The associativity domain numbers are obtained from the hypervisor through
registers and written into memory by the guest: the packed array passed to
vphn_unpack_associativity() is then native-endian, unlike what was assumed
in the following commit:
commit b08a2a12e44eaec5024b2b969f4fcb98169d1ca3
Author: Alistair Popple [off-list ref]
Date: Wed Aug 7 02:01:44 2013 +1000
powerpc: Make NUMA device node code endian safe
If a CPU home node changes, the topology gets filled with
bogus values. This leads to severe performance breakdowns.
This patch does two things:
- extract values from the packed array with shifts, in order to be endian
neutral
- convert the resulting values to be32 as expected
Suggested-by: Anton Blanchard <redacted>
Signed-off-by: Greg Kurz <redacted>
---
I could test this code in a userland program and obtain the same
result in little and big endian. If the 64-bit packed values
are:
0x8001ffffffff0000
0x112200003344ffff
0xffff000055660000
0x7788000099aaffff
0xffff8002ffffffff
0xffffffffffffffff
then the unpacked array is:
0x00000007
0x00000001
0xffffffff
0x00001122
0x00003344
0xffffffff
0x00005566
0x00007788
0x000099aa
0xffffffff
0x00000002
0xffffffff
0xffffffff
I could not test in a PowerVM guest though, hence the RFC.
--
Greg
arch/powerpc/mm/numa.c | 50 ++++++++++++++++++++++++++++++++++++++----------
1 file changed, 40 insertions(+), 10 deletions(-)
diff --git a/arch/powerpc/mm/numa.c b/arch/powerpc/mm/numa.c
index b835bf0..06af179 100644
--- a/arch/powerpc/mm/numa.c
+++ b/arch/powerpc/mm/numa.c
@@ -1369,38 +1369,68 @@ static int update_cpu_associativity_changes_mask(void)
#define VPHN_ASSOC_BUFSIZE (6*sizeof(u64)/sizeof(u32) + 1)
/*
+ * The associativity values are either 16-bit (VPHN_FIELD_MSB) or 32-bit (data
+ * or VPHN_FIELD_UNUSED). We hence need to parse the packed array into 16-bit
+ * chunks. Let's do that with bit shifts to be endian neutral.
+ *
+ * --- 16-bit chunks -->
+ * _________________________
+ * | 0 | 1 | 2 | 3 | packed[0]
+ * -------------------------
+ * _________________________
+ * | 4 | 5 | 6 | 7 | packed[1]
+ * -------------------------
+ * ...
+ * _________________________
+ * | 20 | 21 | 22 | 23 | packed[5]
+ * -------------------------
+ *
+ * >>48 >>32 >>16 >>0 <-- needed bit shift
+ *
+ * We only care for the 2 lower bits of the chunk index to compute the shift.
+ */
+static inline u16 read_vphn_chunk(const long *packed, unsigned int i)
+{
+ return packed[i >> 2] >> ((~i & 3) << 4);
+}
+
+/*
* Convert the associativity domain numbers returned from the hypervisor
* to the sequence they would appear in the ibm,associativity property.
*/
static int vphn_unpack_associativity(const long *packed, __be32 *unpacked)
{
- int i, nr_assoc_doms = 0;
+ unsigned int i, j, nr_assoc_doms = 0;
const __be16 *field = (const __be16 *) packed;
#define VPHN_FIELD_UNUSED (0xffff)
#define VPHN_FIELD_MSB (0x8000)
#define VPHN_FIELD_MASK (~VPHN_FIELD_MSB)
- for (i = 1; i < VPHN_ASSOC_BUFSIZE; i++) {
- if (be16_to_cpup(field) == VPHN_FIELD_UNUSED) {
+ for (i = 1, j = 0; i < VPHN_ASSOC_BUFSIZE; i++) {
+ u16 field = read_vphn_chunk(packed, j);
+
+ if (field == VPHN_FIELD_UNUSED) {
/* All significant fields processed, and remaining
* fields contain the reserved value of all 1's.
* Just store them.
