![Benjamin Herrenschmidt](/assets/img/avatar_default.png)
The basic issue is to be able to do what hugetlbfs does but with different page sizes for some other special filesystems; more specifically, my need is: - Huge pages - SPE local store mappings using 64K pages on a 4K base page size kernel on Cell - Some special 4K segments in 64K-page kernels for mapping a dodgy type of powerpc-specific infiniband hardware that requires 4K MMU mappings for various reasons I won't explain here. The main issues are: - To maintain/keep track of the page size per "segment" (as we can only have one page size per segment on powerpc, which are 256MB divisions of the address space). - To make sure special mappings stay within their allotted "segments" (including MAP_FIXED crap) - To make sure everybody else doesn't mmap/brk/grow_stack into a "segment" that is used for a special mapping Some of the necessary mechanisms to handle that were present in the hugetlbfs code, but mostly in ways not suitable for anything else. The patch relies on some changes to the generic get_unmapped_area() that just got merged. It still hijacks hugetlb callbacks here or there as the generic code hasn't been entirely cleaned up yet but that shouldn't be a problem. So what is a slice ? Well, I re-used the mechanism used formerly by our hugetlbfs implementation which divides the address space in "meta-segments" which I called "slices". The division is done using 256MB slices below 4G, and 1T slices above. Thus the address space is divided currently into 16 "low" slices and 16 "high" slices. (Special case: high slice 0 is the area between 4G and 1T). Doing so simplifies significantly the tracking of segments and avoids having to keep track of all the 256MB segments in the address space. While I used the "concepts" of hugetlbfs, I mostly re-implemented everything in a more generic way and "ported" hugetlbfs to it. Slices can have an associated page size, which is encoded in the mmu context and used by the SLB miss handler to set the segment sizes. The hash code currently doesn't care, it has a specific check for hugepages, though I might add a mechanism to provide per-slice hash mapping functions in the future. The slice code provide a pair of "generic" get_unmapped_area() (bottomup and topdown) functions that should work with any slice size. There is some trickiness here so I would appreciate people to have a look at the implementation of these and let me know if I got something wrong. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
262 lines
7.1 KiB
ArmAsm
262 lines
7.1 KiB
ArmAsm
/*
|
|
* Low-level SLB routines
|
|
*
|
|
* Copyright (C) 2004 David Gibson <dwg@au.ibm.com>, IBM
|
|
*
|
|
* Based on earlier C version:
|
|
* Dave Engebretsen and Mike Corrigan {engebret|mikejc}@us.ibm.com
|
|
* Copyright (c) 2001 Dave Engebretsen
|
|
* Copyright (C) 2002 Anton Blanchard <anton@au.ibm.com>, IBM
|
|
*
|
|
* 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 the Free Software Foundation; either version
|
|
* 2 of the License, or (at your option) any later version.
|
|
*/
|
|
|
|
#include <asm/processor.h>
|
|
#include <asm/ppc_asm.h>
|
|
#include <asm/asm-offsets.h>
|
|
#include <asm/cputable.h>
|
|
#include <asm/page.h>
|
|
#include <asm/mmu.h>
|
|
#include <asm/pgtable.h>
|
|
#include <asm/firmware.h>
|
|
|
|
/* void slb_allocate_realmode(unsigned long ea);
|
|
*
|
|
* Create an SLB entry for the given EA (user or kernel).
|
|
* r3 = faulting address, r13 = PACA
|
|
* r9, r10, r11 are clobbered by this function
|
|
* No other registers are examined or changed.
