 *crclist, *poll;
+	int i, inbufs, outbufs, handles, perferr;
 
 	inbufs = REMOTE_SCALARS_INBUFS(ctx->sc);
 	outbufs = REMOTE_SCALARS_OUTBUFS(ctx->sc);
@@ -1074,6 +1148,8 @@ static int fastrpc_put_args(struct fastrpc_invoke_ctx *ctx,
 	pages = fastrpc_phy_page_start(list, ctx->nscalars);
 	fdlist = (u64 *)(pages + inbufs + outbufs + handles);
 	crclist = (u32 *)(fdlist + FASTRPC_MAX_FDLIST);
+	poll = (u32 *)(crclist + FASTRPC_MAX_CRCLIST);
+	perf_dsp_list = (u64 *)(poll + 1);
 
 	for (i = inbufs; i < ctx->nbufs; ++i) {
 		if (!ctx->maps[i]) {
@@ -1098,8 +1174,16 @@ static int fastrpc_put_args(struct fastrpc_invoke_ctx *ctx,
 	}
 
 	if (ctx->crc && crclist && rpra) {
-		if (copy_to_user((void __user *)ctx->crc, crclist, FASTRPC_MAX_CRCLIST * sizeof(u32)))
+		if (copy_to_user((void __user *)ctx->crc, crclist,
+					FASTRPC_MAX_CRCLIST * sizeof(u32))) {
 			return -EFAULT;
+		}
+	}
+	if (ctx->perf_dsp && perf_dsp_list) {
+		perferr = copy_to_user((void __user *)ctx->perf_dsp,
+				perf_dsp_list, FASTRPC_DSP_PERF_LIST * sizeof(u64));
+		if (perferr)
+			dev_info(fl->sctx->dev, "Warning: failed to copy perf data %d\n", perferr);
 	}
 	return 0;
 }
@@ -1136,6 +1220,20 @@ static int fastrpc_invoke_send(struct fastrpc_session_ctx *sctx,
 
 }
 
+static void fastrpc_update_invoke_count(u32 handle, u64 *perf_counter,
+					struct timespec64 *invoket)
+{
+	u64 *invcount, *count;
+
+	invcount = GET_COUNTER(perf_counter, PERF_INVOKE);
+	if (invcount)
+		*invcount += getnstimediff(invoket);
+
+	count = GET_COUNTER(perf_counter, PERF_COUNT);
+	if (count)
+		*count += 1;
+}
+
 static int fastrpc_internal_invoke(struct fastrpc_user *fl,  u32 kernel,
 				   struct fastrpc_enhanced_invoke *invoke)
 {
@@ -1143,7 +1241,12 @@ static int fastrpc_internal_invoke(struct fastrpc_user *fl,  u32 kernel,
 	struct fastrpc_buf *buf, *b;
 	struct fastrpc_invoke *inv = &invoke->inv;
 	u32 handle, sc;
-	int err = 0;
+	u64 *perf_counter = NULL;
+	int err = 0, perferr = 0;
+	struct timespec64 invoket = {0};
+
+	if (fl->profile)
+		ktime_get_real_ts64(&invoket);
 
 	if (!fl->sctx)
 		return -EINVAL;
@@ -1162,18 +1265,24 @@ static int fastrpc_internal_invoke(struct fastrpc_user *fl,  u32 kernel,
 	if (IS_ERR(ctx))
 		return PTR_ERR(ctx);
 
+	if (fl->profile)
+		perf_counter = (u64 *)ctx->perf + PERF_COUNT;
+	PERF(fl->profile, GET_COUNTER(perf_counter, PERF_GETARGS),
 	if (ctx->nscalars) {
 		err = fastrpc_get_args(kernel, ctx);
 		if (err)
 			goto bail;
 	}
+	PERF_END);
 
