powerpc: Make VSID_BITS* dependency explicit
PATCH 1/2 powerpc: Make VSID_BITS* dependency explicit
From: "Aneesh Kumar K.V" <[email protected]>
VSID_BITS and VSID_BITS_1T depends on the context bits and user esid bits. Make the dependency explicit
--- arch/powerpc/include/asm/mmu-hash64.h | 11 ++++++----- 1 file changed, 6 insertions(+), 5 deletions(-)
arch/powerpc/include/asm/mmu-hash64.h +6 -5
@@ -381,21 +381,22 @@ extern void slb_set_size(u16 size);* hash collisions. */ +#define CONTEXT_BITS 19 +#define USER_ESID_BITS 18 +#define USER_ESID_BITS_1T 6 +/* * This should be computed such that protovosid * vsid_mulitplier * doesn't overflow 64 bits. It should also be co-prime to vsid_modulus */ #define VSID_MULTIPLIER_256M ASM_CONST(12538073)/* 24-bit prime */ -#define VSID_BITS_256M 38 +#define VSID_BITS_256M (CONTEXT_BITS + USER_ESID_BITS + 1)#define VSID_MODULUS_256M ((1UL<<VSID_BITS_256M)-1)#define VSID_MULTIPLIER_1T ASM_CONST(12538073)/* 24-bit prime */ -#define VSID_BITS_1T 26 +#define VSID_BITS_1T (CONTEXT_BITS + USER_ESID_BITS_1T + 1)#define VSID_MODULUS_1T ((1UL<<VSID_BITS_1T)-1) -#define CONTEXT_BITS 19 -#define USER_ESID_BITS 18 -#define USER_ESID_BITS_1T 6#define USER_VSID_RANGE (1UL << (USER_ESID_BITS + SID_SHIFT))
Signature · 1 line
1.7.10
PATCH 2/2 powerpc: Make context bits depend on virtual addr size.
From: "Aneesh Kumar K.V" <[email protected]>
Different platforms supports different virtual addr size(n bits). We need to mak sure 0:77-n bits of the VA generated is forced to zero.
--- arch/powerpc/include/asm/mmu-hash64.h | 22 +++++++++++++++++++++- 1 file changed, 21 insertions(+), 1 deletion(-)
arch/powerpc/include/asm/mmu-hash64.h +21 -1
@@ -381,7 +381,27 @@ extern void slb_set_size(u16 size);* hash collisions. */ -#define CONTEXT_BITS 19 +/* +* Be careful with this value. This determines the VSID_MODULUS_* and that +* need to be co-prime with VSID_MULTIPLIER* +*/ +#if defined(CONFIG_POWER6_CPU ) || defined(CONFIG_POWER7_CPU ) +/* +* Even if cpu support 68 bits, we limit this to 66 because +* we support only 2^19 context. +*/ +#define MAX_VIRTUAL_ADDR_BITS 66 +#else +/* power4,power 5 and cell is 65 */ +#define MAX_VIRTUAL_ADDR_BITS 65 +#endif + +/* +* One bit is taken by the kernel, only the rest of space is available for the +* user space. +*/ +#define CONTEXT_BITS (MAX_VIRTUAL_ADDR_BITS - \ + (USER_ESID_BITS + SID_SHIFT + 1))#define USER_ESID_BITS 18#define USER_ESID_BITS_1T 6
Signature · 1 line
1.7.10
Re: [PATCH 2/2] powerpc: Make context bits depend on virtual addr size.
Quoted 4 lines · Aneesh Kumar K.V
+ * Be careful with this value. This determines the VSID_MODULUS_* and that
+ * need to be co-prime with VSID_MULTIPLIER*
+ */arch/powerpc/include/asm/mmu-hash64.h +1
-0+#if defined(CONFIG_POWER6_CPU ) || defined(CONFIG_POWER7_CPU )
arch/powerpc/include/asm/mmu-hash64.h extern void slb_set_size(u16 size); · line 398
Quoted 10 lines
arch/powerpc/include/asm/mmu-hash64.h +10
-0+/* +* Even if cpu support 68 bits, we limit this to 66 because +* we support only 2^19 context. +*/ +#define MAX_VIRTUAL_ADDR_BITS 66 +#else +/* power4,power 5 and cell is 65 */ +#define MAX_VIRTUAL_ADDR_BITS 65 +#endif +
A compile option ? Really ? Ugh...
Ben.
