Skip to contentSkip to conversation

drivers/misc/sgi-gru: add return on error

PATCH In review Sudip Mukherjee 3 participants 15 messages Sep 3, 2015 – Sep 21, 2015 Char and misc drivers

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/5 drivers/misc/sgi-gru: add return on error

If the buffer is too small then return the error and in the process remove the variables which became unused.

---

v1: only removed variables.

 drivers/misc/sgi-gru/grukdump.c | 14 ++++++--------
 1 file changed, 6 insertions(+), 8 deletions(-)
drivers/misc/sgi-gru/grukdump.c +6 -8
index a3700a5..7e9aae5 100644
@@ -78,11 +78,10 @@ static int gru_dump_tfm(struct gru_state *gru,
 		void __user *ubuf, void __user *ubufend)
 {
 	struct gru_tlb_fault_map *tfm;
-	int i, ret, bytes;
+	int i;
 
-	bytes = GRU_NUM_TFM * GRU_CACHE_LINE_BYTES;
-	if (bytes > ubufend - ubuf)
-		ret = -EFBIG;
+	if (GRU_NUM_TFM * GRU_CACHE_LINE_BYTES > ubufend - ubuf)
+		return -EFBIG;
 
 	for (i = 0; i < GRU_NUM_TFM; i++) {
 		tfm = get_tfm(gru->gs_gru_base_vaddr, i);
@@ -99,11 +98,10 @@ static int gru_dump_tgh(struct gru_state *gru,
 		void __user *ubuf, void __user *ubufend)
 {
 	struct gru_tlb_global_handle *tgh;
-	int i, ret, bytes;
+	int i;
 
-	bytes = GRU_NUM_TGH * GRU_CACHE_LINE_BYTES;
-	if (bytes > ubufend - ubuf)
-		ret = -EFBIG;
+	if (GRU_NUM_TGH * GRU_CACHE_LINE_BYTES > ubufend - ubuf)
+		return -EFBIG;
 
 	for (i = 0; i < GRU_NUM_TGH; i++) {
 		tgh = get_tgh(gru->gs_gru_base_vaddr, i);
Signature · 1 line
1.9.1
#1 lore raw

PATCH v2 2/5 drivers/misc/sgi-gru: make functions static

The functions gru_get_cb_exception_detail_str() and gru_abort() were only called locally from that file. We can make them static.

---

v2: no change, sent as a part of the series.

 drivers/misc/sgi-gru/grukservices.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/misc/sgi-gru/grukservices.c +3 -3
index 913de07..490b79a 100644
@@ -429,8 +429,8 @@ int gru_get_cb_exception_detail(void *cb,
 	return 0;
 }
 
-char *gru_get_cb_exception_detail_str(int ret, void *cb,
-				      char *buf, int size)
+static char *gru_get_cb_exception_detail_str(int ret, void *cb,
+					     char *buf, int size)
 {
 	struct gru_control_block_status *gen = (void *)cb;
 	struct control_block_extended_exc_detail excdet;
@@ -505,7 +505,7 @@ int gru_wait_proc(void *cb)
 	return ret;
 }
 
-void gru_abort(int ret, void *cb, char *str)
+static void gru_abort(int ret, void *cb, char *str)
 {
 	char buf[GRU_EXC_STR_SIZE];
 
Signature · 1 line
1.9.1
#2 lore raw

PATCH v2 3/5 drivers/misc/sgi-gru: remove always false condition

In reply to Sudip Mukherjee

The member gid in struct gru_dump_chiplet_state_req is unsigned int. So it can never be less than 0.

---

v2: no change, sent as a part of the series.

 drivers/misc/sgi-gru/grukdump.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/misc/sgi-gru/grukdump.c +1 -1
index 7e9aae5..313da31 100644
@@ -194,7 +194,7 @@ int gru_dump_chiplet_request(unsigned long arg)
 		return -EFAULT;
 
 	/* Currently, only dump by gid is implemented */
-	if (req.gid >= gru_max_gids || req.gid < 0)
+	if (req.gid >= gru_max_gids)
 		return -EINVAL;
 
 	gru = GID_TO_GRU(req.gid);
Signature · 1 line
1.9.1
#3 lore raw

PATCH v2 4/5 drivers/misc/sgi-gru: fix dereference of ERR_PTR

In reply to Sudip Mukherjee

gru_alloc_gts() can fail and it can return ERR_PTR(errvalue). We should not dereference it if it has returned error. And incase it has returned error then wait for some time and try again.

