diff options
author | Aaron Durbin <adurbin@chromium.org> | 2013-03-26 14:09:47 -0500 |
---|---|---|
committer | Stefan Reinauer <stefan.reinauer@coreboot.org> | 2013-03-29 19:53:43 +0100 |
commit | bb4e79a332f0a4f79d402c91b61010157d8a7886 (patch) | |
tree | c886bd2f1d047392dac8dcdae78d7387d79904a3 | |
parent | 28adb6ead6d82073f32e4e786728e27326ccbc6c (diff) | |
download | coreboot-bb4e79a332f0a4f79d402c91b61010157d8a7886.tar.xz |
x86: add new mtrr implementation
The old MTRR code had issues using too many variable
MTRRs depending on the physical address space layout dictated
by the device resources. This new implementation calculates
the default MTRR type by comparing the number of variable MTRRs
used for each type. This avoids the need for IORESOURE_UMA_FB
because in many of those situations setting the default type to WB
frees up the variable MTTRs to set that space to UC.
Additionally, it removes the need for IORESOURCE_IGNORE_MTRR
becuase the new mtrr uses the memrange library which does merging
of resources.
Lastly, the sandybridge gma has its speedup optimization removed
for the graphics memory by writing a pre-determined MTRR index.
That will be fixed in an upcoming patch once write-combining support
is added to the resources.
Slight differences from previous MTRR code:
- The number of reserved OS MTRRs is not a hard limit. It's now advisory
as PAT can be used by the OS to setup the regions to the caching
policy desired.
- The memory types are calculated once by the first CPU to run the code.
After that all other CPUs use that value.
- CONFIG_CACHE_ROM support was dropped. It will be added back in its own
change.
A pathological case that was previously fixed by changing vendor code
to adjust the IO hole location looked like the following:
MTRR: Physical address space:
0x0000000000000000 - 0x00000000000a0000 size 0x000a0000 type 6
0x00000000000a0000 - 0x00000000000c0000 size 0x00020000 type 0
0x00000000000c0000 - 0x00000000ad800000 size 0xad740000 type 6
0x00000000ad800000 - 0x00000000d0000000 size 0x22800000 type 0
0x00000000d0000000 - 0x00000000e0000000 size 0x10000000 type 1
0x00000000e0000000 - 0x0000000100000000 size 0x20000000 type 0
0x0000000100000000 - 0x000000014f600000 size 0x4f600000 type 6
As noted by the output below it's impossible to accomodate those
ranges even with 10 variable MTRRS. However, because the code
can select WB as the default MTRR type it can be done in 6 MTRRs:
MTRR: default type WB/UC MTRR counts: 6/14.
MTRR: WB selected as default type.
MTRR: 0 base 0x00000000ad800000 mask 0x0000007fff800000 type 0
MTRR: 1 base 0x00000000ae000000 mask 0x0000007ffe000000 type 0
MTRR: 2 base 0x00000000b0000000 mask 0x0000007ff0000000 type 0
MTRR: 3 base 0x00000000c0000000 mask 0x0000007ff0000000 type 0
MTRR: 4 base 0x00000000d0000000 mask 0x0000007ff0000000 type 1
MTRR: 5 base 0x00000000e0000000 mask 0x0000007fe0000000 type 0
Change-Id: Idfcc78d9afef9d44c769a676716aae3ff2bd79de
Signed-off-by: Aaron Durbin <adurbin@chromium.org>
Reviewed-on: http://review.coreboot.org/2889
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
-rw-r--r-- | src/cpu/x86/mtrr/mtrr.c | 755 | ||||
-rw-r--r-- | src/include/cpu/x86/mtrr.h | 34 | ||||
-rw-r--r-- | src/northbridge/intel/sandybridge/gma.c | 16 |
3 files changed, 426 insertions, 379 deletions
diff --git a/src/cpu/x86/mtrr/mtrr.c b/src/cpu/x86/mtrr/mtrr.c index 7076a7e213..81ea6da91d 100644 --- a/src/cpu/x86/mtrr/mtrr.c +++ b/src/cpu/x86/mtrr/mtrr.c @@ -4,6 +4,7 @@ * Derived from intel_set_mtrr in intel_subr.c and mtrr.c in linux kernel * * Copyright 2000 Silicon Integrated System Corporation + * Copyright 2013 Google Inc. * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -23,14 +24,9 @@ * Reference: Intel Architecture Software Developer's Manual, Volume 3: System Programming */ -/* - 2005.1 yhlu add NC support to spare mtrrs for 64G memory above installed - 2005.6 Eric add address bit in x86_setup_mtrrs - 2005.6 yhlu split x86_setup_var_mtrrs and x86_setup_fixed_mtrrs, - for AMD, it will not use x86_setup_fixed_mtrrs -*/ - #include <stddef.h> +#include <stdlib.h> +#include <string.h> #include <console/console.h> #include <device/device.h> #include <cpu/x86/msr.h> @@ -39,6 +35,7 @@ #include <cpu/x86/lapic.h> #include <arch/cpu.h> #include <arch/acpi.h> +#include <memrange.h> #if CONFIG_X86_AMD_FIXED_MTRRS #include <cpu/amd/mtrr.h> #define MTRR_FIXED_WRBACK_BITS (MTRR_READ_MEM | MTRR_WRITE_MEM) @@ -46,12 +43,6 @@ #define MTRR_FIXED_WRBACK_BITS 0 #endif -static unsigned int mtrr_msr[] = { - MTRRfix64K_00000_MSR, MTRRfix16K_80000_MSR, MTRRfix16K_A0000_MSR, - MTRRfix4K_C0000_MSR, MTRRfix4K_C8000_MSR, MTRRfix4K_D0000_MSR, MTRRfix4K_D8000_MSR, - MTRRfix4K_E0000_MSR, MTRRfix4K_E8000_MSR, MTRRfix4K_F0000_MSR, MTRRfix4K_F8000_MSR, -}; - /* 2 MTRRS are reserved for the operating system */ #define BIOS_MTRRS 6 #define OS_MTRRS 2 @@ -75,79 +66,20 @@ void enable_fixed_mtrr(void) msr_t msr; msr = rdmsr(MTRRdefType_MSR); - msr.lo |= 0xc00; + msr.lo |= MTRRdefTypeEn | MTRRdefTypeFixEn; wrmsr(MTRRdefType_MSR, msr); } -static void enable_var_mtrr(void) +static void enable_var_mtrr(unsigned char deftype) { msr_t msr; msr = rdmsr(MTRRdefType_MSR); - msr.lo |= MTRRdefTypeEn; + msr.lo &= ~0xff; + msr.lo |= MTRRdefTypeEn | deftype; wrmsr(MTRRdefType_MSR, msr); } -/* setting variable mtrr, comes from linux kernel source */ -void set_var_mtrr( - unsigned int reg, unsigned long basek, unsigned long sizek, - unsigned char type, unsigned address_bits) -{ - msr_t base, mask; - unsigned address_mask_high; - - if (reg >= total_mtrrs) - return; - - // it is recommended that we disable and enable cache when we - // do this. - if (sizek == 0) { - disable_cache(); - - msr_t zero; - zero.lo = zero.hi = 0; - /* The invalid bit is kept in the mask, so we simply clear the - relevant mask register to disable a range. */ - wrmsr (MTRRphysMask_MSR(reg), zero); - - enable_cache(); - return; - } - - - address_mask_high = ((1u << (address_bits - 32u)) - 1u); - - base.hi = basek >> 22; - base.lo = basek << 10; - - if (sizek < 4*1024*1024) { - mask.hi = address_mask_high; - mask.lo = ~((sizek << 10) -1); - } - else { - mask.hi = address_mask_high & (~((sizek >> 22) -1)); - mask.lo = 0; - } - - // it is recommended that we disable and enable cache when we - // do this. - disable_cache(); - - /* Bit 32-35 of MTRRphysMask should be set to 1 */ - base.lo |= type; - mask.lo |= MTRRphysMaskValid; - wrmsr (MTRRphysBase_MSR(reg), base); - wrmsr (MTRRphysMask_MSR(reg), mask); - - enable_cache(); - - printk(BIOS_DEBUG, "Setting variable MTRR %d, base: %4ldMB, range: %4ldMB, type %s\n", - reg, basek >>10, sizek >> 10, - (type==MTRR_TYPE_UNCACHEABLE)?"UC": - ((type==MTRR_TYPE_WRBACK)?"WB":"Other") - ); -} - /* fms: find most sigificant bit set, stolen from Linux Kernel Source. */ static inline unsigned int fms(unsigned int x) { @@ -172,343 +104,458 @@ static inline unsigned int fls(unsigned int x) return r; } -/* setting up variable and fixed mtrr - * - * From Intel Vol. III Section 9.12.4, the Range Size and Base Alignment has some kind of requirement: - * 1. The range size must be 2^N byte for N >= 12 (i.e 4KB minimum). - * 2. The base address must be 2^N aligned, where the N here is equal to the N in previous - * requirement. So a 8K range must be 8K aligned not 4K aligned. - * - * These requirement is meet by "decompositing" the ramsize into Sum(Cn * 2^n, n = [0..N], Cn = [0, 1]). - * For Cm = 1, there is a WB range of 2^m size at base address Sum(Cm * 2^m, m = [N..n]). - * A 124MB (128MB - 4MB SMA) example: - * ramsize = 124MB == 64MB (at 0MB) + 32MB (at 64MB) + 16MB (at 96MB ) + 8MB (at 112MB) + 4MB (120MB). - * But this wastes a lot of MTRR registers so we use another more "aggresive" way with Uncacheable Regions. - * - * In the Uncacheable Region scheme, we try to cover the whole ramsize by one WB region as possible, - * If (an only if) this can not be done we will try to decomposite the ramesize, the mathematical formula - * whould be ramsize = Sum(Cn * 2^n, n = [0..N], Cn = [-1, 0, 1]). For Cn = -1, a Uncachable Region is used. - * The same 124MB example: - * ramsize = 124MB == 128MB WB (at 0MB) + 4MB UC (at 124MB) - * or a 156MB (128MB + 32MB - 4MB SMA) example: - * ramsize = 156MB == 128MB WB (at 0MB) + 32MB WB (at 128MB) + 4MB UC (at 156MB) - */ +#define MTRR_VERBOSE_LEVEL BIOS_NEVER + +/* MTRRs are at a 4KiB granularity. Therefore all address calculations can + * be done with 32-bit numbers. This allows for the MTRR code to handle + * up to 2^44 bytes (16 TiB) of address space. */ +#define RANGE_SHIFT 12 +#define