All of lore.kernel.org
 help / color / mirror / Atom feed
* [kvm-unit-tests PATCH] s390x: Fix uv_call() exception behavior
@ 2021-01-18 14:03 Janosch Frank
  2021-01-18 14:40 ` Thomas Huth
  2021-01-18 15:09 ` [kvm-unit-tests PATCH v2] " Janosch Frank
  0 siblings, 2 replies; 6+ messages in thread
From: Janosch Frank @ 2021-01-18 14:03 UTC (permalink / raw)
  To: kvm390 mailing list
  Cc: thuth, david, borntraeger, imbrenda, cohuck, linux-s390

On a program exception we usually skip the instruction that caused the
exception and continue. That won't work for UV calls since a "brc 3,0b"
will retry the instruction if the CC is > 1.

Signed-off-by: Janosch Frank <frankja@linux.ibm.com>
---

I know this isn't very pretty.
I'm open for suggestions.

---
 lib/s390x/asm/uv.h | 14 ++++++++------
 1 file changed, 8 insertions(+), 6 deletions(-)

diff --git a/lib/s390x/asm/uv.h b/lib/s390x/asm/uv.h
index 4c2fc48..252f1a3 100644
--- a/lib/s390x/asm/uv.h
+++ b/lib/s390x/asm/uv.h
@@ -53,21 +53,23 @@ struct uv_cb_share {
 static inline int uv_call(unsigned long r1, unsigned long r2)
 {
 	int cc;
+	struct lowcore *lc = 0x0;
 
 	/*
-	 * The brc instruction will take care of the cc 2/3 case where
-	 * we need to continue the execution because we were
-	 * interrupted. The inline assembly will only return on
-	 * success/error i.e. cc 0/1.
-	*/
+	 * CC 2 and 3 tell us to re-execute because the instruction
+	 * hasn't yet finished.
+	 */
+	lc->pgm_int_code = 0;
+retry:
 	asm volatile(
 		"0:	.insn rrf,0xB9A40000,%[r1],%[r2],0,0\n"
-		"		brc	3,0b\n"
 		"		ipm	%[cc]\n"
 		"		srl	%[cc],28\n"
 		: [cc] "=d" (cc)
 		: [r1] "a" (r1), [r2] "a" (r2)
 		: "memory", "cc");
+	if (!lc->pgm_int_code && cc > 1)
+		goto retry;
 	return cc;
 }
 
-- 
2.25.1

^ permalink raw reply related	[flat|nested] 6+ messages in thread

* Re: [kvm-unit-tests PATCH] s390x: Fix uv_call() exception behavior
  2021-01-18 14:03 [kvm-unit-tests PATCH] s390x: Fix uv_call() exception behavior Janosch Frank
@ 2021-01-18 14:40 ` Thomas Huth
  2021-01-18 14:46   ` Janosch Frank
  2021-01-18 15:09 ` [kvm-unit-tests PATCH v2] " Janosch Frank
  1 sibling, 1 reply; 6+ messages in thread
From: Thomas Huth @ 2021-01-18 14:40 UTC (permalink / raw)
  To: Janosch Frank; +Cc: david, borntraeger, imbrenda, cohuck, linux-s390

On 18/01/2021 15.03, Janosch Frank wrote:
> On a program exception we usually skip the instruction that caused the
> exception and continue. That won't work for UV calls since a "brc 3,0b"
> will retry the instruction if the CC is > 1.
> 
> Signed-off-by: Janosch Frank <frankja@linux.ibm.com>
> ---
> 
> I know this isn't very pretty.
> I'm open for suggestions.
> 
> ---
>   lib/s390x/asm/uv.h | 14 ++++++++------
>   1 file changed, 8 insertions(+), 6 deletions(-)
> 
> diff --git a/lib/s390x/asm/uv.h b/lib/s390x/asm/uv.h
> index 4c2fc48..252f1a3 100644
> --- a/lib/s390x/asm/uv.h
> +++ b/lib/s390x/asm/uv.h
> @@ -53,21 +53,23 @@ struct uv_cb_share {
>   static inline int uv_call(unsigned long r1, unsigned long r2)
>   {
>   	int cc;
> +	struct lowcore *lc = 0x0;
>   
>   	/*
> -	 * The brc instruction will take care of the cc 2/3 case where
> -	 * we need to continue the execution because we were
> -	 * interrupted. The inline assembly will only return on
> -	 * success/error i.e. cc 0/1.
> -	*/
> +	 * CC 2 and 3 tell us to re-execute because the instruction
> +	 * hasn't yet finished.
> +	 */
> +	lc->pgm_int_code = 0;
> +retry:
>   	asm volatile(
>   		"0:	.insn rrf,0xB9A40000,%[r1],%[r2],0,0\n"
> -		"		brc	3,0b\n"
>   		"		ipm	%[cc]\n"
>   		"		srl	%[cc],28\n"
>   		: [cc] "=d" (cc)
>   		: [r1] "a" (r1), [r2] "a" (r2)
>   		: "memory", "cc");
> +	if (!lc->pgm_int_code && cc > 1)
> +		goto retry;

Why not simply:

	do {
		asm volatile(...);
	} while (!lc->pgm_int_code && cc > 1)

?