 	/* make sure that all CPU memory writes are seen by DSP */
 	dma_wmb();
+	PERF(fl->profile, GET_COUNTER(perf_counter, PERF_LINK),
 	/* Send invoke buffer to remote dsp */
 	err = fastrpc_invoke_send(fl->sctx, ctx, kernel, handle);
 	if (err)
 		goto bail;
+	PERF_END);
 
 	if (kernel) {
 		if (!wait_for_completion_timeout(&ctx->work, 10 * HZ))
@@ -1193,10 +1302,12 @@ static int fastrpc_internal_invoke(struct fastrpc_user *fl,  u32 kernel,
 	if (ctx->nscalars) {
 		/* make sure that all memory writes by DSP are seen by CPU */
 		dma_rmb();
+		PERF(fl->profile, GET_COUNTER(perf_counter, PERF_PUTARGS),
 		/* populate all the output buffers with results */
 		err = fastrpc_put_args(ctx, kernel);
 		if (err)
 			goto bail;
+		PERF_END);
 	}
 
 bail:
@@ -1213,6 +1324,15 @@ static int fastrpc_internal_invoke(struct fastrpc_user *fl,  u32 kernel,
 			list_del(&buf->node);
 			list_add_tail(&buf->node, &fl->cctx->invoke_interrupted_mmaps);
 		}
+	} else if (ctx) {
+		if (fl->profile && !err)
+			fastrpc_update_invoke_count(handle, perf_counter, &invoket);
+		if (fl->profile && ctx->perf && ctx->perf_kernel) {
+			perferr = copy_to_user((void __user *)ctx->perf_kernel,
+						ctx->perf, FASTRPC_KERNEL_PERF_LIST * sizeof(u64));
+			if (perferr)
+				dev_info(fl->sctx->dev, "Warning: failed to copy perf data %d\n", perferr);
+		}
 	}
 
 	if (err)
@@ -1706,6 +1826,7 @@ static int fastrpc_multimode_invoke(struct fastrpc_user *fl, char __user *argp)
 	struct fastrpc_invoke_args *args = NULL;
 	struct fastrpc_ioctl_multimode_invoke invoke;
 	u32 nscalars;
+	u64 *perf_kernel;
 	int err;
 
 	if (copy_from_user(&invoke, argp, sizeof(invoke)))
@@ -1729,6 +1850,9 @@ static int fastrpc_multimode_invoke(struct fastrpc_user *fl, char __user *argp)
 				return -EFAULT;
 			}
 		}
+		perf_kernel = (u64 *)(uintptr_t)einv.perf_kernel;
+		if (perf_kernel)
+			fl->profile = true;
 		einv.inv.args = (__u64)args;
 		err = fastrpc_internal_invoke(fl, false, &einv);
 		kfree(args);
diff --git a/include/uapi/misc/fastrpc.h b/include/uapi/misc/fastrpc.h
index 36694fb..c64cf6a 100644
--- a/include/uapi/misc/fastrpc.h
+++ b/include/uapi/misc/fastrpc.h
@@ -165,4 +165,18 @@ struct fastrpc_ioctl_capability {
 	__u32 reserved[4];
 };
 
+enum fastrpc_perfkeys {
+	PERF_COUNT = 0,
+	PERF_RESERVED1 = 1,
+	PERF_MAP = 2,
+	PERF_COPY = 3,
+	PERF_LINK = 4,
+	PERF_GETARGS = 5,
+	PERF_PUTARGS = 6,
+	PERF_RESERVED2 = 7,
+	PERF_INVOKE = 8,
+	PERF_RESERVED3 = 9,
+	PERF_KEY_MAX = 10,
+};
+
 #endif /* __QCOM_FASTRPC_H__ */
-- 
2.7.4

[PATCH v2 3/5] misc: fastrpc: Add support to save and restore interrupted

[Ekansh Gupta]

For any remote call, driver sends a message to DSP using RPMSG
framework. After message is sent, there is a wait on a completion
object at driver which is completed when DSP response is received.

There is a possibility that a signal is received while waiting
causing the wait function to return -ERESTARTSYS. In this case
the context should be saved and it should get restored for the
next invocation for the thread.

