korg/linux
0
0
Fork 0
mirror of https://mirrors.bfsu.edu.cn/git/linux.git synced 2024-11-11 04:18:39 +08:00
Code Issues Packages Projects Releases Wiki Activity

ARC: mm: retire support for aliasing VIPT D$

Browse Source 
Legacy ARC700 processors (first generation of MMU enabled ARC cores) had
VIPT cached which could be configured such that they could alias.
Corresponding support in kernel (with all the obnoxious cache flush overhead)
was added in ARC port 10 years ago to support 1 silicon. That is long bygone
and we can let it RIP.

Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Vineet Gupta <vgupta@kernel.org>
This commit is contained in:
Vineet Gupta 2023-02-15 21:06:33 -08:00
parent 3a02ec2f0b
commit 6732c0e494
5 changed files with 14 additions and 207 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
arch/arc/Kconfig
View File

@ -49,7 +49,6 @@ config ARC
select OF
select OF_EARLY_FLATTREE
select PCI_SYSCALL if PCI
select PERF_USE_VMALLOC if ARC_CACHE_VIPT_ALIASING
select HAVE_ARCH_JUMP_LABEL if ISA_ARCV2 && !CPU_ENDIAN_BE32
select TRACE_IRQFLAGS_SUPPORT
@ -232,10 +231,6 @@ config ARC_CACHE_PAGES
Note that Global I/D ENABLE + Per Page DISABLE works but corollary
Global DISABLE + Per Page ENABLE won't work
config ARC_CACHE_VIPT_ALIASING
bool "Support VIPT Aliasing D$"
depends on ARC_HAS_DCACHE && ISA_ARCOMPACT
endif #ARC_CACHE
config ARC_HAS_ICCM

43
arch/arc/include/asm/cacheflush.h
View File

@ -44,31 +44,10 @@ void dma_cache_wback(phys_addr_t start, unsigned long sz);
#define flush_cache_dup_mm(mm) /* called on fork (VIVT only) */
#ifndef CONFIG_ARC_CACHE_VIPT_ALIASING
#define flush_cache_mm(mm) /* called on munmap/exit */
#define flush_cache_range(mm, u_vstart, u_vend)
#define flush_cache_page(vma, u_vaddr, pfn) /* PF handling/COW-break */
#else /* VIPT aliasing dcache */
/* To clear out stale userspace mappings */
void flush_cache_mm(struct mm_struct *mm);
void flush_cache_range(struct vm_area_struct *vma,
unsigned long start,unsigned long end);
void flush_cache_page(struct vm_area_struct *vma,
unsigned long user_addr, unsigned long page);
/*
* To make sure that userspace mapping is flushed to memory before
* get_user_pages() uses a kernel mapping to access the page
*/
#define ARCH_HAS_FLUSH_ANON_PAGE
void flush_anon_page(struct vm_area_struct *vma,
struct page *page, unsigned long u_vaddr);
#endif /* CONFIG_ARC_CACHE_VIPT_ALIASING */
/*
* A new pagecache page has PG_arch_1 clear - thus dcache dirty by default
* This works around some PIO based drivers which don't call flush_dcache_page
@ -76,28 +55,6 @@ void flush_anon_page(struct vm_area_struct *vma,
*/
#define PG_dc_clean PG_arch_1
#define CACHE_COLORS_NUM 4
#define CACHE_COLORS_MSK (CACHE_COLORS_NUM - 1)
#define CACHE_COLOR(addr) (((unsigned long)(addr) >> (PAGE_SHIFT)) & CACHE_COLORS_MSK)
/*
* Simple wrapper over config option
* Bootup code ensures that hardware matches kernel configuration
*/
static inline int cache_is_vipt_aliasing(void)
{
return IS_ENABLED(CONFIG_ARC_CACHE_VIPT_ALIASING);
}
/*
* checks if two addresses (after page aligning) index into same cache set
*/
#define addr_not_cache_congruent(addr1, addr2) \
({ \
cache_is_vipt_aliasing() ? \
(CACHE_COLOR(addr1) != CACHE_COLOR(addr2)) : 0; \
})
#define copy_to_user_page(vma, page, vaddr, dst, src, len) \
do { \
memcpy(dst, src, len); \

