Submitter | Aneesh Kumar K.V |
---|---|

Date | Sept. 6, 2012, 3:29 p.m. |

Message ID | <1346945351-7672-12-git-send-email-aneesh.kumar@linux.vnet.ibm.com> |

Download | mbox | patch |

Permalink | /patch/182225/ |

State | Changes Requested |

Delegated to: | Benjamin Herrenschmidt |

Headers | show |

## Comments

## Patch

diff --git a/arch/powerpc/include/asm/mmu-hash64.h b/arch/powerpc/include/asm/mmu-hash64.h index de9cfed..428f23e 100644 --- a/arch/powerpc/include/asm/mmu-hash64.h +++ b/arch/powerpc/include/asm/mmu-hash64.h @@ -324,51 +324,45 @@ extern void slb_set_size(u16 size); #endif /* __ASSEMBLY__ */ /* - * VSID allocation + * VSID allocation (256MB segment) * - * We first generate a 36-bit "proto-VSID". For kernel addresses this - * is equal to the ESID, for user addresses it is: - * (context << 15) | (esid & 0x7fff) + * We first generate a 38-bit "proto-VSID". For kernel addresses this + * is equal to the ESID | 1 << 37, for user addresses it is: + * (context << USER_ESID_BITS) | (esid & ((1U << USER_ESID_BITS) - 1) * - * The two forms are distinguishable because the top bit is 0 for user - * addresses, whereas the top two bits are 1 for kernel addresses. - * Proto-VSIDs with the top two bits equal to 0b10 are reserved for - * now. + * This splits the proto-VSID into the below range + * 0 - (2^(CONTEXT_BITS + USER_ESID_BITS) - 1) : User proto-VSID range + * 2^(CONTEXT_BITS + USER_ESID_BITS) - 2^(VSID_BITS) : Kernel proto-VSID range + * + * We also have CONTEXT_BITS + USER_ESID_BITS = VSID_BITS - 1 + * That is, we assign half of the space to user processes and half + * to the kernel. * * The proto-VSIDs are then scrambled into real VSIDs with the * multiplicative hash: * * VSID = (proto-VSID * VSID_MULTIPLIER) % VSID_MODULUS - * where VSID_MULTIPLIER = 268435399 = 0xFFFFFC7 - * VSID_MODULUS = 2^36-1 = 0xFFFFFFFFF * - * This scramble is only well defined for proto-VSIDs below - * 0xFFFFFFFFF, so both proto-VSID and actual VSID 0xFFFFFFFFF are - * reserved. VSID_MULTIPLIER is prime, so in particular it is + * VSID_MULTIPLIER is prime, so in particular it is * co-prime to VSID_MODULUS, making this a 1:1 scrambling function. * Because the modulus is 2^n-1 we can compute it efficiently without * a divide or extra multiply (see below). * * This scheme has several advantages over older methods: * - * - We have VSIDs allocated for every kernel address + * - We have VSIDs allocated for every kernel address * (i.e. everything above 0xC000000000000000), except the very top * segment, which simplifies several things. * - * - We allow for 16 significant bits of ESID and 19 bits of - * context for user addresses. i.e. 16T (44 bits) of address space for - * up to half a million contexts. + * - We allow for USER_ESID_BITS significant bits of ESID and + * CONTEXT_BITS bits of context for user addresses. + * i.e. 64T (46 bits) of address space for up to half a million contexts. * - * - The scramble function gives robust scattering in the hash + * - The scramble function gives robust scattering in the hash * table (at least based on some initial results). The previous * method was more susceptible to pathological cases giving excessive * hash collisions. */ -/* - * WARNING - If you change these you must make sure the asm - * implementations in slb_allocate (slb_low.S), do_stab_bolted - * (head.S) and ASM_VSID_SCRAMBLE (below) are changed accordingly. - */ /* * This should be computed such that protovosid * vsid_mulitplier diff --git a/arch/powerpc/mm/mmu_context_hash64.c b/arch/powerpc/mm/mmu_context_hash64.c index daa076c..40bc5b0 100644 --- a/arch/powerpc/mm/mmu_context_hash64.c +++ b/arch/powerpc/mm/mmu_context_hash64.c @@ -30,9 +30,11 @@ static DEFINE_SPINLOCK(mmu_context_lock); static DEFINE_IDA(mmu_context_ida); /* - * The proto-VSID space has 2^35 - 1 segments available for user mappings. - * Each segment contains 2^28 bytes. Each context maps 2^44 bytes, - * so we can support 2^19-1 contexts (19 == 35 + 28 - 44). + * 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)