Adding changes to support saving and restoring of interrupted
fastrpc contexts.

Signed-off-by: Ekansh Gupta <quic_ekangupt@quicinc.com>
---
 drivers/misc/fastrpc.c | 76 +++++++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 72 insertions(+), 4 deletions(-)

diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c
index f8c1c381..1b26718 100644
--- a/drivers/misc/fastrpc.c
+++ b/drivers/misc/fastrpc.c
@@ -332,6 +332,7 @@ struct fastrpc_user {
 	struct list_head user;
 	struct list_head maps;
 	struct list_head pending;
+	struct list_head interrupted;
 	struct list_head mmaps;
 
 	struct fastrpc_channel_ctx *cctx;
@@ -711,6 +712,40 @@ static struct fastrpc_invoke_ctx *fastrpc_context_alloc(
 	return ERR_PTR(ret);
 }
 
+static struct fastrpc_invoke_ctx *fastrpc_context_restore_interrupted(
+			struct fastrpc_user *fl, struct fastrpc_invoke *inv)
+{
+	struct fastrpc_invoke_ctx *ctx = NULL, *ictx = NULL, *n;
+
+	spin_lock(&fl->lock);
+	list_for_each_entry_safe(ictx, n, &fl->interrupted, node) {
+		if (ictx->pid == current->pid) {
+			if (inv->sc != ictx->sc || ictx->fl != fl) {
+				dev_err(ictx->fl->sctx->dev,
+					"interrupted sc (0x%x) or fl (%pK) does not match with invoke sc (0x%x) or fl (%pK)\n",
+					ictx->sc, ictx->fl, inv->sc, fl);
+				spin_unlock(&fl->lock);
+				return ERR_PTR(-EINVAL);
+			}
+			ctx = ictx;
+			list_del(&ctx->node);
+			list_add_tail(&ctx->node, &fl->pending);
+			break;
+		}
+	}
+	spin_unlock(&fl->lock);
+	return ctx;
+}
+
+static void fastrpc_context_save_interrupted(
+			struct fastrpc_invoke_ctx *ctx)
+{
+	spin_lock(&ctx->fl->lock);
+	list_del(&ctx->node);
+	list_add_tail(&ctx->node, &ctx->fl->interrupted);
+	spin_unlock(&ctx->fl->lock);
+}
+
 static struct sg_table *
 fastrpc_map_dma_buf(struct dma_buf_attachment *attachment,
 		    enum dma_data_direction dir)
@@ -1261,6 +1296,14 @@ static int fastrpc_internal_invoke(struct fastrpc_user *fl,  u32 kernel,
 		return -EPERM;
 	}
 
+	if (!kernel) {
+		ctx = fastrpc_context_restore_interrupted(fl, inv);
+		if (IS_ERR(ctx))
+			return PTR_ERR(ctx);
+		if (ctx)
+			goto wait;
+	}
+
 	ctx = fastrpc_context_alloc(fl, kernel, sc, invoke);
 	if (IS_ERR(ctx))
 		return PTR_ERR(ctx);
@@ -1284,6 +1327,7 @@ static int fastrpc_internal_invoke(struct fastrpc_user *fl,  u32 kernel,
 		goto bail;
 	PERF_END);
 
+wait:
 	if (kernel) {
 		if (!wait_for_completion_timeout(&ctx->work, 10 * HZ))
 			err = -ETIMEDOUT;
@@ -1320,6 +1364,9 @@ static int fastrpc_internal_invoke(struct fastrpc_user *fl,  u32 kernel,
 	}
 
 	if (err == -ERESTARTSYS) {
+		if (ctx)
+			fastrpc_context_save_interrupted(ctx);
+
 		list_for_each_entry_safe(buf, b, &fl->mmaps, node) {
 			list_del(&buf->node);
 			list_add_tail(&buf->node, &fl->cctx->invoke_interrupted_mmaps);
@@ -1620,6 +1667,25 @@ static void fastrpc_session_free(struct fastrpc_channel_ctx *cctx,
 	spin_unlock_irqrestore(&cctx->lock, flags);
 }
 