*/
- unpacked[i] = *((__be32 *)field);
- field += 2;
+ unpacked[i] = (VPHN_FIELD_UNUSED << 16 |
+ VPHN_FIELD_UNUSED);
+ j += 2;
} else if (be16_to_cpup(field) & VPHN_FIELD_MSB) {
/* Data is in the lower 15 bits of this field */
- unpacked[i] = cpu_to_be32(
- be16_to_cpup(field) & VPHN_FIELD_MASK);
- field++;
+ unpacked[i] = cpu_to_be32(field & VPHN_FIELD_MASK);
+ j++;
nr_assoc_doms++;
} else {
/* Data is in the lower 15 bits of this field
* concatenated with the next 16 bit field
*/
- unpacked[i] = *((__be32 *)field);
- field += 2;
+ unpacked[i] =
+ cpu_to_be32((u32) field << 16 |
+ read_vphn_chunk(packed, j + 1));
+ j += 2;
nr_assoc_doms++;
}
}On 02.10.2014 [23:59:15 +0200], Greg Kurz wrote:
quoted hunk
The associativity domain numbers are obtained from the hypervisor through
registers and written into memory by the guest: the packed array passed to
vphn_unpack_associativity() is then native-endian, unlike what was assumed
in the following commit:
commit b08a2a12e44eaec5024b2b969f4fcb98169d1ca3
Author: Alistair Popple [off-list ref]
Date: Wed Aug 7 02:01:44 2013 +1000
powerpc: Make NUMA device node code endian safe
If a CPU home node changes, the topology gets filled with
bogus values. This leads to severe performance breakdowns.
This patch does two things:
- extract values from the packed array with shifts, in order to be endian
neutral
- convert the resulting values to be32 as expected
Suggested-by: Anton Blanchard <redacted>
Signed-off-by: Greg Kurz <redacted>
---
I could test this code in a userland program and obtain the same
result in little and big endian. If the 64-bit packed values
are:
0x8001ffffffff0000
0x112200003344ffff
0xffff000055660000
0x7788000099aaffff
0xffff8002ffffffff
0xffffffffffffffff
then the unpacked array is:
0x00000007
0x00000001
0xffffffff
0x00001122
0x00003344
0xffffffff
0x00005566
0x00007788
0x000099aa
0xffffffff
0x00000002
0xffffffff
0xffffffff
I could not test in a PowerVM guest though, hence the RFC.
--
Greg
arch/powerpc/mm/numa.c | 50 ++++++++++++++++++++++++++++++++++++++----------
1 file changed, 40 insertions(+), 10 deletions(-)
diff --git a/arch/powerpc/mm/numa.c b/arch/powerpc/mm/numa.c
index b835bf0..06af179 100644
--- a/arch/powerpc/mm/numa.c
+++ b/arch/powerpc/mm/numa.c
@@ -1369,38 +1369,68 @@ static int update_cpu_associativity_changes_mask(void)
#define VPHN_ASSOC_BUFSIZE (6*sizeof(u64)/sizeof(u32) + 1)
/*
+ * The associativity values are either 16-bit (VPHN_FIELD_MSB) or 32-bit (data
+ * or VPHN_FIELD_UNUSED). We hence need to parse the packed array into 16-bit
+ * chunks. Let's do that with bit shifts to be endian neutral.
+ *
+ * --- 16-bit chunks -->
+ * _________________________
+ * | 0 | 1 | 2 | 3 | packed[0]
+ * -------------------------
+ * _________________________
+ * | 4 | 5 | 6 | 7 | packed[1]
+ * -------------------------
+ * ...
+ * _________________________
+ * | 20 | 21 | 22 | 23 | packed[5]
+ * -------------------------
+ *
+ * >>48 >>32 >>16 >>0 <-- needed bit shift
+ *
+ * We only care for the 2 lower bits of the chunk index to compute the shift.
+ */
+static inline u16 read_vphn_chunk(const long *packed, unsigned int i)
+{
+ return packed[i >> 2] >> ((~i & 3) << 4);
This is some excellent magic and the comment *should* be sufficient, but
maybe an example would be good? i is the index we want to read from in
the packed array?
quoted hunk
+}
+
+/*
* Convert the associativity domain numbers returned from the hypervisor
* to the sequence they would appear in the ibm,associativity property.
