VPHN parsing fixes
v2 cover VPHN parsing fixes
Hi,
This series addresses remarks from Ben and Michael (see individual patches). The most notable changes are:
- the parsing code being pull out into a separate file in patch 3/4. This allows to write userland tests like the one below.
- a full rewrite of the parsing logic in patch 4/4
--
#include <stdio.h>
#include <byteswap.h>
typedef unsigned long u64;
typedef unsigned int u32;
typedef unsigned short u16;
typedef enum { false = 0, true } bool;
#if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
#define cpu_to_be32(x) bswap_32(x)
#define be32_to_cpu(x) bswap_32(x)
#define be16_to_cpup(x) bswap_16(*x)
#define cpu_to_be64(x) bswap_64(x)
#else
#define cpu_to_be32(x) (x)
#define be32_to_cpu(x) (x)
#define be16_to_cpup(x) (*x)
#define cpu_to_be64(x) (x)
#endif
#define pr_debug(...) printf(__VA_ARGS__)
#include "vphn.c"
void print_packed(const long *packed)
{
char *p = (char*) packed;
int i;
printf("\nRegisters:\n");
for (i = 0; i < VPHN_REGISTER_COUNT; i++)
printf("0x%016lx\n", packed[i]);
printf("\nMemory layout:\n");
for (i = 0; i < 6; i++) {
printf("0x %02hhx %02hhx %02hhx %02hhx"
" %02hhx %02hhx %02hhx %02hhx\n",
*(p + 0), *(p + 1), *(p + 2), *(p + 3),
*(p + 4), *(p + 5), *(p + 6), *(p + 7));
p += 8;
}
putchar('\n');
}void print_unpacked(const __be32 *unpacked)
{
int i;
printf("\nVPHN associativity:\n");
for (i = 0; i <= be32_to_cpu(unpacked[0]); i++)
printf("0x%08x\n", be32_to_cpu(unpacked[i]));
putchar('\n');
}int main(int argc, char **argv)
{
int i;
struct {
const char *descr;
long packed[VPHN_REGISTER_COUNT];
} data[] = {
{
"16-bit and 32-bit",
0x8001800280038004,
0x8005800680078008,
0x000000090000000a,
0x0000000b0000000c,
0xffffffffffffffff,
0xffffffffffffffff
},
{
"filled with 16-bit",
0x8001800280038004,
0x8005800680078008,
0x8009800a800b800c,
0x800d800e800f8010,
0x8011801280138014,
0x8015801680178018,
},
{
"filled with 32-bit",
0x0000000100000002,
0x0000000300000004,
0x0000000500000006,
0x0000000700000008,
0x000000090000000a,
0x0000000b0000000c,
},
{
"32-bit has all ones in 16 lower bits",
0x0001ffff80028003,
0xffffffffffffffff,
0xffffffffffffffff,
0xffffffffffffffff,
0xffffffffffffffff,
0xffffffffffffffff,
},
{
"32-bit accross two 64-bit registers",
0x8001000000020000,
0x0003000000048005,
0xffffffffffffffff,
0xffffffffffffffff,
0xffffffffffffffff,
0xffffffffffffffff,
},
{
"Truncated last 32-bit",
0x0000000100000002,
0x0000000300000004,
0x0000000500000006,
0x0000000700000008,
0x000000090000000a,
0x0000000b800c0bad,
},
};
for (i = 0; i < sizeof(data) / sizeof(data[0]); i++) {
__be32 unpacked[VPHN_ASSOC_BUFSIZE] = { 0 };
printf("\n==================================================\n");
printf("\nSet #%d: %s\n", i, data[i].descr);
printf("\n==================================================\n");
print_packed(data[i].packed);
vphn_unpack_associativity(data[i].packed, unpacked);
print_unpacked(unpacked);
}
return 0;
}---
Greg Kurz (4):
powerpc/vphn: clarify the H_HOME_NODE_ASSOCIATIVITY API powerpc/vphn: move endianness fixing to vphn_unpack_associativity() powerpc/vphn: move VPHN parsing logic to a separate file powerpc/vphn: parsing code rewrite
arch/powerpc/mm/Makefile | 1 + arch/powerpc/mm/numa.c | 55 ++---------------------------------- arch/powerpc/mm/vphn.c | 70 ++++++++++++++++++++++++++++++++++++++++++++++ arch/powerpc/mm/vphn.h | 16 +++++++++++ 4 files changed, 90 insertions(+), 52 deletions(-) create mode 100644 arch/powerpc/mm/vphn.c create mode 100644 arch/powerpc/mm/vphn.h
--
Greg
PATCH v2 1/4 powerpc/vphn: clarify the H_HOME_NODE_ASSOCIATIVITY API
The number of values returned by the H_HOME_NODE_ASSOCIATIVITY h_call deserves to be explicitly defined, for a better understanding of the code.