+static void fastrpc_context_list_free(struct fastrpc_user *fl)
+{
+	struct fastrpc_invoke_ctx *ctx, *n;
+
+	list_for_each_entry_safe(ctx, n, &fl->interrupted, node) {
+		spin_lock(&fl->lock);
+		list_del(&ctx->node);
+		spin_unlock(&fl->lock);
+		fastrpc_context_put(ctx);
+	}
+
+	list_for_each_entry_safe(ctx, n, &fl->pending, node) {
+		spin_lock(&fl->lock);
+		list_del(&ctx->node);
+		spin_unlock(&fl->lock);
+		fastrpc_context_put(ctx);
+	}
+}
+
 static int fastrpc_release_current_dsp_process(struct fastrpc_user *fl)
 {
 	struct fastrpc_invoke_args args[1];
@@ -1656,10 +1722,7 @@ static int fastrpc_device_release(struct inode *inode, struct file *file)
 	if (fl->init_mem)
 		fastrpc_buf_free(fl->init_mem);
 
-	list_for_each_entry_safe(ctx, n, &fl->pending, node) {
-		list_del(&ctx->node);
-		fastrpc_context_put(ctx);
-	}
+	fastrpc_context_list_free(fl);
 
 	list_for_each_entry_safe(map, m, &fl->maps, node)
 		fastrpc_map_put(map);
@@ -1700,6 +1763,7 @@ static int fastrpc_device_open(struct inode *inode, struct file *filp)
 	spin_lock_init(&fl->lock);
 	mutex_init(&fl->mutex);
 	INIT_LIST_HEAD(&fl->pending);
+	INIT_LIST_HEAD(&fl->interrupted);
 	INIT_LIST_HEAD(&fl->maps);
 	INIT_LIST_HEAD(&fl->mmaps);
 	INIT_LIST_HEAD(&fl->user);
@@ -2555,6 +2619,10 @@ static void fastrpc_notify_users(struct fastrpc_user *user)
 		ctx->retval = -EPIPE;
 		complete(&ctx->work);
 	}
+	list_for_each_entry(ctx, &user->interrupted, node) {
+		ctx->retval = -EPIPE;
+		complete(&ctx->work);
+	}
 	spin_unlock(&user->lock);
 }
 
-- 
2.7.4

[PATCH v2 4/5] misc: fastrpc: Add support to save and restore interrupted

[Ekansh Gupta]

For any remote call, driver sends a message to DSP using RPMSG
framework. After message is sent, there is a wait on a completion
object at driver which is completed when DSP response is received.

There is a possibility that a signal is received while waiting
causing the wait function to return -ERESTARTSYS. In this case
the context should be saved and it should get restored for the
next invocation for the thread.

Adding changes to support saving and restoring of interrupted
fastrpc contexts.

Signed-off-by: Ekansh Gupta <quic_ekangupt@quicinc.com>
---
Changes in v2:
  - Fixed compile time warnings

 drivers/misc/fastrpc.c | 23 +++++++++++------------
 1 file changed, 11 insertions(+), 12 deletions(-)

diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c
index 1b26718..1c625571 100644
--- a/drivers/misc/fastrpc.c
+++ b/drivers/misc/fastrpc.c
@@ -1488,7 +1488,7 @@ static int fastrpc_init_create_static_process(struct fastrpc_user *fl,
 
 	ioctl.inv.handle = FASTRPC_INIT_HANDLE;
 	ioctl.inv.sc = FASTRPC_SCALARS(FASTRPC_RMID_INIT_CREATE_STATIC, 3, 0);
-	ioctl.inv.args = (__u64)args;
+	ioctl.inv.args = (u64)args;
 