*/
static int vphn_unpack_associativity(const long *packed, __be32 *unpacked)
{
- int i, nr_assoc_doms = 0;
+ unsigned int i, j, nr_assoc_doms = 0;
const __be16 *field = (const __be16 *) packed;
#define VPHN_FIELD_UNUSED (0xffff)
#define VPHN_FIELD_MSB (0x8000)
#define VPHN_FIELD_MASK (~VPHN_FIELD_MSB)
- for (i = 1; i < VPHN_ASSOC_BUFSIZE; i++) {
- if (be16_to_cpup(field) == VPHN_FIELD_UNUSED) {
+ for (i = 1, j = 0; i < VPHN_ASSOC_BUFSIZE; i++) {
+ u16 field = read_vphn_chunk(packed, j);
Maybe I'm super dense here, but isn't there are already a variable named
field in this context? With a different annotation?
quoted hunk
+
+ if (field == VPHN_FIELD_UNUSED) {
/* All significant fields processed, and remaining
* fields contain the reserved value of all 1's.
* Just store them.
*/
- unpacked[i] = *((__be32 *)field);
- field += 2;
+ unpacked[i] = (VPHN_FIELD_UNUSED << 16 |
+ VPHN_FIELD_UNUSED);
+ j += 2;
} else if (be16_to_cpup(field) & VPHN_FIELD_MSB) {
Why do we need to do be16_to_cpup(field) here if field is now a u16? Or
more explicitly, if we don't performe be16_to_cpup(field) anywhere else
in this function now, why do we need to here?
/* Data is in the lower 15 bits of this field */
- unpacked[i] = cpu_to_be32(
- be16_to_cpup(field) & VPHN_FIELD_MASK);
- field++;
+ unpacked[i] = cpu_to_be32(field & VPHN_FIELD_MASK);
+ j++;
nr_assoc_doms++;
} else {
/* Data is in the lower 15 bits of this field
* concatenated with the next 16 bit field
*/
- unpacked[i] = *((__be32 *)field);
- field += 2;
+ unpacked[i] =
+ cpu_to_be32((u32) field << 16 |
+ read_vphn_chunk(packed, j + 1));
+ j += 2;
nr_assoc_doms++;
}
}
_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev@lists.ozlabs.org
https://lists.ozlabs.org/listinfo/linuxppc-dev
On Thu, 2 Oct 2014 16:43:41 -0700
Nishanth Aravamudan [off-list ref] wrote:
On 02.10.2014 [23:59:15 +0200], Greg Kurz wrote:
quoted
The associativity domain numbers are obtained from the hypervisor through
registers and written into memory by the guest: the packed array passed to
vphn_unpack_associativity() is then native-endian, unlike what was assumed
in the following commit:
commit b08a2a12e44eaec5024b2b969f4fcb98169d1ca3
Author: Alistair Popple [off-list ref]
Date: Wed Aug 7 02:01:44 2013 +1000
powerpc: Make NUMA device node code endian safe
If a CPU home node changes, the topology gets filled with
bogus values. This leads to severe performance breakdowns.
This patch does two things:
- extract values from the packed array with shifts, in order to be endian
neutral
- convert the resulting values to be32 as expected
Suggested-by: Anton Blanchard <redacted>
Signed-off-by: Greg Kurz <redacted>
---
I could test this code in a userland program and obtain the same
result in little and big endian. If the 64-bit packed values
are:
0x8001ffffffff0000
0x112200003344ffff
0xffff000055660000
0x7788000099aaffff
0xffff8002ffffffff
0xffffffffffffffff
then the unpacked array is:
0x00000007
0x00000001
0xffffffff
0x00001122
0x00003344
0xffffffff
0x00005566
0x00007788
0x000099aa
0xffffffff
0x00000002
0xffffffff
0xffffffff
I could not test in a PowerVM guest though, hence the RFC.
--
Greg
arch/powerpc/mm/numa.c | 50 ++++++++++++++++++++++++++++++++++++++----------
1 file changed, 40 insertions(+), 10 deletions(-)
diff --git a/arch/powerpc/mm/numa.c b/arch/powerpc/mm/numa.c
index b835bf0..06af179 100644
--- a/arch/powerpc/mm/numa.c
+++ b/arch/powerpc/mm/numa.c
@@ -1369,38 +1369,68 @@ static int update_cpu_associativity_changes_mask(void)
#define VPHN_ASSOC_BUFSIZE (6*sizeof(u64)/sizeof(u32) + 1)
/*
+ * The associativity values are either 16-bit (VPHN_FIELD_MSB) or 32-bit (data
+ * or VPHN_FIELD_UNUSED). We hence need to parse the packed array into 16-bit
+ * chunks. Let's do that with bit shifts to be endian neutral.