---
No changes in v2.
arch/powerpc/mm/numa.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-)
arch/powerpc/mm/numa.c +6 -2
@@ -1247,11 +1247,15 @@ static int update_cpu_associativity_changes_mask(void)return cpumask_weight (changes); } +/* The H_HOME_NODE_ASSOCIATIVITY h_call returns 6 64-bit registers. +*/ +#define VPHN_REGISTER_COUNT 6 +/* * 6 64-bit registers unpacked into 12 32-bit associativity values. To form * the complete property we have to add the length in the first cell. */ -#define VPHN_ASSOC_BUFSIZE (6*sizeof(u64)/sizeof(u32) + 1) +#define VPHN_ASSOC_BUFSIZE (VPHN_REGISTER_COUNT*sizeof(u64)/sizeof(u32) + 1)/* * Convert the associativity domain numbers returned from the hypervisor
@@ -1309,7 +1313,7 @@ static long hcall_vphn(unsigned long cpu, __be32 *associativity)int i; rc =plpar_hcall9 (H_HOME_NODE_ASSOCIATIVITY , retbuf, flags, hwcpu); -for (i =0 ; i <6 ; i++) +for (i =0 ; i <VPHN_REGISTER_COUNT ; i++) retbuf[i] =cpu_to_be64 (retbuf[i]);vphn_unpack_associativity (retbuf, associativity);
PATCH v2 2/4 powerpc/vphn: move endianness fixing to vphn_unpack_associativity()
In reply to Greg KurzThe first argument to vphn_unpack_associativity() is a const long *, but the parsing code expects __be64 values actually. Let's move the endian fixing down for consistency.
---
v2: only move the endian fixing. The logic is reworked in another patch.
arch/powerpc/mm/numa.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-)
arch/powerpc/mm/numa.c +6 -4
@@ -1263,13 +1263,18 @@ static int update_cpu_associativity_changes_mask(void)*/ static int vphn_unpack_associativity (const long *packed,__be32 *unpacked) { +__be64 be_packed[VPHN_REGISTER_COUNT ];int i, nr_assoc_doms =0 ; -const __be16 *field = (const __be16 *) packed; +const __be16 *field = (const __be16 *) be_packed;#define VPHN_FIELD_UNUSED (0xffff)#define VPHN_FIELD_MSB (0x8000)#define VPHN_FIELD_MASK (~VPHN_FIELD_MSB) +/* Let's recreate the original stream. */ +for (i =0 ; i <VPHN_REGISTER_COUNT ; i++) + be_packed[i] =cpu_to_be64 (packed[i]); +for (i =1 ; i <VPHN_ASSOC_BUFSIZE ; i++) {if (be16_to_cpup (field) ==VPHN_FIELD_UNUSED ) {/* All significant fields processed, and remaining
@@ -1310,11 +1315,8 @@ static long hcall_vphn(unsigned long cpu, __be32 *associativity)long retbuf[PLPAR_HCALL9_BUFSIZE ] = {0 };u64 flags =1 ;int hwcpu =get_hard_smp_processor_id (cpu); -int i; rc =plpar_hcall9 (H_HOME_NODE_ASSOCIATIVITY , retbuf, flags, hwcpu); -for (i =0 ; i <VPHN_REGISTER_COUNT ; i++) - retbuf[i] =cpu_to_be64 (retbuf[i]);vphn_unpack_associativity (retbuf, associativity);return rc;
PATCH v2 3/4 powerpc/vphn: move VPHN parsing logic to a separate file
In reply to Greg KurzThe goal behind this patch is to be able to write userland tests for the VPHN parsing code.