---

v2: on error retry after msleep(1).
v1: returned error.

 drivers/misc/sgi-gru/grukservices.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)
drivers/misc/sgi-gru/grukservices.c +8 -1
index 490b79a..d64f2a0 100644
@@ -160,7 +160,14 @@ static void gru_load_kernel_context(struct gru_blade_state *bs, int blade_id)
 	down_write(&bs->bs_kgts_sema);
 
 	if (!bs->bs_kgts) {
-		bs->bs_kgts = gru_alloc_gts(NULL, 0, 0, 0, 0, 0);
+		do {
+			bs->bs_kgts = gru_alloc_gts(NULL, 0, 0, 0, 0, 0);
+			if (IS_ERR(bs->bs_kgts)) {
+				msleep(1);
+				continue;
+			}
+			break;
+		} while (true);
 		bs->bs_kgts->ts_user_blade_id = blade_id;
 	}
 	kgts = bs->bs_kgts;
Signature · 1 line
1.9.1
#4 lore raw

Re: [PATCH v2 4/5] drivers/misc/sgi-gru: fix dereference of ERR_PTR

Quoted 20 lines · Sudip Mukherjee

gru_alloc_gts() can fail and it can return ERR_PTR(errvalue). We should not dereference it if it has returned error. And incase it has returned error then wait for some time and try again.

---

v2: on error retry after msleep(1).
v1: returned error.

 drivers/misc/sgi-gru/grukservices.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)
drivers/misc/sgi-gru/grukservices.c +0 -0
index 490b79a..d64f2a0 100644
@@ -160,7 +160,14 @@ static void gru_load_kernel_context(struct gru_blade_state *bs, int blade_id)
 	down_write(&bs->bs_kgts_sema);
 

drivers/misc/sgi-gru/grukservices.c static void gru_load_kernel_context(struct gru_blade_state *bs, int blade_id) · line 170

Quoted 10 lines
drivers/misc/sgi-gru/grukservices.c +8 -1
 	if (!bs->bs_kgts) {
-		bs->bs_kgts = gru_alloc_gts(NULL, 0, 0, 0, 0, 0);
+		do {
+			bs->bs_kgts = gru_alloc_gts(NULL, 0, 0, 0, 0, 0);
+			if (IS_ERR(bs->bs_kgts)) {
+				msleep(1);
+				continue;
+			}
+			break;
+		} while (true);

How about this way (for a little more compactness)?

+ do {
+ bs->bs_kgts = gru_alloc_gts(NULL, 0, 0, 0, 0, 0);
+ if (!IS_ERR(bs->bs_kgts))
+ break;
+ msleep(1)
+ } while (true);

Quoted 5 lines
drivers/misc/sgi-gru/grukservices.c +0 -0
 		bs->bs_kgts->ts_user_blade_id = blade_id;
 	}
 	kgts = bs->bs_kgts;
Signature · 1 line
1.9.1
#5 lore raw

PATCH v3 1/1 drivers/misc/sgi-gru: fix dereference of ERR_PTR

In reply to Sudip Mukherjee

gru_alloc_gts() can fail and it can return ERR_PTR(errvalue). We should not dereference it if it has returned error. And incase it has returned error then wait for some time and try again.

---

v3: compact code for if logic
v2: on error retry after msleep(1).
v1: returned error.

 drivers/misc/sgi-gru/grukservices.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/misc/sgi-gru/grukservices.c +6 -1
index 490b79a..7812e34 100644
@@ -160,7 +160,12 @@ static void gru_load_kernel_context(struct gru_blade_state *bs, int blade_id)
 	down_write(&bs->bs_kgts_sema);
 
 	if (!bs->bs_kgts) {
-		bs->bs_kgts = gru_alloc_gts(NULL, 0, 0, 0, 0, 0);
+		do {
+			bs->bs_kgts = gru_alloc_gts(NULL, 0, 0, 0, 0, 0);
+			if (!IS_ERR(bs->bs_kgts))
+				break;
+			msleep(1);
+		} while (true);
 		bs->bs_kgts->ts_user_blade_id = blade_id;
 	}
 	kgts = bs->bs_kgts;
Signature · 1 line
1.9.1
Acked-by on v3 1/1
#6 lore raw

Re: [PATCH v3] drivers/misc/sgi-gru: fix dereference of ERR_PTR

Quoted 35 lines · Sudip Mukherjee

gru_alloc_gts() can fail and it can return ERR_PTR(errvalue). We should not dereference it if it has returned error. And incase it has returned error then wait for some time and try again.