 	err = fastrpc_internal_invoke(fl, true, &ioctl);
 	if (err)
@@ -1616,7 +1616,7 @@ static int fastrpc_init_create_process(struct fastrpc_user *fl,
 	ioctl.inv.sc = FASTRPC_SCALARS(FASTRPC_RMID_INIT_CREATE, 4, 0);
 	if (init.attrs)
 		ioctl.inv.sc = FASTRPC_SCALARS(FASTRPC_RMID_INIT_CREATE_ATTR, 4, 0);
-	ioctl.inv.args = (__u64)args;
+	ioctl.inv.args = (u64)args;
 
 	err = fastrpc_internal_invoke(fl, true, &ioctl);
 	if (err)
@@ -1699,7 +1699,7 @@ static int fastrpc_release_current_dsp_process(struct fastrpc_user *fl)
 
 	ioctl.inv.handle = FASTRPC_INIT_HANDLE;
 	ioctl.inv.sc = FASTRPC_SCALARS(FASTRPC_RMID_INIT_RELEASE, 1, 0);
-	ioctl.inv.args = (__u64)args;
+	ioctl.inv.args = (u64)args;
 
 	return fastrpc_internal_invoke(fl, true, &ioctl);
 }
@@ -1708,7 +1708,6 @@ static int fastrpc_device_release(struct inode *inode, struct file *file)
 {
 	struct fastrpc_user *fl = (struct fastrpc_user *)file->private_data;
 	struct fastrpc_channel_ctx *cctx = fl->cctx;
-	struct fastrpc_invoke_ctx *ctx, *n;
 	struct fastrpc_map *map, *m;
 	struct fastrpc_buf *buf, *b;
 	unsigned long flags;
@@ -1845,7 +1844,7 @@ static int fastrpc_init_attach(struct fastrpc_user *fl, int pd)
 
 	ioctl.inv.handle = FASTRPC_INIT_HANDLE;
 	ioctl.inv.sc = FASTRPC_SCALARS(FASTRPC_RMID_INIT_ATTACH, 1, 0);
-	ioctl.inv.args = (__u64)args;
+	ioctl.inv.args = (u64)args;
 
 	return fastrpc_internal_invoke(fl, true, &ioctl);
 }
@@ -1876,7 +1875,7 @@ static int fastrpc_invoke(struct fastrpc_user *fl, char __user *argp)
 	}
 
 	ioctl.inv = inv;
-	ioctl.inv.args = (__u64)args;
+	ioctl.inv.args = (u64)args;
 
 	err = fastrpc_internal_invoke(fl, false, &ioctl);
 	kfree(args);
@@ -1917,7 +1916,7 @@ static int fastrpc_multimode_invoke(struct fastrpc_user *fl, char __user *argp)
 		perf_kernel = (u64 *)(uintptr_t)einv.perf_kernel;
 		if (perf_kernel)
 			fl->profile = true;
-		einv.inv.args = (__u64)args;
+		einv.inv.args = (u64)args;
 		err = fastrpc_internal_invoke(fl, false, &einv);
 		kfree(args);
 		break;
@@ -1947,7 +1946,7 @@ static int fastrpc_get_info_from_dsp(struct fastrpc_user *fl, uint32_t *dsp_attr
 
 	ioctl.inv.handle = FASTRPC_DSP_UTILITIES_HANDLE;
 	ioctl.inv.sc = FASTRPC_SCALARS(0, 1, 1);
-	ioctl.inv.args = (__u64)args;
+	ioctl.inv.args = (u64)args;
 
 	return fastrpc_internal_invoke(fl, true, &ioctl);
 }
@@ -2050,7 +2049,7 @@ static int fastrpc_req_munmap_impl(struct fastrpc_user *fl, struct fastrpc_buf *
 
 	ioctl.inv.handle = FASTRPC_INIT_HANDLE;
 	ioctl.inv.sc = FASTRPC_SCALARS(FASTRPC_RMID_INIT_MUNMAP, 1, 0);
-	ioctl.inv.args = (__u64)args;
+	ioctl.inv.args = (u64)args;
 