|
|
*/
|
|
_GLOBAL(slb_allocate_realmode)
|
|
/* r3 = faulting address */
|
|
|
|
srdi r9,r3,60 /* get region */
|
|
srdi r10,r3,28 /* get esid */
|
|
cmpldi cr7,r9,0xc /* cmp PAGE_OFFSET for later use */
|
|
|
|
/* r3 = address, r10 = esid, cr7 = <> PAGE_OFFSET */
|
|
blt cr7,0f /* user or kernel? */
|
|
|
|
/* kernel address: proto-VSID = ESID */
|
|
/* WARNING - MAGIC: we don't use the VSID 0xfffffffff, but
|
|
* this code will generate the protoVSID 0xfffffffff for the
|
|
* top segment. That's ok, the scramble below will translate
|
|
* it to VSID 0, which is reserved as a bad VSID - one which
|
|
* will never have any pages in it. */
|
|
|
|
/* Check if hitting the linear mapping of the vmalloc/ioremap
|
|
* kernel space
|
|
*/
|
|
bne cr7,1f
|
|
|
|
/* Linear mapping encoding bits, the "li" instruction below will
|
|
* be patched by the kernel at boot
|
|
*/
|
|
_GLOBAL(slb_miss_kernel_load_linear)
|
|
li r11,0
|
|
b slb_finish_load
|
|
|
|
1: /* vmalloc/ioremap mapping encoding bits, the "li" instructions below
|
|
* will be patched by the kernel at boot
|
|
*/
|
|
BEGIN_FTR_SECTION
|
|
/* check whether this is in vmalloc or ioremap space */
|
|
clrldi r11,r10,48
|
|
cmpldi r11,(VMALLOC_SIZE >> 28) - 1
|
|
bgt 5f
|
|
lhz r11,PACAVMALLOCSLLP(r13)
|
|
b slb_finish_load
|
|
5:
|
|
END_FTR_SECTION_IFCLR(CPU_FTR_CI_LARGE_PAGE)
|
|
_GLOBAL(slb_miss_kernel_load_io)
|
|
li r11,0
|
|
b slb_finish_load
|
|
|
|
|
|
0: /* user address: proto-VSID = context << 15 | ESID. First check
|
|
* if the address is within the boundaries of the user region
|
|
*/
|
|
srdi. r9,r10,USER_ESID_BITS
|
|
bne- 8f /* invalid ea bits set */
|
|
|
|
|
|
/* when using slices, we extract the psize off the slice bitmaps
|
|
* and then we need to get the sllp encoding off the mmu_psize_defs
|
|
* array.
|
|
*
|
|
* XXX This is a bit inefficient especially for the normal case,
|
|
* so we should try to implement a fast path for the standard page
|
|
* size using the old sllp value so we avoid the array. We cannot
|
|
* really do dynamic patching unfortunately as processes might flip
|
|
* between 4k and 64k standard page size
|
|
*/
|
|
#ifdef CONFIG_PPC_MM_SLICES
|
|
cmpldi r10,16
|
|
|
|
/* Get the slice index * 4 in r11 and matching slice size mask in r9 */
|
|
ld r9,PACALOWSLICESPSIZE(r13)
|
|
sldi r11,r10,2
|
|
blt 5f
|
|
ld r9,PACAHIGHSLICEPSIZE(r13)
|
|
srdi r11,r10,(SLICE_HIGH_SHIFT - SLICE_LOW_SHIFT - 2)
|
|
andi. r11,r11,0x3c
|
|
|
|
5: /* Extract the psize and multiply to get an array offset */
|
|
srd r9,r9,r11
|
|
andi. r9,r9,0xf
|
|
mulli r9,r9,MMUPSIZEDEFSIZE
|
|
|
|
/* Now get to the array and obtain the sllp
|
|
*/
|
|
ld r11,PACATOC(r13)
|
|
ld r11,mmu_psize_defs@got(r11)
|
|
add r11,r11,r9
|
|
ld r11,MMUPSIZESLLP(r11)
|
|
ori r11,r11,SLB_VSID_USER
|
|
#else
|
|
/* paca context sllp already contains the SLB_VSID_USER bits */
|
|
lhz r11,PACACONTEXTSLLP(r13)
|
|
#endif /* CONFIG_PPC_MM_SLICES */
|
|
|
|
ld r9,PACACONTEXTID(r13)
|
|
rldimi r10,r9,USER_ESID_BITS,0
|
|
b slb_finish_load
|
|
|
|
8: /* invalid EA */
|
|
li r10,0 /* BAD_VSID */
|
|
li r11,SLB_VSID_USER /* flags don't much matter */
|
|
b slb_finish_load
|
|
|
|
#ifdef __DISABLED__
|
|
|
|
/* void slb_allocate_user(unsigned long ea);
|
|
*
|
|
* Create an SLB entry for the given EA (user or kernel).
|
|
* r3 = faulting address, r13 = PACA
|
|
* r9, r10, r11 are clobbered by this function
|
|
* No other registers are examined or changed.
|
|
*
|
|
* It is called with translation enabled in order to be able to walk the
|
|
* page tables. This is not currently used.