ADDR_SHIFT_TO_RANGE_SHIFT(x) \ + (((x) > RANGE_SHIFT) ? ((x) - RANGE_SHIFT) : RANGE_SHIFT) +#define PHYS_TO_RANGE_ADDR(x) ((x) >> RANGE_SHIFT) +#define RANGE_TO_PHYS_ADDR(x) (((resource_t)(x)) << RANGE_SHIFT) +#define NUM_FIXED_MTRRS (NUM_FIXED_RANGES / RANGES_PER_FIXED_MTRR) + +/* The minimum alignment while handling variable MTRR ranges is 64MiB. */ +#define MTRR_MIN_ALIGN PHYS_TO_RANGE_ADDR(64 << 20) +/* Helpful constants. */ +#define RANGE_1MB PHYS_TO_RANGE_ADDR(1 << 20) +#define RANGE_4GB (1 << (ADDR_SHIFT_TO_RANGE_SHIFT(32))) + +static inline uint32_t range_entry_base_mtrr_addr(struct range_entry *r) +{ + return PHYS_TO_RANGE_ADDR(range_entry_base(r)); +} -static void set_fixed_mtrrs(unsigned int first, unsigned int last, unsigned char type) +static inline uint32_t range_entry_end_mtrr_addr(struct range_entry *r) { - unsigned int i; - unsigned int fixed_msr = NUM_FIXED_RANGES >> 3; - msr_t msr; - msr.lo = msr.hi = 0; /* Shut up gcc */ - for(i = first; i < last; i++) { - /* When I switch to a new msr read it in */ - if (fixed_msr != i >> 3) { - /* But first write out the old msr */ - if (fixed_msr < (NUM_FIXED_RANGES >> 3)) { - disable_cache(); - wrmsr(mtrr_msr[fixed_msr], msr); - enable_cache(); - } - fixed_msr = i>>3; - msr = rdmsr(mtrr_msr[fixed_msr]); - } - if ((i & 7) < 4) { - msr.lo &= ~(0xff << ((i&3)*8)); - msr.lo |= type << ((i&3)*8); - } else { - msr.hi &= ~(0xff << ((i&3)*8)); - msr.hi |= type << ((i&3)*8); - } - } - /* Write out the final msr */ - if (fixed_msr < (NUM_FIXED_RANGES >> 3)) { - disable_cache(); - wrmsr(mtrr_msr[fixed_msr], msr); - enable_cache(); - } + return PHYS_TO_RANGE_ADDR(range_entry_end(r)); } -static unsigned fixed_mtrr_index(unsigned long addrk) +static struct memranges *get_physical_address_space(void) { - unsigned index; - index = (addrk - 0) >> 6; - if (index >= 8) { - index = ((addrk - 8*64) >> 4) + 8; + static struct memranges *addr_space; + static struct memranges addr_space_storage; + + /* In order to handle some chipsets not being able to pre-determine + * uncacheable ranges, such as graphics memory, at resource inseration + * time remove unacheable regions from the cacheable ones. */ + if (addr_space == NULL) { + struct range_entry *r; + const unsigned long mask = IORESOURCE_CACHEABLE; + + addr_space = &addr_space_storage; + + /* Collect cacheable and uncacheable address ranges. The + * uncacheable regions take precedence over the cacheable + * regions. */ + memranges_init(addr_space, mask, mask, MTRR_TYPE_WRBACK); + memranges_add_resources(addr_space, mask, 0, + MTRR_TYPE_UNCACHEABLE); + + /* The address space below 4GiB is special. It needs to be + * covered entirly by range entries so that MTRR calculations + * can be properly done for the full 32-bit address space. + * Therefore, ensure holes are filled up to 4GiB as + * uncacheable */ + memranges_fill_holes_up_to(addr_space, + RANGE_TO_PHYS_ADDR(RANGE_4GB), + MTRR_TYPE_UNCACHEABLE); + + printk(BIOS_DEBUG, "MTRR: Physical address space:\n"); + memranges_each_entry(r, addr_space) + printk(BIOS_DEBUG, + "0x%016llx - 0x%016llx size 0x%08llx type %ld\n", + range_entry_base(r), range_entry_end(r), + range_entry_size(r), range_entry_tag(r)); } - if (index >= 24) { - index = ((addrk - (8*64 + 16*16)) >> 2) + 24; - } - if (index > NUM_FIXED_RANGES) { - index = NUM_FIXED_RANGES; - } - return index; + + return addr_space; } -static unsigned int range_to_mtrr(unsigned int reg, - unsigned long range_startk, unsigned long range_sizek, - unsigned long next_range_startk, unsigned char type, - unsigned int address_bits, unsigned int above4gb) +/* Fixed MTRR descriptor. This structure defines the step size and begin + * and end (exclusive) address covered by a set of fixe MTRR MSRs. + * It also describes the offset in byte intervals to store the calculated MTRR + * type in an array. */ +struct fixed_mtrr_desc { + uint32_t begin; + uint32_t end; + uint32_t step; + int range_index; + int msr_index_base; +}; + +/* Shared MTRR calculations. Can be reused by APs. */ +static uint8_t fixed_mtrr_types[NUM_FIXED_RANGES]; + +/* Fixed MTRR descriptors. */ +static const struct fixed_mtrr_desc fixed_mtrr_desc[] = { + { PHYS_TO_RANGE_ADDR(0x000000), PHYS_TO_RANGE_ADDR(0x080000), + PHYS_TO_RANGE_ADDR(64 * 1024), 0, MTRRfix64K_00000_MSR }, + { PHYS_TO_RANGE_ADDR(0x080000), PHYS_TO_RANGE_ADDR(0x0C0000), + PHYS_TO_RANGE_ADDR(16 * 1024), 8, MTRRfix16K_80000_MSR }, + { PHYS_TO_RANGE_ADDR(0x0C0000), PHYS_TO_RANGE_ADDR(0x100000), + PHYS_TO_RANGE_ADDR(4 * 1024), 24, MTRRfix4K_C0000_MSR }, +}; + +static void calc_fixed_mtrrs(void) { - unsigned long hole_startk = 0, hole_sizek = 0; - - if (!range_sizek) { - /* If there's no MTRR hole, this function will bail out - * here when called for the hole. - */ - printk(BIOS_SPEW, "Zero-sized MTRR range @%ldKB\n", range_startk); - return reg; - } + static int fixed_mtrr_types_initialized; + struct memranges *phys_addr_space; + struct range_entry *r; + const struct fixed_mtrr_desc *desc; + const struct fixed_mtrr_desc *last_desc; + uint32_t begin; + uint32_t end; + int type_index; + + if (fixed_mtrr_types_initialized) + return; - if (reg >= bios_mtrrs) { - printk(BIOS_ERR, "Warning: Out of MTRRs for base: %4ldMB, range: %ldMB, type %s\n", - range_startk >>10, range_sizek >> 10, - (type==MTRR_TYPE_UNCACHEABLE)?"UC": - ((type==MTRR_TYPE_WRBACK)?"WB":"Other") ); - return reg; - } + phys_addr_space = get_physical_address_space(); -#define MIN_ALIGN 0x10000 /* 64MB */ - - if (above4gb == 2 && type == MTRR_TYPE_WRBACK && - range_sizek > MIN_ALIGN && range_sizek % MIN_ALIGN) { - /* - * If this range is not divisible then instead - * make a larger range and carve out an uncached hole. - */ - hole_startk = range_startk + range_sizek; - hole_sizek = MIN_ALIGN - (range_sizek % MIN_ALIGN); - range_sizek += hole_sizek; - } + /* Set all fixed ranges to uncacheable first. */ + memset(&fixed_mtrr_types[0], MTRR_TYPE_UNCACHEABLE, NUM_FIXED_RANGES); - while(range_sizek) { - unsigned long max_align, align; - unsigned long sizek; - /* Compute the maximum size I can make a range */ - max_align = fls(range_startk); - align = fms(range_sizek); - if (align > max_align) { - align = max_align; - } - sizek = 1 << align; - - /* if range is above 4GB, MTRR is needed - * only if above4gb flag is set - */ - if (range_startk < 0x100000000ull / 1024 || above4gb) - set_var_mtrr(reg++, range_startk, sizek, type, address_bits); - range_startk += sizek; - range_sizek -= sizek; - if (reg >= bios_mtrrs) { - printk(BIOS_ERR, "Running out of variable MTRRs!\n"); + desc = &fixed_mtrr_desc[0]; + last_desc = &fixed_mtrr_desc[ARRAY_SIZE(fixed_mtrr_desc) - 1]; + type_index = desc->range_index; + + memranges_each_entry(r, phys_addr_space) { + begin = range_entry_base_mtrr_addr(r); + end = range_entry_end_mtrr_addr(r); + + if (begin >= last_desc->end) break; + + if (end > last_desc->end) + end = last_desc->end; + + /* Get to the correct fixed mtrr descriptor. */ + while (begin >= desc->end) + desc++; + + type_index = desc->range_index; + type_index += (begin - desc->begin) / desc->step; + + while (begin != end) { + unsigned char type; + + type = range_entry_tag(r); + printk(MTRR_VERBOSE_LEVEL, + "MTRR addr 0x%x-0x%x set to %d type @ %d\n", + begin, begin + desc->step, type, type_index); + if (type == MTRR_TYPE_WRBACK) + type |= MTRR_FIXED_WRBACK_BITS; + fixed_mtrr_types[type_index] = type; + type_index++; + begin += desc->step; + if (begin == desc->end) + desc++; } } + fixed_mtrr_types_initialized = 1; +} + +static void commit_fixed_mtrrs(void) +{ + int i; + int j; + int msr_num; + int type_index; + /* 8 ranges per msr. */ + msr_t fixed_msrs[NUM_FIXED_MTRRS]; + unsigned long msr_index[NUM_FIXED_MTRRS]; + + memset(&fixed_msrs, 0, sizeof(fixed_msrs)); - if (hole_sizek) { - printk(BIOS_DEBUG, "Adding hole at %ldMB-%ldMB\n", - hole_startk >> 10, (hole_startk + hole_sizek) >> 10); - reg = range_to_mtrr(reg, hole_startk, hole_sizek, - next_range_startk, MTRR_TYPE_UNCACHEABLE, - address_bits, above4gb); + disable_cache(); + + msr_num = 0; + type_index = 0; + for (i = 0; i < ARRAY_SIZE(fixed_mtrr_desc); i++) { + const struct fixed_mtrr_desc *desc; + int num_ranges; + + desc = &fixed_mtrr_desc[i]; + num_ranges = (desc->end - desc->begin) / desc->step; + for (j = 0; j < num_ranges; j += RANGES_PER_FIXED_MTRR) { + msr_index[msr_num] = desc->msr_index_base + + (j / RANGES_PER_FIXED_MTRR); + fixed_msrs[msr_num].lo |= + fixed_mtrr_types[type_index++] << 0; + fixed_msrs[msr_num].lo |= + fixed_mtrr_types[type_index++] << 8; + fixed_msrs[msr_num].lo |= + fixed_mtrr_types[type_index++] << 16; + fixed_msrs[msr_num].lo |= + fixed_mtrr_types[type_index++] << 24; + fixed_msrs[msr_num].hi |= + fixed_mtrr_types[type_index++] << 0; + fixed_msrs[msr_num].hi |= + fixed_mtrr_types[type_index++] << 8; + fixed_msrs[msr_num].hi |= + fixed_mtrr_types[type_index++] << 16; + fixed_msrs[msr_num].hi |= + fixed_mtrr_types[type_index++] << 24; + msr_num++; + } } - return reg; + for (i = 0; i < ARRAY_SIZE(fixed_msrs); i++) { + printk(BIOS_DEBUG, "MTRR: Fixed MSR 0x%lx 0x%08x%08x\n", + msr_index[i], fixed_msrs[i].hi, fixed_msrs[i].lo); + wrmsr(msr_index[i], fixed_msrs[i]); + } + + enable_cache(); } -static unsigned long resk(uint64_t value) +void x86_setup_fixed_mtrrs_no_enable(void) { - unsigned long resultk; - if (value < (1ULL << 42)) { - resultk = value >> 10; - } - else { - resultk = 0xffffffff; - } - return resultk; + calc_fixed_mtrrs(); + commit_fixed_mtrrs(); } -static void set_fixed_mtrr_resource(void *gp, struct device *dev, struct resource *res) +void x86_setup_fixed_mtrrs(void) { - unsigned int start_mtrr; - unsigned int last_mtrr; - const unsigned char type = MTRR_TYPE_WRBACK | MTRR_FIXED_WRBACK_BITS; - start_mtrr = fixed_mtrr_index(resk(res->base)); - last_mtrr = fixed_mtrr_index(resk((res->base + res->size))); - if (start_mtrr >= NUM_FIXED_RANGES) { - return; - } - printk(BIOS_DEBUG, "Setting fixed MTRRs(%d-%d) Type: WB\n", - start_mtrr, last_mtrr); - set_fixed_mtrrs(start_mtrr, last_mtrr, type); + x86_setup_fixed_mtrrs_no_enable(); + printk(BIOS_SPEW, "call enable_fixed_mtrr()\n"); + enable_fixed_mtrr(); } struct var_mtrr_state { - unsigned long range_startk, range_sizek; - unsigned int reg; - unsigned long hole_startk, hole_sizek; - unsigned int address_bits; - unsigned int above4gb; /* Set if MTRRs are needed for DRAM above 4GB */ + struct memranges *addr_space; + int above4gb; + int address_bits; + int commit_mtrrs; + int mtrr_index; + int def_mtrr_type; }; -void set_var_mtrr_resource(void *gp, struct device *dev, struct resource *res) +static void clear_var_mtrr(int index) { - struct var_mtrr_state *state = gp; - unsigned long basek, sizek; - if (state->reg >= bios_mtrrs) - return; + msr_t msr_val; - basek = resk(res->base); - sizek = resk(res->size); + msr_val = rdmsr(MTRRphysMask_MSR(index)); + msr_val.lo &= ~MTRRphysMaskValid; + wrmsr(MTRRphysMask_MSR(index), msr_val); +} - if (res->flags & IORESOURCE_UMA_FB) { - /* FIXME: could I use Write-Combining for Frame Buffer ? */ - state->reg = range_to_mtrr(state->reg, basek, sizek, 0, - MTRR_TYPE_UNCACHEABLE, state->address_bits, state->above4gb); +static void write_var_mtrr(struct var_mtrr_state *var_state, + uint32_t base, uint32_t size, int mtrr_type) +{ + msr_t msr_val; + unsigned long msr_index; + resource_t rbase; + resource_t rsize; + resource_t mask; + + /* Some variable MTRRs are attempted to be saved for the OS use. + * However, it's more important to try to map the full address space + * properly. */ + if (var_state->mtrr_index >= bios_mtrrs) + printk(BIOS_WARNING, "Taking a reserved OS MTRR.\n"); + if (var_state->mtrr_index >= total_mtrrs) { + printk(BIOS_ERR, "ERROR: Not enough MTTRs available!\n"); return; } - if (res->flags & IORESOURCE_IGNORE_MTRR) { - return; - } + rbase = base; + rsize = size; - if (!(res->flags & IORESOURCE_CACHEABLE)) - return; + rbase = RANGE_TO_PHYS_ADDR(rbase); + rsize = RANGE_TO_PHYS_ADDR(rsize); + rsize = -rsize; - /* See if I can merge with the last range - * Either I am below 1M and the fixed mtrrs handle it, or - * the ranges touch. - */ - if ((basek <= 1024) || (state->range_startk + state->range_sizek == basek)) { - unsigned long endk = basek + sizek; - state->range_sizek = endk - state->range_startk; - return; - } - /* Write the range mtrrs */ - if (state->range_sizek != 0) { - if (state->hole_sizek == 0 && state->above4gb != 2) { - /* We need to put that on to hole */ - unsigned long endk = basek + sizek; - state->hole_startk = state->range_startk + state->range_sizek; - state->hole_sizek = basek - state->hole_startk; - state->range_sizek = endk - state->range_startk; - return; - } - state->reg = range_to_mtrr(state->reg, state->range_startk, - state->range_sizek, basek, MTRR_TYPE_WRBACK, - state->address_bits, state->above4gb); - - state->reg = range_to_mtrr(state->reg, state->hole_startk, - state->hole_sizek, basek, MTRR_TYPE_UNCACHEABLE, - state->address_bits, state->above4gb); - - state->range_startk = 0; - state->range_sizek = 0; - state->hole_startk = 0; - state->hole_sizek = 0; - } - /* Allocate an msr */ - printk(BIOS_SPEW, " Allocate an msr - basek = %08lx, sizek = %08lx,\n", basek, sizek); - state->range_startk = basek; - state->range_sizek = sizek; + mask = (1ULL << var_state->address_bits) - 1; + rsize = rsize & mask; + + printk(BIOS_DEBUG, "MTRR: %d base 0x%016llx mask 0x%016llx type %d\n", + var_state->mtrr_index, rbase, rsize, mtrr_type); + + msr_val.lo = rbase; + msr_val.lo |= mtrr_type; + + msr_val.hi = rbase >> 32; + msr_index = MTRRphysBase_MSR(var_state->mtrr_index); + wrmsr(msr_index, msr_val); + + msr_val.lo = rsize; + msr_val.lo |= MTRRphysMaskValid; + msr_val.hi = rsize >> 32; + msr_index = MTRRphysMask_MSR(var_state->mtrr_index); + wrmsr(msr_index, msr_val); } -void x86_setup_fixed_mtrrs_no_enable(void) +static void calc_var_mtrr_range(struct var_mtrr_state *var_state, + uint32_t base, uint32_t size, int mtrr_type) { - /* Try this the simple way of incrementally adding together - * mtrrs. If this doesn't work out we can get smart again - * and clear out the mtrrs. - */ - - printk(BIOS_DEBUG, "\n"); - /* Initialized the fixed_mtrrs to uncached */ - printk(BIOS_DEBUG, "Setting fixed MTRRs(%d-%d) Type: UC\n", - 0, NUM_FIXED_RANGES); - set_fixed_mtrrs(0, NUM_FIXED_RANGES, MTRR_TYPE_UNCACHEABLE); - - /* Now see which of the fixed mtrrs cover ram. - */ - search_global_resources( - IORESOURCE_MEM | IORESOURCE_CACHEABLE, IORESOURCE_MEM | IORESOURCE_CACHEABLE, - set_fixed_mtrr_resource, NULL); - printk(BIOS_DEBUG, "DONE fixed MTRRs\n"); + while (size != 0) { + uint32_t addr_lsb; + uint32_t size_msb; + uint32_t mtrr_size; + + addr_lsb = fls(base); + size_msb = fms(size); + + /* All MTRR entries need to have their base aligned to the mask + * size. The maximum size is calculated by a function of the + * min base bit set and maximum size bit set. */ + if (addr_lsb > size_msb) + mtrr_size = 1 << size_msb; + else + mtrr_size = 1 << addr_lsb; + + if (var_state->commit_mtrrs) + write_var_mtrr(var_state, base, mtrr_size, mtrr_type); + + size -= mtrr_size; + base += mtrr_size; + var_state->mtrr_index++; + } } -void x86_setup_fixed_mtrrs(void) +static void setup_var_mtrrs_by_state(struct var_mtrr_state *var_state) { - x86_setup_fixed_mtrrs_no_enable(); + struct range_entry *r; + + /* + * For each range that meets the non-default type process it in the + * following manner: + * +------------------+ c2 = end + * | 0 or more bytes | + * +------------------+ b2 = c1 = ALIGN_DOWN(end) + * | | + * +------------------+ b1 = a2 = ALIGN_UP(begin) + * | 0 or more bytes | + * +------------------+ a1 = begin + * + * Thus, there are 3 sub-ranges to configure variable MTRRs for. + */ + memranges_each_entry(r, var_state->addr_space) { + uint32_t a1, a2, b1, b2, c1, c2; + int mtrr_type = range_entry_tag(r); + + /* Skip default type. */ + if (var_state->def_mtrr_type == mtrr_type) + continue; + + a1 = range_entry_base_mtrr_addr(r); + c2 = range_entry_end_mtrr_addr(r); + + /* The end address is under 1MiB. The fixed MTRRs take + * precedence over the variable ones. Therefore this range + * can be ignored. */ + if (c2 < RANGE_1MB) + continue; + + /* Again, the fixed MTRRs take precedence so the beginning + * of the range can be set to 0 if it starts below 1MiB. */ + if (a1 < RANGE_1MB) + a1 = 0; + + /* If the range starts above 4GiB the processing is done. */ + if (!var_state->above4gb && a1 >= RANGE_4GB) + break; - printk(BIOS_SPEW, "call enable_fixed_mtrr()\n"); - enable_fixed_mtrr(); -} + /* Clip the upper address to 4GiB if addresses above 4GiB + * are not being processed. */ + if (!var_state->above4gb && c2 > RANGE_4GB) + c2 = RANGE_4GB; + /* Don't align up or down on the range if it is smaller + * than the minimum granularity. */ + if ((c2 - a1) < MTRR_MIN_ALIGN) { + calc_var_mtrr_range(var_state, a1, c2 - a1, mtrr_type); + continue; + } -void x86_setup_var_mtrrs(unsigned int address_bits, unsigned int above4gb) -/* this routine needs to know how many address bits a given processor - * supports. CPUs get grumpy when you set too many bits in - * their mtrr registers :( I would generically call cpuid here - * and find out how many physically supported