   Thomas

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [kvm-unit-tests PATCH] s390x: Fix uv_call() exception behavior
  2021-01-18 14:40 ` Thomas Huth
@ 2021-01-18 14:46   ` Janosch Frank
  0 siblings, 0 replies; 6+ messages in thread
From: Janosch Frank @ 2021-01-18 14:46 UTC (permalink / raw)
  To: Thomas Huth; +Cc: david, borntraeger, imbrenda, cohuck, linux-s390

On 1/18/21 3:40 PM, Thomas Huth wrote:
> On 18/01/2021 15.03, Janosch Frank wrote:
>> On a program exception we usually skip the instruction that caused the
>> exception and continue. That won't work for UV calls since a "brc 3,0b"
>> will retry the instruction if the CC is > 1.
>>
>> Signed-off-by: Janosch Frank <frankja@linux.ibm.com>
>> ---
>>
>> I know this isn't very pretty.
>> I'm open for suggestions.
>>
>> ---
>>   lib/s390x/asm/uv.h | 14 ++++++++------
>>   1 file changed, 8 insertions(+), 6 deletions(-)
>>
>> diff --git a/lib/s390x/asm/uv.h b/lib/s390x/asm/uv.h
>> index 4c2fc48..252f1a3 100644
>> --- a/lib/s390x/asm/uv.h
>> +++ b/lib/s390x/asm/uv.h
>> @@ -53,21 +53,23 @@ struct uv_cb_share {
>>   static inline int uv_call(unsigned long r1, unsigned long r2)
>>   {
>>   	int cc;
>> +	struct lowcore *lc = 0x0;
>>   
>>   	/*
>> -	 * The brc instruction will take care of the cc 2/3 case where
>> -	 * we need to continue the execution because we were
>> -	 * interrupted. The inline assembly will only return on
>> -	 * success/error i.e. cc 0/1.
>> -	*/
>> +	 * CC 2 and 3 tell us to re-execute because the instruction
>> +	 * hasn't yet finished.
>> +	 */
>> +	lc->pgm_int_code = 0;
>> +retry:
>>   	asm volatile(
>>   		"0:	.insn rrf,0xB9A40000,%[r1],%[r2],0,0\n"
>> -		"		brc	3,0b\n"
>>   		"		ipm	%[cc]\n"
>>   		"		srl	%[cc],28\n"
>>   		: [cc] "=d" (cc)
>>   		: [r1] "a" (r1), [r2] "a" (r2)
>>   		: "memory", "cc");
>> +	if (!lc->pgm_int_code && cc > 1)
>> +		goto retry;
> 
> Why not simply:
> 
> 	do {
> 		asm volatile(...);
> 	} while (!lc->pgm_int_code && cc > 1)
> 
> ?

That would also be an option but it would basically be the same horrible
looking quick fix.

Claudio proposed implementing a one shot uv_call that doesn't branch
back. We should be able to use that purely for privilege checks.

> 
>    Thomas
> 

^ permalink raw reply	[flat|nested] 6+ messages in thread

* [kvm-unit-tests PATCH v2] s390x: Fix uv_call() exception behavior
  2021-01-18 14:03 [kvm-unit-tests PATCH] s390x: Fix uv_call() exception behavior Janosch Frank
  2021-01-18 14:40 ` Thomas Huth
@ 2021-01-18 15:09 ` Janosch Frank
  2021-01-18 15:15   ` Thomas Huth
  2021-01-19 16:28   ` Cornelia Huck
  1 sibling, 2 replies; 6+ messages in thread
From: Janosch Frank @ 2021-01-18 15:09 UTC (permalink / raw)
  To: kvm390 mailing list
  Cc: thuth, david, borntraeger, imbrenda, cohuck, linux-s390

On a program exception we usually skip the instruction that caused the
exception and continue. That won't work for UV calls since a "brc
3,0b" will retry the instruction if the CC is > 1. Let's forgo the brc
when checking for privilege exceptions and use a uv_call_once().