+ *
+ * --- 16-bit chunks -->
+ * _________________________
+ * | 0 | 1 | 2 | 3 | packed[0]
+ * -------------------------
+ * _________________________
+ * | 4 | 5 | 6 | 7 | packed[1]
+ * -------------------------
+ * ...
+ * _________________________
+ * | 20 | 21 | 22 | 23 | packed[5]
+ * -------------------------
+ *
+ * >>48 >>32 >>16 >>0 <-- needed bit shift
+ *
+ * We only care for the 2 lower bits of the chunk index to compute the shift.
+ */
+static inline u16 read_vphn_chunk(const long *packed, unsigned int i)
+{
+ return packed[i >> 2] >> ((~i & 3) << 4);
This is some excellent magic and the comment *should* be sufficient, but
maybe an example would be good? i is the index we want to read from in
the packed array?
Yes, i is the index of the 16-bit chunk we want to read. I will add
numerical examples:
chunk #0 is (u16) packed[0] >> 48
chunk #1 is (u16) packed[0] >> 32
chunk #2 is (u16) packed[0] >> 16
chunk #3 is (u16) packed[0] >> 0
chunk #4 is (u16) packed[1] >> 48
chunk #5 is (u16) packed[1] >> 32
...
chunk #22 is (u16) packed[1] >> 16
chunk #23 is (u16) packed[1] >> 0
quoted
+}
+
+/*
* Convert the associativity domain numbers returned from the hypervisor
* to the sequence they would appear in the ibm,associativity property.
*/
static int vphn_unpack_associativity(const long *packed, __be32 *unpacked)
{
- int i, nr_assoc_doms = 0;
+ unsigned int i, j, nr_assoc_doms = 0;
const __be16 *field = (const __be16 *) packed;
#define VPHN_FIELD_UNUSED (0xffff)
#define VPHN_FIELD_MSB (0x8000)
#define VPHN_FIELD_MASK (~VPHN_FIELD_MSB)
- for (i = 1; i < VPHN_ASSOC_BUFSIZE; i++) {
- if (be16_to_cpup(field) == VPHN_FIELD_UNUSED) {
+ for (i = 1, j = 0; i < VPHN_ASSOC_BUFSIZE; i++) {
+ u16 field = read_vphn_chunk(packed, j);
Maybe I'm super dense here, but isn't there are already a variable named
field in this context? With a different annotation?
Oops ! The u16 one is supposed to supersede the __be16 * one :)
- const __be16 *field = (const __be16 *) packed;
quoted
+
+ if (field == VPHN_FIELD_UNUSED) {
/* All significant fields processed, and remaining
* fields contain the reserved value of all 1's.
* Just store them.
*/
- unpacked[i] = *((__be32 *)field);
- field += 2;
+ unpacked[i] = (VPHN_FIELD_UNUSED << 16 |
+ VPHN_FIELD_UNUSED);
+ j += 2;
} else if (be16_to_cpup(field) & VPHN_FIELD_MSB) {
Why do we need to do be16_to_cpup(field) here if field is now a u16? Or
more explicitly, if we don't performe be16_to_cpup(field) anywhere else
in this function now, why do we need to here?
Same as above ^^
- } else if (be16_to_cpup(field) & VPHN_FIELD_MSB) {
+ } else if (field & VPHN_FIELD_MSB) {
Bad copy/paste from the prototype code... :\
Thanks for the review.
quoted
/* Data is in the lower 15 bits of this field */
- unpacked[i] = cpu_to_be32(
- be16_to_cpup(field) & VPHN_FIELD_MASK);
- field++;
+ unpacked[i] = cpu_to_be32(field & VPHN_FIELD_MASK);
+ j++;
nr_assoc_doms++;
} else {
/* Data is in the lower 15 bits of this field
* concatenated with the next 16 bit field
*/
- unpacked[i] = *((__be32 *)field);
- field += 2;
+ unpacked[i] =
+ cpu_to_be32((u32) field << 16 |
+ read_vphn_chunk(packed, j + 1));
+ j += 2;
nr_assoc_doms++;
}
}
_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev@lists.ozlabs.org
https://lists.ozlabs.org/listinfo/linuxppc-dev
_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev@lists.ozlabs.org
https://lists.ozlabs.org/listinfo/linuxppc-dev