--- arch/powerpc/mm/Makefile | 1 + arch/powerpc/mm/numa.c | 61 ++-------------------------------------------- arch/powerpc/mm/vphn.c | 50 ++++++++++++++++++++++++++++++++++++++ arch/powerpc/mm/vphn.h | 16 ++++++++++++ 4 files changed, 70 insertions(+), 58 deletions(-) create mode 100644 arch/powerpc/mm/vphn.c create mode 100644 arch/powerpc/mm/vphn.h
arch/powerpc/mm/Makefile +1 -0
@@ -24,6 +24,7 @@ obj-$(CONFIG_40x) += 40x_mmu.o obj-$(CONFIG_44x) += 44x_mmu.o obj-$(CONFIG_PPC_FSL_BOOK3E) += fsl_booke_mmu.o obj-$(CONFIG_NEED_MULTIPLE_NODES) += numa.o +obj-$(CONFIG_PPC_SPLPAR) += vphn.o obj-$(CONFIG_PPC_MM_SLICES) += slice.o obj-y += hugetlbpage.o ifeq ($(CONFIG_HUGETLB_PAGE),y)
arch/powerpc/mm/numa.c +3 -58
73 lines not on this page · raw message
arch/powerpc/mm/vphn.c new file +50 -0
50 lines not on this page · raw message
arch/powerpc/mm/vphn.h new file +16 -0
@@ -0,0 +1,16 @@ +#ifndef _ARCH_POWERPC_MM_VPHN_H_ +#define _ARCH_POWERPC_MM_VPHN_H_ + +/* The H_HOME_NODE_ASSOCIATIVITY h_call returns 6 64-bit registers. +*/ +#define VPHN_REGISTER_COUNT 6 + +/* +* 6 64-bit registers unpacked into 12 32-bit associativity values. To form +* the complete property we have to add the length in the first cell. +*/ +#define VPHN_ASSOC_BUFSIZE (VPHN_REGISTER_COUNT*sizeof(u64)/sizeof(u32) + 1) + +extern int vphn_unpack_associativity (const long *packed,__be32 *unpacked); + +#endif
PATCH v2 4/4 powerpc/vphn: parsing code rewrite
In reply to Greg KurzThe current VPHN parsing logic has some flaws that this patch aims to fix:
1) when the value 0xffff is read, the value 0xffffffff gets added to the the output list and its element count isn't incremented. This is wrong. According to PAPR+ the domain identifiers are packed into a sequence terminated by the "reserved value of all ones". This means that 0xffff is a stream terminator.
2) the combination of byteswaps and casts make the code hardly readable. Let's parse the stream one 16-bit field at a time instead.
3) it is assumed that the hypercall returns 12 32-bit values packed into 6 64-bit registers. According to PAPR+, the domain identifiers may be streamed as 16-bit values. Let's increase the number of expected numbers to 24.