---

v3: compact code for if logic
v2: on error retry after msleep(1).
v1: returned error.

 drivers/misc/sgi-gru/grukservices.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/misc/sgi-gru/grukservices.c +6 -1
index 490b79a..7812e34 100644
@@ -160,7 +160,12 @@ static void gru_load_kernel_context(struct gru_blade_state *bs, int blade_id)
 	down_write(&bs->bs_kgts_sema);
 
 	if (!bs->bs_kgts) {
-		bs->bs_kgts = gru_alloc_gts(NULL, 0, 0, 0, 0, 0);
+		do {
+			bs->bs_kgts = gru_alloc_gts(NULL, 0, 0, 0, 0, 0);
+			if (!IS_ERR(bs->bs_kgts))
+				break;
+			msleep(1);
+		} while (true);
 		bs->bs_kgts->ts_user_blade_id = blade_id;
 	}
 	kgts = bs->bs_kgts;
Signature · 1 line
1.9.1
#7 lore raw

PATCH v2 5/5 drivers/misc/sgi-gru: remove unused variable

In reply to Sudip Mukherjee

dw was only assigned some value and was never reused.

---

v2: It was part of first patch in v1. But since one logical change in one patch so it had to be removed from first patch and became separate.

 drivers/misc/sgi-gru/grukservices.c | 2 --
 1 file changed, 2 deletions(-)
drivers/misc/sgi-gru/grukservices.c +0 -2
index d64f2a0..ba2d2e5 100644
@@ -1004,7 +1004,6 @@ static int quicktest1(unsigned long arg)
 {
 	struct gru_message_queue_desc mqd;
 	void *p, *mq;
-	unsigned long *dw;
 	int i, ret = -EIO;
 	char mes[GRU_CACHE_LINE_BYTES], *m;
 
@@ -1014,7 +1013,6 @@ static int quicktest1(unsigned long arg)
 		return -ENOMEM;
 	mq = ALIGNUP(p, 1024);
 	memset(mes, 0xee, sizeof(mes));
-	dw = mq;
 
 	gru_create_message_queue(&mqd, mq, 8 * GRU_CACHE_LINE_BYTES, 0, 0, 0);
 	for (i = 0; i < 6; i++) {
Signature · 1 line
1.9.1
Acked-by on v2 5/5
#8 lore raw

Re: [PATCH v2 5/5] drivers/misc/sgi-gru: remove unused variable

Quoted 33 lines · Sudip Mukherjee

dw was only assigned some value and was never reused.

---

v2: It was part of first patch in v1. But since one logical change in one patch so it had to be removed from first patch and became separate.

 drivers/misc/sgi-gru/grukservices.c | 2 --
 1 file changed, 2 deletions(-)
drivers/misc/sgi-gru/grukservices.c +0 -2
index d64f2a0..ba2d2e5 100644
@@ -1004,7 +1004,6 @@ static int quicktest1(unsigned long arg)
 {
 	struct gru_message_queue_desc mqd;
 	void *p, *mq;
-	unsigned long *dw;
 	int i, ret = -EIO;
 	char mes[GRU_CACHE_LINE_BYTES], *m;
 
@@ -1014,7 +1013,6 @@ static int quicktest1(unsigned long arg)
 		return -ENOMEM;
 	mq = ALIGNUP(p, 1024);
 	memset(mes, 0xee, sizeof(mes));
-	dw = mq;
 
 	gru_create_message_queue(&mqd, mq, 8 * GRU_CACHE_LINE_BYTES, 0, 0, 0);
 	for (i = 0; i < 6; i++) {
Signature · 1 line
1.9.1
Acked-by on v2 1/5
#9 lore raw
In reply to Sudip Mukherjee
Quoted 47 lines · Sudip Mukherjee

If the buffer is too small then return the error and in the process remove the variables which became unused.

---

v1: only removed variables.

 drivers/misc/sgi-gru/grukdump.c | 14 ++++++--------
 1 file changed, 6 insertions(+), 8 deletions(-)
drivers/misc/sgi-gru/grukdump.c +6 -8
index a3700a5..7e9aae5 100644
@@ -78,11 +78,10 @@ static int gru_dump_tfm(struct gru_state *gru,
 		void __user *ubuf, void __user *ubufend)
 {
 	struct gru_tlb_fault_map *tfm;
-	int i, ret, bytes;
+	int i;
 
-	bytes = GRU_NUM_TFM * GRU_CACHE_LINE_BYTES;
-	if (bytes > ubufend - ubuf)
-		ret = -EFBIG;
+	if (GRU_NUM_TFM * GRU_CACHE_LINE_BYTES > ubufend - ubuf)
+		return -EFBIG;
 
 	for (i = 0; i < GRU_NUM_TFM; i++) {
 		tfm = get_tfm(gru->gs_gru_base_vaddr, i);
@@ -99,11 +98,10 @@ static int gru_dump_tgh(struct gru_state *gru,
 		void __user *ubuf, void __user *ubufend)
 {
 	struct gru_tlb_global_handle *tgh;
-	int i, ret, bytes;
+	int i;
 