 	err = fastrpc_internal_invoke(fl, true, &ioctl);
 	if (!err) {
@@ -2144,7 +2143,7 @@ static int fastrpc_req_mmap(struct fastrpc_user *fl, char __user *argp)
 
 	ioctl.inv.handle = FASTRPC_INIT_HANDLE;
 	ioctl.inv.sc = FASTRPC_SCALARS(FASTRPC_RMID_INIT_MMAP, 2, 1);
-	ioctl.inv.args = (__u64)args;
+	ioctl.inv.args = (u64)args;
 
 	err = fastrpc_internal_invoke(fl, true, &ioctl);
 	if (err) {
@@ -2229,7 +2228,7 @@ static int fastrpc_req_mem_unmap_impl(struct fastrpc_user *fl, struct fastrpc_me
 
 	ioctl.inv.handle = FASTRPC_INIT_HANDLE;
 	ioctl.inv.sc = FASTRPC_SCALARS(FASTRPC_RMID_INIT_MEM_UNMAP, 1, 0);
-	ioctl.inv.args = (__u64)args;
+	ioctl.inv.args = (u64)args;
 
 	err = fastrpc_internal_invoke(fl, true, &ioctl);
 	fastrpc_map_put(map);
@@ -2298,7 +2297,7 @@ static int fastrpc_req_mem_map(struct fastrpc_user *fl, char __user *argp)
 
 	ioctl.inv.handle = FASTRPC_INIT_HANDLE;
 	ioctl.inv.sc = FASTRPC_SCALARS(FASTRPC_RMID_INIT_MEM_MAP, 3, 1);
-	ioctl.inv.args = (__u64)args;
+	ioctl.inv.args = (u64)args;
 
 	err = fastrpc_internal_invoke(fl, true, &ioctl);
 	if (err) {
-- 
2.7.4

[PATCH v2 5/5] misc: fastrpc: Add support to allocate shared context bank

[Ekansh Gupta]

Context banks could be set as a shared one using a DT propery
"qcom,nsessions". The property takes the number of session to
be created of the context bank. This change provides a control
mechanism for user to use shared context banks for light weight
processes. The session is set as shared while its creation and if
a user requests for shared context bank, the same will be allocated
during process initialization.

Signed-off-by: Ekansh Gupta <quic_ekangupt@quicinc.com>
---
Changes in v2:
  - Fixed missing definition
  - Fixes compile time issue

 drivers/misc/fastrpc.c      | 122 ++++++++++++++++++++++++++++++--------------
 include/uapi/misc/fastrpc.h |  12 +++++
 2 files changed, 95 insertions(+), 39 deletions(-)

diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c
index 1c625571..6deef32 100644
--- a/drivers/misc/fastrpc.c
+++ b/drivers/misc/fastrpc.c
@@ -296,6 +296,7 @@ struct fastrpc_session_ctx {
 	int sid;
 	bool used;
 	bool valid;
+	bool sharedcb;
 };
 
 struct fastrpc_channel_ctx {
@@ -343,12 +344,22 @@ struct fastrpc_user {
 	int tgid;
 	int pd;
 	bool is_secure_dev;
+	bool sharedcb;
 	/* Lock for lists */
 	spinlock_t lock;
 	/* lock for allocations */
 	struct mutex mutex;
 };
 
+struct fastrpc_ctrl_smmu {
+	u32 sharedcb;	/* Set to SMMU share context bank */
+};
+
+struct fastrpc_internal_control {
+	u32 req;
+	struct fastrpc_ctrl_smmu smmu;
+};
+
 static inline int64_t getnstimediff(struct timespec64 *start)
 {
 	int64_t ns;
@@ -850,6 +861,37 @@ static const struct dma_buf_ops fastrpc_dma_buf_ops = {
 	.release = fastrpc_release,
 };
 