--- arch/powerpc/mm/vphn.c | 54 +++++++++++++++++++++++++++++++++--------------- arch/powerpc/mm/vphn.h | 6 +++-- 2 files changed, 40 insertions(+), 20 deletions(-)
arch/powerpc/mm/vphn.c +37 -17
@@ -2,44 +2,64 @@#include "vphn.h" /* -* Convert the associativity domain numbers returned from the hypervisor -* to the sequence they would appear in the ibm,associativity property. +* The associativity domain numbers are returned from the hypervisor as a +* stream of mixed 16-bit and 32-bit fields. The stream is terminated by the +* special value of "all ones" (aka. 0xffff) and its size may not exceed 48 +* bytes. +* +* --- 16-bit fields --> +* _________________________ +* | 0 | 1 | 2 | 3 | be_packed[0] +* ------+-----+-----+------ +* _________________________ +* | 4 | 5 | 6 | 7 | be_packed[1] +* ------------------------- +* ... +* _________________________ +* | 20 | 21 | 22 | 23 | be_packed[5] +* ------------------------- +* +* Convert to the sequence they would appear in the ibm,associativity property. */ int vphn_unpack_associativity (const long *packed,__be32 *unpacked) {__be64 be_packed[VPHN_REGISTER_COUNT ];int i, nr_assoc_doms =0 ;const __be16 *field = (const __be16 *) be_packed; +u16 last =0 ; +bool is_32bit = false;#define VPHN_FIELD_UNUSED (0xffff)#define VPHN_FIELD_MSB (0x8000)#define VPHN_FIELD_MASK (~VPHN_FIELD_MSB) -/* Let's recreate the original stream. */ +/* Let's fix the values returned by plpar_hcall9() */ for (i =0 ; i <VPHN_REGISTER_COUNT ; i++) be_packed[i] =cpu_to_be64 (packed[i]);for (i =1 ; i <VPHN_ASSOC_BUFSIZE ; i++) { -if (be16_to_cpup (field) ==VPHN_FIELD_UNUSED ) { -/* All significant fields processed, and remaining -* fields contain the reserved value of all 1's. -* Just store them. +u16 new =be16_to_cpup (field++); + +if (is_32bit) { +/* Let's concatenate the 16 bits of this field to the +* 15 lower bits of the previous field */ - unpacked[i] = *((__be32 *)field); - field +=2 ; - }else if (be16_to_cpup (field) &VPHN_FIELD_MSB ) { + unpacked[++nr_assoc_doms] = +cpu_to_be32 (last <<16 | new); + is_32bit = false; + }else if (new ==VPHN_FIELD_UNUSED ) +/* This is the list terminator */ +break ; +else if (new &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++; - nr_assoc_doms++; + unpacked[++nr_assoc_doms] = +cpu_to_be32 (new &VPHN_FIELD_MASK ); }else {/* Data is in the lower 15 bits of this field * concatenated with the next 16 bit field */ - unpacked[i] = *((__be32 *)field); - field +=2 ; - nr_assoc_doms++; + last = new; + is_32bit = true; } }
arch/powerpc/mm/vphn.h +3 -3
@@ -6,10 +6,10 @@#define VPHN_REGISTER_COUNT 6/* -* 6 64-bit registers unpacked into 12 32-bit associativity values. To form -* the complete property we have to add the length in the first cell. +* 6 64-bit registers unpacked into up to 24 be32 associativity values. To +* form the complete property we have to add the length in the first cell. */ -#define VPHN_ASSOC_BUFSIZE (VPHN_REGISTER_COUNT*sizeof(u64)/sizeof(u32) + 1) +#define VPHN_ASSOC_BUFSIZE (VPHN_REGISTER_COUNT*sizeof(u64)/sizeof(u16) + 1)extern int vphn_unpack_associativity (const long *packed,__be32 *unpacked);
Quoted 8 lines · On Wed, 17 Dec 2014 10:40:46 +0100 Greg Kurz <[email protected]> wrote:
Hi,
This series addresses remarks from Ben and Michael (see individual patches). The most notable changes are:
- the parsing code being pull out into a separate file in patch 3/4. This allows to write userland tests like the one below.
- a full rewrite of the parsing logic in patch 4/4
Ping ?