but some cpus are - * buggy, and report more bits then they actually support. - * If above4gb flag is set, variable MTRR ranges must be used to - * set cacheability of DRAM above 4GB. If above4gb flag is clear, - * some other mechanism is controlling cacheability of DRAM above 4GB. - */ + b1 = a2 = ALIGN_UP(a1, MTRR_MIN_ALIGN); + b2 = c1 = ALIGN_DOWN(c2, MTRR_MIN_ALIGN); + + calc_var_mtrr_range(var_state, a1, a2 - a1, mtrr_type); + calc_var_mtrr_range(var_state, b1, b2 - b1, mtrr_type); + calc_var_mtrr_range(var_state, c1, c2 - c1, mtrr_type); + } +} + +static int calc_var_mtrrs(struct memranges *addr_space, + int above4gb, int address_bits) { - /* Try this the simple way of incrementally adding together - * mtrrs. If this doesn't work out we can get smart again - * and clear out the mtrrs. - */ + int wb_deftype_count; + int uc_deftype_count; struct var_mtrr_state var_state; - /* Cache as many memory areas as possible */ - /* FIXME is there an algorithm for computing the optimal set of mtrrs? - * In some cases it is definitely possible to do better. - */ - var_state.range_startk = 0; - var_state.range_sizek = 0; - var_state.hole_startk = 0; - var_state.hole_sizek = 0; - var_state.reg = 0; - var_state.address_bits = address_bits; + /* The default MTRR cacheability type is determined by calculating + * the number of MTTRs required for each MTTR type as if it was the + * default. */ + var_state.addr_space = addr_space; var_state.above4gb = above4gb; + var_state.address_bits = address_bits; + var_state.commit_mtrrs = 0; - /* Detect number of variable MTRRs */ - if (above4gb == 2) - detect_var_mtrrs(); - - search_global_resources(IORESOURCE_MEM, IORESOURCE_MEM, - set_var_mtrr_resource, &var_state); + var_state.mtrr_index = 0; + var_state.def_mtrr_type = MTRR_TYPE_WRBACK; + setup_var_mtrrs_by_state(&var_state); + wb_deftype_count = var_state.mtrr_index; - /* Write the last range */ - var_state.reg = range_to_mtrr(var_state.reg, var_state.range_startk, - var_state.range_sizek, 0, MTRR_TYPE_WRBACK, - var_state.address_bits, var_state.above4gb); + var_state.mtrr_index = 0; + var_state.def_mtrr_type = MTRR_TYPE_UNCACHEABLE; + setup_var_mtrrs_by_state(&var_state); + uc_deftype_count = var_state.mtrr_index; - var_state.reg = range_to_mtrr(var_state.reg, var_state.hole_startk, - var_state.hole_sizek, 0, MTRR_TYPE_UNCACHEABLE, - var_state.address_bits, var_state.above4gb); + printk(BIOS_DEBUG, "MTRR: default type WB/UC MTRR counts: %d/%d.\n", + wb_deftype_count, uc_deftype_count); - printk(BIOS_DEBUG, "DONE variable MTRRs\n"); - printk(BIOS_DEBUG, "Clear out the extra MTRR's\n"); - /* Clear out the extra MTRR's */ - while(var_state.reg < total_mtrrs) { - set_var_mtrr(var_state.reg++, 0, 0, 0, var_state.address_bits); + if (wb_deftype_count < uc_deftype_count) { + printk(BIOS_DEBUG, "MTRR: WB selected as default type.\n"); + return MTRR_TYPE_WRBACK; } + printk(BIOS_DEBUG, "MTRR: UC selected as default type.\n"); + return MTRR_TYPE_UNCACHEABLE; +} -#if CONFIG_CACHE_ROM - /* Enable Caching and speculative Reads for the - * complete ROM now that we actually have RAM. - */ - if (boot_cpu() && (acpi_slp_type != 3)) { - set_var_mtrr(total_mtrrs - 1, (4096 - 8)*1024, 8 * 1024, - MTRR_TYPE_WRPROT, address_bits); +static void commit_var_mtrrs(struct memranges *addr_space, int def_type, + int above4gb, int address_bits) +{ + struct var_mtrr_state var_state; + int i; + + var_state.addr_space = addr_space; + var_state.above4gb = above4gb; + var_state.address_bits = address_bits; + /* Write the MSRs. */ + var_state.commit_mtrrs = 1; + var_state.mtrr_index = 0; + var_state.def_mtrr_type = def_type; + setup_var_mtrrs_by_state(&var_state); + + /* Clear all remaining variable MTTRs. */ + for (i = var_state.mtrr_index; i < total_mtrrs; i++) + clear_var_mtrr(i); +} + +void x86_setup_var_mtrrs(unsigned int address_bits, unsigned int above4gb) +{ + static int mtrr_default_type = -1; + struct memranges *addr_space; + + addr_space = get_physical_address_space(); + + if (mtrr_default_type == -1) { + if (above4gb == 2) + detect_var_mtrrs(); + mtrr_default_type = + calc_var_mtrrs(addr_space, !!above4gb, address_bits); } -#endif - printk(BIOS_SPEW, "call enable_var_mtrr()\n"); - enable_var_mtrr(); - printk(BIOS_SPEW, "Leave %s\n", __func__); - post_code(0x6A); + disable_cache(); + commit_var_mtrrs(addr_space, mtrr_default_type, !!above4gb, + address_bits); + enable_var_mtrr(mtrr_default_type); + enable_cache(); } void x86_setup_mtrrs(void) diff --git a/src/include/cpu/x86/mtrr.h b/src/include/cpu/x86/mtrr.h index fe85ad0a83..618a93c44e 100644 --- a/src/include/cpu/x86/mtrr.h +++ b/src/include/cpu/x86/mtrr.h @@ -26,6 +26,7 @@ #define MTRRphysMaskValid (1 << 11) #define NUM_FIXED_RANGES 88 +#define RANGES_PER_FIXED_MTRR 8 #define MTRRfix64K_00000_MSR 0x250 #define MTRRfix16K_80000_MSR 0x258 #define MTRRfix16K_A0000_MSR 0x259 @@ -39,22 +40,33 @@ #define MTRRfix4K_F8000_MSR 0x26f #if !defined (__ASSEMBLER__) && !defined(__PRE_RAM__) -#include <device/device.h> -/* You should almost NEVER use this function. - * N.B. We worked on a lot of ways to make this continue as static, - * but just making it available ended up being the simplest solution. + +/* + * The MTRR code has some side effects that the callers should be aware for. + * 1. The call sequence matters. x86_setup_mtrrs() calls + * x86_setup_fixed_mtrrs_no_enable() then enable_fixed_mtrrs() (equivalent + * of x86_setup_fixed_mtrrs()) then x86_setup_var_mtrrs(). If the callers + * want to call the components of x86_setup_mtrrs() because of other + * rquirements the ordering should still preserved. + * 2. enable_fixed_mtrr() will enable both variable and fixed MTRRs because + * of the nature of the global MTRR enable flag. Therefore, all direct + * or indirect callers of enable_fixed_mtrr() should ensure that the + * variable MTRR MSRs do not contain bad ranges. */ -void set_var_mtrr( - unsigned int reg, unsigned long basek, unsigned long sizek, - unsigned char type, unsigned address_bits); -void enable_fixed_mtrr(void); -void x86_setup_var_mtrrs(unsigned int address_bits, unsigned int above4gb); void x86_setup_mtrrs(void); -int x86_mtrr_check(void); -void set_var_mtrr_resource(void *gp, struct device *dev, struct resource *res); +/* + * x86_setup_var_mtrrs() parameters: + * address_bits - number of physical address bits supported by cpu + * above4gb - 2 means dynamically detect number of variable MTRRs available. + * non-zero means handle memory ranges above 4GiB. + * 0 means ignore memory ranges above 4GiB + */ +void x86_setup_var_mtrrs(unsigned int address_bits, unsigned int above4gb); +void enable_fixed_mtrr(void); void x86_setup_fixed_mtrrs(void); /* Set up fixed MTRRs but do not enable them. */ void x86_setup_fixed_mtrrs_no_enable(void); +int x86_mtrr_check(void); #endif #if !defined(CONFIG_RAMTOP) diff --git a/src/northbridge/intel/sandybridge/gma.c b/src/northbridge/intel/sandybridge/gma.c index 4a043ebb35..b9a07a2d72 100644 --- a/src/northbridge/intel/sandybridge/gma.c +++ b/src/northbridge/intel/sandybridge/gma.c @@ -622,25 +622,12 @@ static void gma_pm_init_post_vbios(struct device *dev) static void gma_func0_init(struct device *dev) { u32 reg32; - u32 graphics_base, graphics_size; /* IGD needs to be Bus Master */ reg32 = pci_read_config32(dev, PCI_COMMAND); reg32 |= PCI_COMMAND_MASTER | PCI_COMMAND_MEMORY | PCI_COMMAND_IO; pci_write_config32(dev, PCI_COMMAND, reg32); - /* Set up an MTRR for the graphics memory BAR to vastly improve - * speed of VGA initialization (and later access). To stay out of - * the way of the MTRR init code, we are using MTRR #8 to cover - * that range. - */ - graphics_base = dev->resource_list[1].base; - graphics_size = dev->resource_list[1].size; - printk(BIOS_DEBUG, "Setting up MTRR for graphics 0x%08x (%dK)\n", - graphics_base, graphics_size / 1024); - set_var_mtrr(8, graphics_base >> 10, graphics_size >> 10, - MTRR_TYPE_WRCOMB, 0x24); - /* Init graphics power management */ gma_pm_init_pre_vbios(dev); @@ -655,10 +642,11 @@ static void gma_func0_init(struct device *dev) #if CONFIG_MAINBOARD_DO_NATIVE_VGA_INIT /* This should probably run before post VBIOS init. */ printk(BIOS_SPEW, "Initializing VGA without OPROM.\n"); - u32 iobase, mmiobase, physbase; + u32 iobase, mmiobase, physbase, graphics_base; iobase = dev->resource_list[2].base; mmiobase = dev->resource_list[0].base; physbase = pci_read_config32(dev, 0x5c) & ~0xf; + graphics_base = dev->resource_list[1].base; int i915lightup(u32 physbase, u32 iobase, u32 mmiobase, u32 gfx); i915lightup(physbase, iobase, mmiobase, graphics_base); |