Re: [PATCH 2/2] powerpc: Make context bits depend on virtual addr size.
Quoted 17 lines · Benjamin Herrenschmidt
Quoted 4 lines · Aneesh Kumar K.V
+ * Be careful with this value. This determines the VSID_MODULUS_* and that
+ * need to be co-prime with VSID_MULTIPLIER*
+ */arch/powerpc/include/asm/mmu-hash64.h +1
-0+#if defined(CONFIG_POWER6_CPU ) || defined(CONFIG_POWER7_CPU )arch/powerpc/include/asm/mmu-hash64.h extern void slb_set_size(u16 size); · line 398
Quoted 10 lines
arch/powerpc/include/asm/mmu-hash64.h +10
-0+/* +* Even if cpu support 68 bits, we limit this to 66 because +* we support only 2^19 context. +*/ +#define MAX_VIRTUAL_ADDR_BITS 66 +#else +/* power4,power 5 and cell is 65 */ +#define MAX_VIRTUAL_ADDR_BITS 65 +#endif +A compile option ? Really ? Ugh...
I actually wanted that to be done in Kconfig.cputype, but haven't found a nice way to do it. Considering we are switching between only two values, I was thinking an #ifdef would work.
-aneesh
Re: [PATCH 2/2] powerpc: Make context bits depend on virtual addr size.
Quoted 5 lines · Aneesh Kumar K.V
Quoted 1 line
A compile option ? Really ? Ugh...
I actually wanted that to be done in
Kconfig.cputype, but haven't found a nice way to do it. Considering we are switching between only two values, I was thinking an #ifdef would work.
No, we want to support all those processor types from a single kernel image.
Ben.
Re: [PATCH 2/2] powerpc: Make context bits depend on virtual addr size.
Quoted 8 lines · Benjamin Herrenschmidt
Quoted 5 lines · Aneesh Kumar K.V
Quoted 1 line
A compile option ? Really ? Ugh...
I actually wanted that to be done in
Kconfig.cputype, but haven't found a nice way to do it. Considering we are switching between only two values, I was thinking an #ifdef would work.No, we want to support all those processor types from a single kernel image.
Ok. How about the below patch. This is based on the suggestion from Paulus. I still have to take care of few comments in the code.
We now split the proto-vsid range differently.
User: 0 to [2^(CONTEXT_BITS) - 4 + 2^(USER_ESID_BITS)]
kernel: [2^(CONTEXT_BITS) - 4 + 2^(USER_ESID_BITS)] to 2^(VSID_BITS) - 1
Phileas and Geoff,
Can we check whether this fix the PS3 boot hang ?
-aneesh
arch/powerpc/include/asm/mmu-hash64.h +37 -33
@@ -381,21 +381,33 @@ extern void slb_set_size(u16 size);* hash collisions. */ +#define CONTEXT_BITS 19 +#define USER_ESID_BITS 18 +#define USER_ESID_BITS_1T 6 + +/* +* 256MB segment +* The proto-VSID space has 2^(CONTEX_BITS + USER_ESID_BITS) - 1 segments +* available for user + kernel mapping. The top 4 contexts are used for +* kernel mapping. Each segment contains 2^28 bytes. Each +* context maps 2^46 bytes (64TB) so we can support 2^19-1 contexts +* (19 == 37 + 28 - 46). +*/ +#define MAX_CONTEXT ((ASM_CONST(1) << CONTEXT_BITS) - 1) + +/* * This should be computed such that protovosid * vsid_mulitplier * doesn't overflow 64 bits. It should also be co-prime to vsid_modulus */ #define VSID_MULTIPLIER_256M ASM_CONST(12538073)/* 24-bit prime */ -#define VSID_BITS_256M 38 +#define VSID_BITS_256M (CONTEXT_BITS + USER_ESID_BITS)#define VSID_MODULUS_256M ((1UL<<VSID_BITS_256M)-1)#define VSID_MULTIPLIER_1T ASM_CONST(12538073)/* 24-bit prime */ -#define VSID_BITS_1T 26 +#define VSID_BITS_1T (CONTEXT_BITS + USER_ESID_BITS_1T)#define VSID_MODULUS_1T ((1UL<<VSID_BITS_1T)-1) -#define CONTEXT_BITS 19 -#define USER_ESID_BITS 18 -#define USER_ESID_BITS_1T 6#define USER_VSID_RANGE (1UL << (USER_ESID_BITS + SID_SHIFT))