+static struct fastrpc_session_ctx *fastrpc_session_alloc(
+					struct fastrpc_channel_ctx *cctx, bool sharedcb)
+{
+	struct fastrpc_session_ctx *session = NULL;
+	unsigned long flags;
+	int i;
+
+	spin_lock_irqsave(&cctx->lock, flags);
+	for (i = 0; i < cctx->sesscount; i++) {
+		if (!cctx->session[i].used && cctx->session[i].valid &&
+			cctx->session[i].sharedcb == sharedcb) {
+			cctx->session[i].used = true;
+			session = &cctx->session[i];
+			break;
+		}
+	}
+	spin_unlock_irqrestore(&cctx->lock, flags);
+
+	return session;
+}
+
+static void fastrpc_session_free(struct fastrpc_channel_ctx *cctx,
+				 struct fastrpc_session_ctx *session)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&cctx->lock, flags);
+	session->used = false;
+	spin_unlock_irqrestore(&cctx->lock, flags);
+}
+
 static int fastrpc_map_create(struct fastrpc_user *fl, int fd,
 			      u64 len, u32 attr, struct fastrpc_map **ppmap)
 {
@@ -1446,6 +1488,10 @@ static int fastrpc_init_create_static_process(struct fastrpc_user *fl,
 		goto err_name;
 	}
 
+	fl->sctx = fastrpc_session_alloc(fl->cctx, fl->sharedcb);
+	if (!fl->sctx)
+		return -EBUSY;
+
 	if (!fl->cctx->remote_heap) {
 		err = fastrpc_remote_heap_alloc(fl, fl->sctx->dev, init.memlen,
 						&fl->cctx->remote_heap);
@@ -1563,6 +1609,10 @@ static int fastrpc_init_create_process(struct fastrpc_user *fl,
 		goto err;
 	}
 
+	fl->sctx = fastrpc_session_alloc(fl->cctx, fl->sharedcb);
+	if (!fl->sctx)
+		return -EBUSY;
+
 	inbuf.pgid = fl->tgid;
 	inbuf.namelen = strlen(current->comm) + 1;
 	inbuf.filelen = init.filelen;
@@ -1637,36 +1687,6 @@ static int fastrpc_init_create_process(struct fastrpc_user *fl,
 	return err;
 }
 
-static struct fastrpc_session_ctx *fastrpc_session_alloc(
-					struct fastrpc_channel_ctx *cctx)
-{
-	struct fastrpc_session_ctx *session = NULL;
-	unsigned long flags;
-	int i;
-
-	spin_lock_irqsave(&cctx->lock, flags);
-	for (i = 0; i < cctx->sesscount; i++) {
-		if (!cctx->session[i].used && cctx->session[i].valid) {
-			cctx->session[i].used = true;
-			session = &cctx->session[i];
-			break;
-		}
-	}
-	spin_unlock_irqrestore(&cctx->lock, flags);
-
-	return session;
-}
-
-static void fastrpc_session_free(struct fastrpc_channel_ctx *cctx,
-				 struct fastrpc_session_ctx *session)
-{
-	unsigned long flags;
-
-	spin_lock_irqsave(&cctx->lock, flags);
-	session->used = false;
-	spin_unlock_irqrestore(&cctx->lock, flags);
-}
-
 static void fastrpc_context_list_free(struct fastrpc_user *fl)
 {
 	struct fastrpc_invoke_ctx *ctx, *n;
@@ -1770,15 +1790,6 @@ static int fastrpc_device_open(struct inode *inode, struct file *filp)
 	fl->cctx = cctx;
 	fl->is_secure_dev = fdevice->secure;
 