Signed-off-by: Janosch Frank <frankja@linux.ibm.com>
Suggested-by: Claudio Imbrenda <imbrenda@linux.ibm.com>

---
 lib/s390x/asm/uv.h | 24 ++++++++++++++++--------
 s390x/uv-guest.c   |  6 +++---
 2 files changed, 19 insertions(+), 11 deletions(-)

diff --git a/lib/s390x/asm/uv.h b/lib/s390x/asm/uv.h
index 4c2fc48..39d2dc0 100644
--- a/lib/s390x/asm/uv.h
+++ b/lib/s390x/asm/uv.h
@@ -50,19 +50,12 @@ struct uv_cb_share {
 	u64 reserved28;
 } __attribute__((packed))  __attribute__((aligned(8)));
 
-static inline int uv_call(unsigned long r1, unsigned long r2)
+static inline int uv_call_once(unsigned long r1, unsigned long r2)
 {
 	int cc;
 
-	/*
-	 * The brc instruction will take care of the cc 2/3 case where
-	 * we need to continue the execution because we were
-	 * interrupted. The inline assembly will only return on
-	 * success/error i.e. cc 0/1.
-	*/
 	asm volatile(
 		"0:	.insn rrf,0xB9A40000,%[r1],%[r2],0,0\n"
-		"		brc	3,0b\n"
 		"		ipm	%[cc]\n"
 		"		srl	%[cc],28\n"
 		: [cc] "=d" (cc)
@@ -71,4 +64,19 @@ static inline int uv_call(unsigned long r1, unsigned long r2)
 	return cc;
 }
 
+static inline int uv_call(unsigned long r1, unsigned long r2)
+{
+	int cc;
+
+	/*
+	 * CC 2 and 3 tell us to re-execute because the instruction
+	 * hasn't yet finished.
+	 */
+	do {
+		cc = uv_call_once(r1, r2);
+	} while (cc > 1);
+
+	return cc;
+}
+
 #endif
diff --git a/s390x/uv-guest.c b/s390x/uv-guest.c
index d47333e..091a19b 100644
--- a/s390x/uv-guest.c
+++ b/s390x/uv-guest.c
@@ -31,7 +31,7 @@ static void test_priv(void)
 	uvcb.len = sizeof(struct uv_cb_qui);
 	expect_pgm_int();
 	enter_pstate();
-	uv_call(0, (u64)&uvcb);
+	uv_call_once(0, (u64)&uvcb);
 	check_pgm_int_code(PGM_INT_CODE_PRIVILEGED_OPERATION);
 	report_prefix_pop();
 
@@ -40,7 +40,7 @@ static void test_priv(void)
 	uvcb.len = sizeof(struct uv_cb_share);
 	expect_pgm_int();
 	enter_pstate();
-	uv_call(0, (u64)&uvcb);
+	uv_call_once(0, (u64)&uvcb);
 	check_pgm_int_code(PGM_INT_CODE_PRIVILEGED_OPERATION);
 	report_prefix_pop();
 
@@ -49,7 +49,7 @@ static void test_priv(void)
 	uvcb.len = sizeof(struct uv_cb_share);
 	expect_pgm_int();
 	enter_pstate();
-	uv_call(0, (u64)&uvcb);
+	uv_call_once(0, (u64)&uvcb);
 	check_pgm_int_code(PGM_INT_CODE_PRIVILEGED_OPERATION);
 	report_prefix_pop();
 
-- 
2.25.1

^ permalink raw reply related	[flat|nested] 6+ messages in thread

* Re: [kvm-unit-tests PATCH v2] s390x: Fix uv_call() exception behavior
  2021-01-18 15:09 ` [kvm-unit-tests PATCH v2] " Janosch Frank
@ 2021-01-18 15:15   ` Thomas Huth
  2021-01-19 16:28   ` Cornelia Huck
  1 sibling, 0 replies; 6+ messages in thread
From: Thomas Huth @ 2021-01-18 15:15 UTC (permalink / raw)
  To: Janosch Frank, KVM; +Cc: david, borntraeger, imbrenda, cohuck, linux-s390