@@ -513,34 +525,6 @@ typedef struct { })#endif /* 1 */ -/* -* This is only valid for addresses >= PAGE_OFFSET -* The proto-VSID space is divided into two class -* User: 0 to 2^(CONTEXT_BITS + USER_ESID_BITS) -1 -* kernel: 2^(CONTEXT_BITS + USER_ESID_BITS) to 2^(VSID_BITS) - 1 -* -* With KERNEL_START at 0xc000000000000000, the proto vsid for -* the kernel ends up with 0xc00000000 (36 bits). With 64TB -* support we need to have kernel proto-VSID in the -* [2^37 to 2^38 - 1] range due to the increased USER_ESID_BITS. -*/ -static inline unsigned long get_kernel_vsid (unsigned long ea,int ssize) -{ -unsigned long proto_vsid; -/* -* We need to make sure proto_vsid for the kernel is -* >= 2^(CONTEXT_BITS + USER_ESID_BITS[_1T]) -*/ -if (ssize ==MMU_SEGSIZE_256M ) { - proto_vsid = ea >>SID_SHIFT ; - proto_vsid |= (1UL << (CONTEXT_BITS +USER_ESID_BITS )); - return vsid_scramble(proto_vsid, 256M); - } - proto_vsid = ea >>SID_SHIFT_1T ; - proto_vsid |= (1UL << (CONTEXT_BITS +USER_ESID_BITS_1T )); - return vsid_scramble(proto_vsid, 1T); -} -/* Returns the segment size indicator for a user address */ static inline int user_segment_size (unsigned long addr) {
@@ -561,6 +545,26 @@ static inline unsigned long get_vsid(unsigned long context, unsigned long ea, | (ea >> SID_SHIFT_1T), 1T); } +/* +* This is only valid for addresses >= PAGE_OFFSET +* The proto-VSID space is divided into two class +* User: 0 to 2^(CONTEXT_BITS) - 4 + 2^(USER_ESID_BITS) +* kernel: 2^(CONTEXT_BITS) - 4 + 2^(USER_ESID_BITS) to 2^(VSID_BITS) - 1 +* +* With KERNEL_START at 0xc000000000000000, the proto vsid for +* the kernel ends up with 0xc00000000 (36 bits). We use one context +* for 0xc, 0xd, 0xe and 0xf. +* +*/ +static inline unsigned long get_kernel_vsid (unsigned long ea,int ssize) +{ +unsigned long context; +/* +* kernel take the top 4 context from the available range +*/ + context = (MAX_CONTEXT -4 ) + ((ea >>60 ) -0xc ); +return get_vsid (context, ea, ssize); +}#endif /* __ASSEMBLY__ */ #endif /* _ASM_POWERPC_MMU_HASH64_H_ */
arch/powerpc/kernel/exceptions-64s.S +13 -6
@@ -1269,16 +1269,23 @@ _GLOBAL(do_stab_bolted) stw r9,PACA_EXSLB+EX_CCR(r13) /* save CR in exc. frame */ std r11,PACA_EXSLB+EX_SRR0(r13) /* save SRR0 in exc. frame */ + mfspr r11,SPRN_DAR + srdi r9,r11,60 /* ea >> 60 */ + srdi r11,r11,SID_SHIFT + + /* Calculate VSID: + * This is the kernel vsid, we take the top for context from + * the range. context = (MAX_CONTEXT - 4) + ((ea >> 60) - 0xc) + */ + subi r9,r9,(0xc + 4 + 1) + lis r10,8 + add r9,r9,r10 + /* Hash to the primary group */ ld r10,PACASTABVIRT(r13) - mfspr r11,SPRN_DAR - srdi r11,r11,28 rldimi r10,r11,7,52 /* r10 = first ste of the group */ + rldimi r11,r9,USER_ESID_BITS,0 /* proto vsid */ - /* Calculate VSID */ - /* This is a kernel address, so protovsid = ESID | 1 << 37 */ - li r9,0x1 - rldimi r11,r9,(CONTEXT_BITS + USER_ESID_BITS),0 ASM_VSID_SCRAMBLE(r11, r9, 256M) rldic r9,r11,12,16 /* r9 = vsid << 12 */
arch/powerpc/mm/mmu_context_hash64.c +1 -10