136
arch/arc/mm/cache.c
View File

@ -145,10 +145,9 @@ dc_chk:
p_dc->sz_k = 1 << (dbcr.sz - 1);
n += scnprintf(buf + n, len - n,
"D-Cache\t\t: %uK, %dway/set, %uB Line, %s%s%s\n",
"D-Cache\t\t: %uK, %dway/set, %uB Line, %s%s\n",
p_dc->sz_k, assoc, p_dc->line_len,
vipt ? "VIPT" : "PIPT",
p_dc->colors > 1 ? " aliasing" : "",
IS_USED_CFG(CONFIG_ARC_HAS_DCACHE));
slc_chk:
@ -703,51 +702,10 @@ static inline void arc_slc_enable(void)
* Exported APIs
*/
/*
* Handle cache congruency of kernel and userspace mappings of page when kernel
* writes-to/reads-from
*
* The idea is to defer flushing of kernel mapping after a WRITE, possible if:
* -dcache is NOT aliasing, hence any U/K-mappings of page are congruent
* -U-mapping doesn't exist yet for page (finalised in update_mmu_cache)
* -In SMP, if hardware caches are coherent
*
* There's a corollary case, where kernel READs from a userspace mapped page.
* If the U-mapping is not congruent to K-mapping, former needs flushing.
*/
void flush_dcache_folio(struct folio *folio)
{
struct address_space *mapping;
if (!cache_is_vipt_aliasing()) {
clear_bit(PG_dc_clean, &folio->flags);
return;
}
/* don't handle anon pages here */
mapping = folio_flush_mapping(folio);
if (!mapping)
return;
/*
* pagecache page, file not yet mapped to userspace
* Make a note that K-mapping is dirty
*/
if (!mapping_mapped(mapping)) {
clear_bit(PG_dc_clean, &folio->flags);
} else if (folio_mapped(folio)) {
/* kernel reading from page with U-mapping */
phys_addr_t paddr = (unsigned long)folio_address(folio);
unsigned long vaddr = folio_pos(folio);
/*
* vaddr is not actually the virtual address, but is
* congruent to every user mapping.
*/
if (addr_not_cache_congruent(paddr, vaddr))
__flush_dcache_pages(paddr, vaddr,
folio_nr_pages(folio));
}
clear_bit(PG_dc_clean, &folio->flags);
return;
}
EXPORT_SYMBOL(flush_dcache_folio);
@ -921,44 +879,6 @@ noinline void flush_cache_all(void)
}
#ifdef CONFIG_ARC_CACHE_VIPT_ALIASING
void flush_cache_mm(struct mm_struct *mm)
{
flush_cache_all();
}
void flush_cache_page(struct vm_area_struct *vma, unsigned long u_vaddr,
unsigned long pfn)
{
phys_addr_t paddr = pfn << PAGE_SHIFT;
u_vaddr &= PAGE_MASK;
__flush_dcache_pages(paddr, u_vaddr, 1);
if (vma->vm_flags & VM_EXEC)
__inv_icache_pages(paddr, u_vaddr, 1);
}
void flush_cache_range(struct vm_area_struct *vma, unsigned long start,
unsigned long end)
{
flush_cache_all();
}
void flush_anon_page(struct vm_area_struct *vma, struct page *page,
unsigned long u_vaddr)
{
/* TBD: do we really need to clear the kernel mapping */
__flush_dcache_pages((phys_addr_t)page_address(page), u_vaddr, 1);
__flush_dcache_pages((phys_addr_t)page_address(page),
(phys_addr_t)page_address(page), 1);
}
#endif
void copy_user_highpage(struct page *to, struct page *from,
unsigned long u_vaddr, struct vm_area_struct *vma)
{
@ -966,46 +886,11 @@ void copy_user_highpage(struct page *to, struct page *from,
struct folio *dst = page_folio(to);
void *kfrom = kmap_atomic(from);
void *kto = kmap_atomic(to);
int clean_src_k_mappings = 0;
/*
* If SRC page was already mapped in userspace AND it's U-mapping is
* not congruent with K-mapping, sync former to physical page so that
* K-mapping in memcpy below, sees the right data
*
* Note that while @u_vaddr refers to DST page's userspace vaddr, it is
* equally valid for SRC page as well
*
* For !VIPT cache, all of this gets compiled out as
* addr_not_cache_congruent() is 0
*/
if (page_mapcount(from) && addr_not_cache_congruent(kfrom, u_vaddr)) {
__flush_dcache_pages((unsigned long)kfrom, u_vaddr, 1);
clean_src_k_mappings = 1;
}
copy_page(kto, kfrom);
/*
* Mark DST page K-mapping as dirty for a later finalization by
* update_mmu_cache(). Although the finalization could have been done
* here as well (given that both vaddr/paddr are available).
* But update_mmu_cache() already has code to do that for other
* non copied user pages (e.g. read faults which wire in pagecache page
* directly).
*/
clear_bit(PG_dc_clean, &dst->flags);
/*
* if SRC was already usermapped and non-congruent to kernel mapping
* sync the kernel mapping back to physical page
*/
if (clean_src_k_mappings) {
__flush_dcache_pages((unsigned long)kfrom,
(unsigned long)kfrom, 1);
} else {
clear_bit(PG_dc_clean, &src->flags);
}
clear_bit(PG_dc_clean, &src->flags);
kunmap_atomic(kto);
kunmap_atomic(kfrom);
@ -1140,17 +1025,8 @@ static noinline void __init arc_cache_init_master(void)
dc->line_len, L1_CACHE_BYTES);
/* check for D-Cache aliasing on ARCompact: ARCv2 has PIPT */
if (is_isa_arcompact()) {
int handled = IS_ENABLED(CONFIG_ARC_CACHE_VIPT_ALIASING);
if (dc->colors > 1) {
if (!handled)
panic("Enable CONFIG_ARC_CACHE_VIPT_ALIASING\n");
if (CACHE_COLORS_NUM != dc->colors)
panic("CACHE_COLORS_NUM not optimized for config\n");
} else if (handled && dc->colors == 1) {
panic("Disable CONFIG_ARC_CACHE_VIPT_ALIASING\n");
}
if (is_isa_arcompact() && dc->colors > 1) {
panic("Aliasing VIPT cache not supported\n");
}
}