Quoted 157 lines
--
#include <stdio.h>
#include <byteswap.h>typedef unsigned long u64;
typedef unsigned int u32;
typedef unsigned short u16;
typedef enum { false = 0, true } bool;#if
__BYTE_ORDER__==__ORDER_LITTLE_ENDIAN__
#definecpu_to_be32(x)bswap_32(x)
#definebe32_to_cpu(x)bswap_32(x)
#definebe16_to_cpup(x)bswap_16(*x)
#definecpu_to_be64(x)bswap_64(x)
#else
#definecpu_to_be32(x)(x)
#definebe32_to_cpu(x)(x)
#definebe16_to_cpup(x)(*x)
#definecpu_to_be64(x)(x)
#endif#define
pr_debug(...)printf(__VA_ARGS__)#include "
vphn.c"void
print_packed(const long *packed){ char *p = (char*) packed; int i; printf("\nRegisters:\n"); for (i = 0; i < VPHN_REGISTER_COUNT; i++) printf("0x%016lx\n", packed[i]); printf("\nMemory layout:\n"); for (i = 0; i < 6; i++) { printf("0x %02hhx %02hhx %02hhx %02hhx" " %02hhx %02hhx %02hhx %02hhx\n", *(p + 0), *(p + 1), *(p + 2), *(p + 3), *(p + 4), *(p + 5), *(p + 6), *(p + 7)); p += 8; } putchar('\n'); }void
print_unpacked(const __be32 *unpacked){ int i; printf("\nVPHN associativity:\n"); for (i = 0; i <= be32_to_cpu(unpacked[0]); i++) printf("0x%08x\n", be32_to_cpu(unpacked[i])); putchar('\n'); }int
main(int argc, char **argv){ int i; struct { const char *descr; long packed[VPHN_REGISTER_COUNT]; } data[] = { { "16-bit and 32-bit", 0x8001800280038004, 0x8005800680078008, 0x000000090000000a, 0x0000000b0000000c, 0xffffffffffffffff, 0xffffffffffffffff }, { "filled with 16-bit", 0x8001800280038004, 0x8005800680078008, 0x8009800a800b800c, 0x800d800e800f8010, 0x8011801280138014, 0x8015801680178018, }, { "filled with 32-bit", 0x0000000100000002, 0x0000000300000004, 0x0000000500000006, 0x0000000700000008, 0x000000090000000a, 0x0000000b0000000c, }, { "32-bit has all ones in 16 lower bits", 0x0001ffff80028003, 0xffffffffffffffff, 0xffffffffffffffff, 0xffffffffffffffff, 0xffffffffffffffff, 0xffffffffffffffff, }, { "32-bit accross two 64-bit registers", 0x8001000000020000, 0x0003000000048005, 0xffffffffffffffff, 0xffffffffffffffff, 0xffffffffffffffff, 0xffffffffffffffff, }, { "Truncated last 32-bit", 0x0000000100000002, 0x0000000300000004, 0x0000000500000006, 0x0000000700000008, 0x000000090000000a, 0x0000000b800c0bad, }, }; for (i = 0; i < sizeof(data) / sizeof(data[0]); i++) { __be32 unpacked[VPHN_ASSOC_BUFSIZE] = { 0 }; printf("\n==================================================\n"); printf("\nSet #%d: %s\n", i, data[i].descr); printf("\n==================================================\n"); print_packed(data[i].packed); vphn_unpack_associativity(data[i].packed, unpacked); print_unpacked(unpacked); } return 0; }---
Greg Kurz (4):
powerpc/vphn: clarify theH_HOME_NODE_ASSOCIATIVITYAPI powerpc/vphn: move endianness fixing tovphn_unpack_associativity()powerpc/vphn: move VPHN parsing logic to a separate file powerpc/vphn: parsing code rewritearch/powerpc/mm/Makefile | 1 + arch/powerpc/mm/numa.c | 55 ++---------------------------------- arch/powerpc/mm/vphn.c | 70 ++++++++++++++++++++++++++++++++++++++++++++++ arch/powerpc/mm/vphn.h | 16 +++++++++++ 4 files changed, 90 insertions(+), 52 deletions(-) create mode 100644 arch/powerpc/mm/vphn.c create mode 100644 arch/powerpc/mm/vphn.h--
Greg_______________________________________________
Linuxppc-dev mailing list
[email protected]
https://lists.ozlabs.org/listinfo/linuxppc-dev
Michael Ellerman
reviewerQuoted 4 lines · Greg Kurz
Quoted 2 lines · On Wed, 17 Dec 2014 10:40:46 +0100 Greg Kurz <[email protected]> wrote:
Hi,
Quoted 8 lines
Quoted 6 lines
This series addresses remarks from Ben and Michael (see individual patches). The most notable changes are:
- the parsing code being pull out into a separate file in patch 3/4. This allows to write userland tests like the one below.