@@ -29,15 +29,6 @@static DEFINE_SPINLOCK (mmu_context_lock);static DEFINE_IDA (mmu_context_ida); -/* -* 256MB segment -* The proto-VSID space has 2^(CONTEX_BITS + USER_ESID_BITS) - 1 segments -* available for user mappings. Each segment contains 2^28 bytes. Each -* context maps 2^46 bytes (64TB) so we can support 2^19-1 contexts -* (19 == 37 + 28 - 46). -*/ -#define MAX_CONTEXT ((1UL << CONTEXT_BITS) - 1) -int __init_new_context (void ) {int index;
@@ -56,7 +47,7 @@ again:else if (err)return err; -if (index >MAX_CONTEXT ) { +if (index > (MAX_CONTEXT -4 )) {spin_lock (&mmu_context_lock);ida_remove (&mmu_context_ida, index);spin_unlock (&mmu_context_lock);
arch/powerpc/mm/slb_low.S +18 -4
@@ -56,12 +56,19 @@ _GLOBAL(slb_allocate_realmode) */ _GLOBAL(slb_miss_kernel_load_linear) li r11,0 - li r9,0x1 + /* + * context = (MAX_CONTEXT - 4) + ((ea >> 60) - 0xc) + */ + srdi r9,r3,60 + subi r9,r9,(0xc + 4 + 1) + lis r10, 8 + add r9,r9,r10 + srdi r10,r3,28 /* FIXME!! doing it twice */ /* * for 1T we shift 12 bits more. slb_finish_load_1T will do * the necessary adjustment */ - rldimi r10,r9,(CONTEXT_BITS + USER_ESID_BITS),0 + rldimi r10,r9,USER_ESID_BITS,0 BEGIN_FTR_SECTION b slb_finish_load END_MMU_FTR_SECTION_IFCLR(MMU_FTR_1T_SEGMENT)
@@ -91,12 +98,19 @@ _GLOBAL(slb_miss_kernel_load_vmemmap) _GLOBAL(slb_miss_kernel_load_io) li r11,0 6: - li r9,0x1 + /* + * context = (MAX_CONTEXT - 4) + ((ea >> 60) - 0xc) + */ + srdi r9,r3,60 + subi r9,r9,(0xc + 4 + 1) + lis r10,8 + add r9,r9,r10 + srdi r10,r3,28 /* FIXME!! doing it twice */ /* * for 1T we shift 12 bits more. slb_finish_load_1T will do * the necessary adjustment */ - rldimi r10,r9,(CONTEXT_BITS + USER_ESID_BITS),0 + rldimi r10,r9,USER_ESID_BITS,0 BEGIN_FTR_SECTION b slb_finish_load END_MMU_FTR_SECTION_IFCLR(MMU_FTR_1T_SEGMENT)
RE: [PATCH 2/2] powerpc: Make context bits depend on virtual addr size.
arch/powerpc/include/asm/mmu-hash64.h extern void slb_set_size(u16 size);
Quoted 7 lines
arch/powerpc/include/asm/mmu-hash64.h +7
-0+#define CONTEXT_BITS 19 +#define USER_ESID_BITS 18 +#define USER_ESID_BITS_1T 6 + +/* +* 256MB segment +* The proto-VSID space has 2^(CONTEX_BITS + USER_ESID_BITS) - 1 =
segments
Quoted 1 line
+ * available for user + kernel mapping. The top 4 contexts are used =
for
Quoted 4 lines
+ * kernel mapping. Each segment contains 2^28 bytes. Each
+ * context maps 2^46 bytes (64TB) so we can support 2^19-1 contexts
+ * (19 =3D=3D 37 + 28 - 46).
+ */
I can't help feeling this would be easier to understand if a full 64? 80? bit address was shown with the various bit ranges identified.
Given the comment, I'd have expected CONTEXT_BITS to be calculated from three other named constants - rather than being set to 19.
David
RE: [PATCH 2/2] powerpc: Make context bits depend on virtual addr size.
Quoted 16 lines · David Laight
arch/powerpc/include/asm/mmu-hash64.h extern void slb_set_size(u16 size); · line 395
Quoted 12 lines
arch/powerpc/include/asm/mmu-hash64.h +12
-0+#define CONTEXT_BITS 19 +#define USER_ESID_BITS 18 +#define USER_ESID_BITS_1T 6 + +/* +* 256MB segment +* The proto-VSID space has 2^(CONTEX_BITS + USER_ESID_BITS) - 1 segments +* available for user + kernel mapping. The top 4 contexts are used for +* kernel mapping. Each segment contains 2^28 bytes. Each +* context maps 2^46 bytes (64TB) so we can support 2^19-1 contexts +* (19 == 37 + 28 - 46). +*/ I can't help feeling this would be easier to understand if a full 64? 80? bit address was shown with the various bit ranges identified.