21
arch/arc/mm/mmap.c
View File

@ -14,10 +14,6 @@
#include <asm/cacheflush.h>
#define COLOUR_ALIGN(addr, pgoff) \
((((addr) + SHMLBA - 1) & ~(SHMLBA - 1)) + \
(((pgoff) << PAGE_SHIFT) & (SHMLBA - 1)))
/*
* Ensure that shared mappings are correctly aligned to
* avoid aliasing issues with VIPT caches.
@ -31,21 +27,13 @@ arch_get_unmapped_area(struct file *filp, unsigned long addr,
{
struct mm_struct *mm = current->mm;
struct vm_area_struct *vma;
int do_align = 0;
int aliasing = cache_is_vipt_aliasing();
struct vm_unmapped_area_info info;
/*
* We only need to do colour alignment if D cache aliases.
*/
if (aliasing)
do_align = filp || (flags & MAP_SHARED);
/*
* We enforce the MAP_FIXED case.
*/
if (flags & MAP_FIXED) {
if (aliasing && flags & MAP_SHARED &&
if (flags & MAP_SHARED &&
(addr - (pgoff << PAGE_SHIFT)) & (SHMLBA - 1))
return -EINVAL;
return addr;
@ -55,10 +43,7 @@ arch_get_unmapped_area(struct file *filp, unsigned long addr,
return -ENOMEM;
if (addr) {
if (do_align)
addr = COLOUR_ALIGN(addr, pgoff);
else
addr = PAGE_ALIGN(addr);
addr = PAGE_ALIGN(addr);
vma = find_vma(mm, addr);
if (TASK_SIZE - len >= addr &&
@ -70,7 +55,7 @@ arch_get_unmapped_area(struct file *filp, unsigned long addr,
info.length = len;
info.low_limit = mm->mmap_base;
info.high_limit = TASK_SIZE;
info.align_mask = do_align ? (PAGE_MASK & (SHMLBA - 1)) : 0;
info.align_mask = 0;
info.align_offset = pgoff << PAGE_SHIFT;
return vm_unmapped_area(&info);
}

16
arch/arc/mm/tlb.c
View File

@ -478,21 +478,15 @@ void update_mmu_cache_range(struct vm_fault *vmf, struct vm_area_struct *vma,
create_tlb(vma, vaddr, ptep);
if (page == ZERO_PAGE(0)) {
if (page == ZERO_PAGE(0))
return;
}
/*
* Exec page : Independent of aliasing/page-color considerations,
* since icache doesn't snoop dcache on ARC, any dirty
* K-mapping of a code page needs to be wback+inv so that
* icache fetch by userspace sees code correctly.
* !EXEC page: If K-mapping is NOT congruent to U-mapping, flush it
* so userspace sees the right data.
* (Avoids the flush for Non-exec + congruent mapping case)
* For executable pages, since icache doesn't snoop dcache, any
* dirty K-mapping of a code page needs to be wback+inv so that
* icache fetch by userspace sees code correctly.
*/
if ((vma->vm_flags & VM_EXEC) ||
addr_not_cache_congruent(paddr, vaddr)) {
if (vma->vm_flags & VM_EXEC) {
struct folio *folio = page_folio(page);
int dirty = !test_and_set_bit(PG_dc_clean, &folio->flags);
if (dirty) {