On 18/01/2021 16.09, Janosch Frank wrote:
> On a program exception we usually skip the instruction that caused the
> exception and continue. That won't work for UV calls since a "brc
> 3,0b" will retry the instruction if the CC is > 1. Let's forgo the brc
> when checking for privilege exceptions and use a uv_call_once().
> 
> Signed-off-by: Janosch Frank <frankja@linux.ibm.com>
> Suggested-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
> 
> ---
>   lib/s390x/asm/uv.h | 24 ++++++++++++++++--------
>   s390x/uv-guest.c   |  6 +++---
>   2 files changed, 19 insertions(+), 11 deletions(-)
> 
> diff --git a/lib/s390x/asm/uv.h b/lib/s390x/asm/uv.h
> index 4c2fc48..39d2dc0 100644
> --- a/lib/s390x/asm/uv.h
> +++ b/lib/s390x/asm/uv.h
> @@ -50,19 +50,12 @@ struct uv_cb_share {
>   	u64 reserved28;
>   } __attribute__((packed))  __attribute__((aligned(8)));
>   
> -static inline int uv_call(unsigned long r1, unsigned long r2)
> +static inline int uv_call_once(unsigned long r1, unsigned long r2)
>   {
>   	int cc;
>   
> -	/*
> -	 * The brc instruction will take care of the cc 2/3 case where
> -	 * we need to continue the execution because we were
> -	 * interrupted. The inline assembly will only return on
> -	 * success/error i.e. cc 0/1.
> -	*/
>   	asm volatile(
>   		"0:	.insn rrf,0xB9A40000,%[r1],%[r2],0,0\n"
> -		"		brc	3,0b\n"
>   		"		ipm	%[cc]\n"
>   		"		srl	%[cc],28\n"
>   		: [cc] "=d" (cc)
> @@ -71,4 +64,19 @@ static inline int uv_call(unsigned long r1, unsigned long r2)
>   	return cc;
>   }
>   
> +static inline int uv_call(unsigned long r1, unsigned long r2)
> +{
> +	int cc;
> +
> +	/*
> +	 * CC 2 and 3 tell us to re-execute because the instruction
> +	 * hasn't yet finished.
> +	 */
> +	do {
> +		cc = uv_call_once(r1, r2);
> +	} while (cc > 1);
> +
> +	return cc;
> +}
> +
>   #endif
> diff --git a/s390x/uv-guest.c b/s390x/uv-guest.c
> index d47333e..091a19b 100644
> --- a/s390x/uv-guest.c
> +++ b/s390x/uv-guest.c
> @@ -31,7 +31,7 @@ static void test_priv(void)
>   	uvcb.len = sizeof(struct uv_cb_qui);
>   	expect_pgm_int();
>   	enter_pstate();
> -	uv_call(0, (u64)&uvcb);
> +	uv_call_once(0, (u64)&uvcb);
>   	check_pgm_int_code(PGM_INT_CODE_PRIVILEGED_OPERATION);
>   	report_prefix_pop();
>   
> @@ -40,7 +40,7 @@ static void test_priv(void)
>   	uvcb.len = sizeof(struct uv_cb_share);
>   	expect_pgm_int();
>   	enter_pstate();
> -	uv_call(0, (u64)&uvcb);
> +	uv_call_once(0, (u64)&uvcb);
>   	check_pgm_int_code(PGM_INT_CODE_PRIVILEGED_OPERATION);
>   	report_prefix_pop();
>   
> @@ -49,7 +49,7 @@ static void test_priv(void)
>   	uvcb.len = sizeof(struct uv_cb_share);
>   	expect_pgm_int();
>   	enter_pstate();
> -	uv_call(0, (u64)&uvcb);
> +	uv_call_once(0, (u64)&uvcb);
>   	check_pgm_int_code(PGM_INT_CODE_PRIVILEGED_OPERATION);
>   	report_prefix_pop();

That looks nicer, indeed.

Reviewed-by: Thomas Huth <thuth@redhat.com>

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [kvm-unit-tests PATCH v2] s390x: Fix uv_call() exception behavior
  2021-01-18 15:09 ` [kvm-unit-tests PATCH v2] " Janosch Frank
  2021-01-18 15:15   ` Thomas Huth
@ 2021-01-19 16:28   ` Cornelia Huck
  1 sibling, 0 replies; 6+ messages in thread
From: Cornelia Huck @ 2021-01-19 16:28 UTC (permalink / raw)
  To: Janosch Frank; +Cc: thuth, david, borntraeger, imbrenda, linux-s390

On Mon, 18 Jan 2021 10:09:22 -0500
Janosch Frank <frankja@linux.ibm.com> wrote:

> On a program exception we usually skip the instruction that caused the
> exception and continue. That won't work for UV calls since a "brc
> 3,0b" will retry the instruction if the CC is > 1. Let's forgo the brc
> when checking for privilege exceptions and use a uv_call_once().
> 
> Signed-off-by: Janosch Frank <frankja@linux.ibm.com>
> Suggested-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
> 
> ---
>  lib/s390x/asm/uv.h | 24 ++++++++++++++++--------
>  s390x/uv-guest.c   |  6 +++---
>  2 files changed, 19 insertions(+), 11 deletions(-)

Reviewed-by: Cornelia Huck <cohuck@redhat.com>

^ permalink raw reply	[flat|nested] 6+ messages in thread

end of thread, other threads:[~2021-01-19 16:30 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-01-18 14:03 [kvm-unit-tests PATCH] s390x: Fix uv_call() exception behavior Janosch Frank
2021-01-18 14:40 ` Thomas Huth
2021-01-18 14:46   ` Janosch Frank
2021-01-18 15:09 ` [kvm-unit-tests PATCH v2] " Janosch Frank
2021-01-18 15:15   ` Thomas Huth
2021-01-19 16:28   ` Cornelia Huck

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.