- a full rewrite of the parsing logic in patch 4/4Ping ?
Sorry, lots of patches needing review.
This looks pretty good at a glance, but did you actually write a userspace test for it? If so please send it. I'm happy to rework it into something that can go in selftests.
cheers
Quoted 12 lines · On Tue, 03 Feb 2015 13:47:35 +1100 Michael Ellerman <[email protected]> wrote:
Quoted 11 lines · Greg Kurz
Quoted 8 lines · On Wed, 17 Dec 2014 10:40:46 +0100 Greg Kurz <[email protected]> wrote:
Hi,
This series addresses remarks from Ben and Michael (see individual patches). The most notable changes are:
- the parsing code being pull out into a separate file in patch 3/4. This allows to write userland tests like the one below.
- a full rewrite of the parsing logic in patch 4/4
Quoted 4 lines
Quoted 1 line
Ping ?
Sorry, lots of patches needing review.
Heh, no problem. :) Since this isn't bugfix, I can even repost later, when the review pressure is lower.
Quoted 4 lines
This looks pretty good at a glance, but did you actually write a userspace test for it? If so please send it. I'm happy to rework it into something that can go in selftests.
Yes, I wrote the premise of a test program. You can find it in the cover mail of this series:
https://lists.ozlabs.org/pipermail/linuxppc-dev/2014-December/123601.html
I'll have a look at selftests.
Quoted 3 lines
cheers
Thanks.
--
Greg
Michael Ellerman
reviewerQuoted 15 lines · Greg Kurz
Quoted 12 lines · On Tue, 03 Feb 2015 13:47:35 +1100 Michael Ellerman <[email protected]> wrote:
Quoted 11 lines · Greg Kurz
Quoted 8 lines · On Wed, 17 Dec 2014 10:40:46 +0100 Greg Kurz <[email protected]> wrote:
Hi,
This series addresses remarks from Ben and Michael (see individual patches). The most notable changes are:
- the parsing code being pull out into a separate file in patch 3/4. This allows to write userland tests like the one below.
- a full rewrite of the parsing logic in patch 4/4
Quoted 6 lines
Quoted 3 lines
Quoted 1 line
Ping ?
Sorry, lots of patches needing review.
Heh, no problem. :) Since this isn't bugfix, I can even repost later, when the review pressure is lower.
Yeah if you can repost in ~2 weeks that would be perfect.
Quoted 11 lines
Quoted 4 lines
This looks pretty good at a glance, but did you actually write a userspace test for it? If so please send it. I'm happy to rework it into something that can go in selftests.
Yes, I wrote the premise of a test program. You can find it in the cover mail of this series:
https://lists.ozlabs.org/pipermail/linuxppc-dev/2014-December/123601.html
I'll have a look at selftests.
Thanks. That test looks like a good start, there might be some more cases to test though?
If you just copy the tools/testing/selftests/powerpc/tm directory and rename to vhpn, you can probably work out the rest.
cheers