-	fl->sctx = fastrpc_session_alloc(cctx);
-	if (!fl->sctx) {
-		dev_err(&cctx->rpdev->dev, "No session available\n");
-		mutex_destroy(&fl->mutex);
-		kfree(fl);
-
-		return -EBUSY;
-	}
-
 	spin_lock_irqsave(&cctx->lock, flags);
 	list_add_tail(&fl->user, &cctx->users);
 	spin_unlock_irqrestore(&cctx->lock, flags);
@@ -1837,6 +1848,10 @@ static int fastrpc_init_attach(struct fastrpc_user *fl, int pd)
 	struct fastrpc_enhanced_invoke ioctl;
 	int tgid = fl->tgid;
 
+	fl->sctx = fastrpc_session_alloc(fl->cctx, fl->sharedcb);
+	if (!fl->sctx)
+		return -EBUSY;
+
 	args[0].ptr = (u64)(uintptr_t) &tgid;
 	args[0].length = sizeof(tgid);
 	args[0].fd = -1;
@@ -1883,11 +1898,33 @@ static int fastrpc_invoke(struct fastrpc_user *fl, char __user *argp)
 	return err;
 }
 
+static int fastrpc_internal_control(struct fastrpc_user *fl,
+					struct fastrpc_internal_control *cp)
+{
+	int err = 0;
+
+	if (!fl)
+		return -EBADF;
+	if (!cp)
+		return -EINVAL;
+
+	switch (cp->req) {
+	case FASTRPC_CONTROL_SMMU:
+		fl->sharedcb = cp->smmu.sharedcb;
+		break;
+	default:
+		err = -EBADRQC;
+		break;
+	}
+	return err;
+}
+
 static int fastrpc_multimode_invoke(struct fastrpc_user *fl, char __user *argp)
 {
 	struct fastrpc_enhanced_invoke einv;
 	struct fastrpc_invoke_args *args = NULL;
 	struct fastrpc_ioctl_multimode_invoke invoke;
+	struct fastrpc_internal_control cp = {0};
 	u32 nscalars;
 	u64 *perf_kernel;
 	int err;
@@ -1920,6 +1957,12 @@ static int fastrpc_multimode_invoke(struct fastrpc_user *fl, char __user *argp)
 		err = fastrpc_internal_invoke(fl, false, &einv);
 		kfree(args);
 		break;
+	case FASTRPC_INVOKE_CONTROL:
+		if (copy_from_user(&cp, (void __user *)(uintptr_t)invoke.invparam, sizeof(cp)))
+			return  -EFAULT;
+
+		err = fastrpc_internal_control(fl, &cp);
+		break;
 	default:
 		err = -ENOTTY;
 		break;
@@ -2420,6 +2463,7 @@ static int fastrpc_cb_probe(struct platform_device *pdev)
 	if (sessions > 0) {
 		struct fastrpc_session_ctx *dup_sess;
 
+		sess->sharedcb = true;
 		for (i = 1; i < sessions; i++) {
 			if (cctx->sesscount >= FASTRPC_MAX_SESSIONS)
 				break;
diff --git a/include/uapi/misc/fastrpc.h b/include/uapi/misc/fastrpc.h
index c64cf6a..c9faecf 100644
--- a/include/uapi/misc/fastrpc.h
+++ b/include/uapi/misc/fastrpc.h
@@ -165,6 +165,18 @@ struct fastrpc_ioctl_capability {
 	__u32 reserved[4];
 };
 
+enum fastrpc_control_type {
+	FASTRPC_CONTROL_LATENCY		=	1,
+	FASTRPC_CONTROL_SMMU		=	2,
+	FASTRPC_CONTROL_KALLOC		=	3,
+	FASTRPC_CONTROL_WAKELOCK	=	4,
+	FASTRPC_CONTROL_PM		=	5,
+	FASTRPC_CONTROL_DSPPROCESS_CLEAN	=	6,
+	FASTRPC_CONTROL_RPC_POLL	=	7,
+	FASTRPC_CONTROL_ASYNC_WAKE	=	8,
+	FASTRPC_CONTROL_NOTIF_WAKE	=	9,
+};
+
 enum fastrpc_perfkeys {
 	PERF_COUNT = 0,
 	PERF_RESERVED1 = 1,
-- 
2.7.4