|
|
*/
|
|
_GLOBAL(slb_allocate_user)
|
|
/* r3 = faulting address */
|
|
srdi r10,r3,28 /* get esid */
|
|
|
|
crset 4*cr7+lt /* set "user" flag for later */
|
|
|
|
/* check if we fit in the range covered by the pagetables*/
|
|
srdi. r9,r3,PGTABLE_EADDR_SIZE
|
|
crnot 4*cr0+eq,4*cr0+eq
|
|
beqlr
|
|
|
|
/* now we need to get to the page tables in order to get the page
|
|
* size encoding from the PMD. In the future, we'll be able to deal
|
|
* with 1T segments too by getting the encoding from the PGD instead
|
|
*/
|
|
ld r9,PACAPGDIR(r13)
|
|
cmpldi cr0,r9,0
|
|
beqlr
|
|
rlwinm r11,r10,8,25,28
|
|
ldx r9,r9,r11 /* get pgd_t */
|
|
cmpldi cr0,r9,0
|
|
beqlr
|
|
rlwinm r11,r10,3,17,28
|
|
ldx r9,r9,r11 /* get pmd_t */
|
|
cmpldi cr0,r9,0
|
|
beqlr
|
|
|
|
/* build vsid flags */
|
|
andi. r11,r9,SLB_VSID_LLP
|
|
ori r11,r11,SLB_VSID_USER
|
|
|
|
/* get context to calculate proto-VSID */
|
|
ld r9,PACACONTEXTID(r13)
|
|
rldimi r10,r9,USER_ESID_BITS,0
|
|
|
|
/* fall through slb_finish_load */
|
|
|
|
#endif /* __DISABLED__ */
|
|
|
|
|
|
/*
|
|
* Finish loading of an SLB entry and return
|
|
*
|
|
* r3 = EA, r10 = proto-VSID, r11 = flags, clobbers r9, cr7 = <> PAGE_OFFSET
|
|
*/
|
|
slb_finish_load:
|
|
ASM_VSID_SCRAMBLE(r10,r9)
|
|
rldimi r11,r10,SLB_VSID_SHIFT,16 /* combine VSID and flags */
|
|
|
|
/* r3 = EA, r11 = VSID data */
|
|
/*
|
|
* Find a slot, round robin. Previously we tried to find a
|
|
* free slot first but that took too long. Unfortunately we
|
|
* dont have any LRU information to help us choose a slot.
|
|
*/
|
|
#ifdef CONFIG_PPC_ISERIES
|
|
BEGIN_FW_FTR_SECTION
|
|
/*
|
|
* On iSeries, the "bolted" stack segment can be cast out on
|
|
* shared processor switch so we need to check for a miss on
|
|
* it and restore it to the right slot.
|
|
*/
|
|
ld r9,PACAKSAVE(r13)
|
|
clrrdi r9,r9,28
|
|
clrrdi r3,r3,28
|
|
li r10,SLB_NUM_BOLTED-1 /* Stack goes in last bolted slot */
|
|
cmpld r9,r3
|
|
beq 3f
|
|
END_FW_FTR_SECTION_IFSET(FW_FEATURE_ISERIES)
|
|
#endif /* CONFIG_PPC_ISERIES */
|
|
|
|
ld r10,PACASTABRR(r13)
|
|
addi r10,r10,1
|
|
/* use a cpu feature mask if we ever change our slb size */
|
|
cmpldi r10,SLB_NUM_ENTRIES
|
|
|
|
blt+ 4f
|
|
li r10,SLB_NUM_BOLTED
|
|
|
|
4:
|
|
std r10,PACASTABRR(r13)
|
|
|
|
3:
|
|
rldimi r3,r10,0,36 /* r3= EA[0:35] | entry */
|
|
oris r10,r3,SLB_ESID_V@h /* r3 |= SLB_ESID_V */
|
|
|
|
/* r3 = ESID data, r11 = VSID data */
|
|
|
|
/*
|
|
* No need for an isync before or after this slbmte. The exception
|
|
* we enter with and the rfid we exit with are context synchronizing.
|
|
*/
|
|
slbmte r11,r10
|
|
|
|
/* we're done for kernel addresses */
|
|
crclr 4*cr0+eq /* set result to "success" */
|
|
bgelr cr7
|
|
|
|
/* Update the slb cache */
|
|
lhz r3,PACASLBCACHEPTR(r13) /* offset = paca->slb_cache_ptr */
|
|
cmpldi r3,SLB_CACHE_ENTRIES
|
|
bge 1f
|
|
|
|
/* still room in the slb cache */
|
|
sldi r11,r3,1 /* r11 = offset * sizeof(u16) */
|
|
rldicl r10,r10,36,28 /* get low 16 bits of the ESID */
|
|
add r11,r11,r13 /* r11 = (u16 *)paca + offset */
|
|
sth r10,PACASLBCACHE(r11) /* paca->slb_cache[offset] = esid */
|
|
addi r3,r3,1 /* offset++ */
|
|
b 2f
|
|
1: /* offset >= SLB_CACHE_ENTRIES */
|
|
li r3,SLB_CACHE_ENTRIES+1
|
|
2:
|
|
sth r3,PACASLBCACHEPTR(r13) /* paca->slb_cache_ptr = offset */
|
|
crclr 4*cr0+eq /* set result to "success" */
|
|
blr
|
|
|