-	bytes = GRU_NUM_TGH * GRU_CACHE_LINE_BYTES;
-	if (bytes > ubufend - ubuf)
-		ret = -EFBIG;
+	if (GRU_NUM_TGH * GRU_CACHE_LINE_BYTES > ubufend - ubuf)
+		return -EFBIG;
 
 	for (i = 0; i < GRU_NUM_TGH; i++) {
 		tgh = get_tgh(gru->gs_gru_base_vaddr, i);
Signature · 1 line
1.9.1
#10 lore raw
Quoted 8 lines · Dimitri Sivanich
Quoted 5 lines · Sudip Mukherjee

If the buffer is too small then return the error and in the process remove the variables which became unused.

---

Hi Greg,
I know its too early to ping you but just wanted to know that this should go through your char-misc tree or through Andrew. Looking at all the previous patches, it all went through Andrew but being a char/misc I thought it will go through you.

regards
sudip

#11 lore raw
Quoted 7 lines · Sudip Mukherjee
Quoted 6 lines · Dimitri Sivanich
Quoted 3 lines · Sudip Mukherjee

If the buffer is too small then return the error and in the process remove the variables which became unused.

Quoted 7 lines
Quoted 2 lines
Quoted 2 lines

---

Hi Greg,
I know its too early to ping you but just wanted to know that this should go through your char-misc tree or through Andrew. Looking at all the previous patches, it all went through Andrew but being a char/misc I thought it will go through you.

I'll take it, thanks.

greg k-h

#12 lore raw
Quoted 10 lines · Greg Kroah-Hartman
Quoted 9 lines · Sudip Mukherjee
Quoted 8 lines · Dimitri Sivanich
Quoted 5 lines · Sudip Mukherjee

If the buffer is too small then return the error and in the process remove the variables which became unused.

---

Quoted 7 lines
Quoted 5 lines

Hi Greg,
I know its too early to ping you but just wanted to know that this should go through your char-misc tree or through Andrew. Looking at all the previous patches, it all went through Andrew but being a char/misc I thought it will go through you.

I'll take it, thanks.

Thanks. But just to let you know that getmaintainer.pl is not showing your and Arnd's name. But MAINTAINER shows drivers/misc/* for you and Arnd. And besides I noticed there are one or two patches still lying around with maintainer ACK but never picked up as they didnot have you or Andrew in the To or CC list.

regards
sudip

#13 lore raw
Quoted 17 lines · Sudip Mukherjee
Quoted 16 lines · Greg Kroah-Hartman
Quoted 14 lines · Sudip Mukherjee
Quoted 8 lines · Dimitri Sivanich
Quoted 5 lines · Sudip Mukherjee

If the buffer is too small then return the error and in the process remove the variables which became unused.

---

Hi Greg,
I know its too early to ping you but just wanted to know that this should go through your char-misc tree or through Andrew. Looking at all the previous patches, it all went through Andrew but being a char/misc I thought it will go through you.

Quoted 6 lines
Quoted 1 line

I'll take it, thanks.

Thanks. But just to let you know that getmaintainer.pl is not showing your and Arnd's name. But MAINTAINER shows drivers/misc/* for you and Arnd. And besides I noticed there are one or two patches still lying around with maintainer ACK but never picked up as they didnot have you or Andrew in the To or CC list.

That's odd, care to forward them on to me?

#14 lore raw
Quoted 18 lines · Greg Kroah-Hartman
Quoted 17 lines · Sudip Mukherjee
Quoted 16 lines · Greg Kroah-Hartman
Quoted 14 lines · Sudip Mukherjee
Quoted 8 lines · Dimitri Sivanich
Quoted 5 lines · Sudip Mukherjee

If the buffer is too small then return the error and in the process remove the variables which became unused.

---

Hi Greg,
I know its too early to ping you but just wanted to know that this should go through your char-misc tree or through Andrew. Looking at all the previous patches, it all went through Andrew but being a char/misc I thought it will go through you.

Quoted 8 lines
Quoted 6 lines
Quoted 1 line

I'll take it, thanks.

Thanks. But just to let you know that getmaintainer.pl is not showing your and Arnd's name. But MAINTAINER shows drivers/misc/* for you and Arnd. And besides I noticed there are one or two patches still lying around with maintainer ACK but never picked up as they didnot have you or Andrew in the To or CC list.

That's odd, care to forward them on to me?

OMG... There are quite a few. Even from 2014 and one of them is a bugfix for a bugzilla.
I am sending them all by the end of the day (IST).

regards
sudip