Quoted 3 lines
Given the comment, I'd have expected
CONTEXT_BITSto be calculated from three other named constants - rather than being set to 19.
May be the comments were misleading. We build proto vsid using a combination of context and ea bits.
Current code does the below:
for kernel:proto_vsid = ea >> SID_SHIFT;proto_vsid |= (1UL << (CONTEXT_BITS + USER_ESID_BITS)); for user:proto_vsid = ea >> SID_SHIFTproto_vsid |= context << USER_ESID_BITS
context range is 0 - (2^19 -1)
With this patch we _don't_ give kernel half the proto vsid range. Instead, we reduce the proto vsid range and then the kernel is given top 4 context. ie, kernel proto vsid is now
for kenel:proto_vsid = ea >> SID_SHIFT;
context = (MAX_CONTEXT - 4) + ((effective address >> 60) - 0xc); proto_vsid |= context << USER_ESID_BITS
-aneesh
RE: [PATCH 2/2] powerpc: Make context bits depend on virtual addr size.
Quoted 36 lines · Aneesh Kumar K.V
Quoted 16 lines · David Laight
arch/powerpc/include/asm/mmu-hash64.h extern void slb_set_size(u16 size); · line 395
Quoted 12 lines
arch/powerpc/include/asm/mmu-hash64.h +12
-0+#define CONTEXT_BITS 19 +#define USER_ESID_BITS 18 +#define USER_ESID_BITS_1T 6 + +/* +* 256MB segment +* The proto-VSID space has 2^(CONTEX_BITS + USER_ESID_BITS) - 1 segments +* available for user + kernel mapping. The top 4 contexts are used for +* kernel mapping. Each segment contains 2^28 bytes. Each +* context maps 2^46 bytes (64TB) so we can support 2^19-1 contexts +* (19 == 37 + 28 - 46). +*/ I can't help feeling this would be easier to understand if a full 64? 80? bit address was shown with the various bit ranges identified.
Quoted 3 lines
Given the comment, I'd have expected
CONTEXT_BITSto be calculated from three other named constants - rather than being set to 19.May be the comments were misleading. We build proto vsid using a combination of context and ea bits.
Current code does the below:
for kernel:
proto_vsid= ea >>SID_SHIFT;proto_vsid|= (1UL << (CONTEXT_BITS+USER_ESID_BITS)); for user:proto_vsid= ea >>SID_SHIFTproto_vsid|= context <<USER_ESID_BITScontext range is 0 - (2^19 -1)
Quoted 9 lines
With this patch we _don't_ give kernel half the proto vsid range. Instead, we reduce the proto vsid range and then the kernel is given top 4 context. ie, kernel proto vsid is now
for kenel:
proto_vsid= ea >>SID_SHIFT;
context = (MAX_CONTEXT- 4) + ((effective address >> 60) -0xc);proto_vsid|= context <<USER_ESID_BITS
Hmm that may be an issue, considering ESID for kernel is 36 bits. We have overlapping bits between shifted value of context and kernel ESID.
-aneesh
Re[2]: [PATCH 2/2] powerpc: Make context bits depend on virtual addr size.
In reply to Aneesh Kumar K.VQuoted 5 lines
Ok. How about the below patch. This is based on the suggestion from Paulus. I still have to take care of few comments in the code.
Quoted 7 lines
We now split the proto-vsid range differently.
User: 0 to [2^(CONTEXT_BITS) - 4 + 2^(USER_ESID_BITS)]
kernel: [2^(CONTEXT_BITS) - 4 + 2^(USER_ESID_BITS)] to 2^(VSID_BITS) - 1Phileas and Geoff,
Can we check whether this fix the PS3 boot hang ?
Tested the new patch with Linux 3.8.0-rc7, it seems to be working well.
regards
Re: [PATCH 2/2] powerpc: Make context bits depend on virtual addr size.
In reply to Aneesh Kumar K.VHi Aneesh,
Quoted 6 lines · Aneesh Kumar K.V
Ok. How about the below patch. This is based on the suggestion from Paulus. I still have to take care of few comments in the code.
Phileas and Geoff,
Can we check whether this fix the PS3 boot hang ?
I did a quick build and boot test, and this one seems to work OK.
-Geoff