diff options
author | Russell King | 2006-12-30 23:17:40 +0000 |
---|---|---|
committer | Russell King | 2007-01-08 19:49:58 +0000 |
commit | 6020dff09252e3670a89edb36baaa4afb9b10d15 (patch) | |
tree | e242232e97e71045f51bda9e0a5bb2e8f8df1c29 /arch/arm/mm | |
parent | a6f36be32622730710b2fadacb6e2649defa4371 (diff) |
[ARM] Resolve fuse and direct-IO failures due to missing cache flushes
fuse does not work on ARM due to cache incoherency issues - fuse wants
to use get_user_pages() to copy data from the current process into
kernel space. However, since this accesses userspace via the kernel
mapping, the kernel mapping can be out of date wrt data written to
userspace.
This can lead to unpredictable behaviour (in the case of fuse) or data
corruption for direct-IO.
This resolves debian bug #402876
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Diffstat (limited to 'arch/arm/mm')
-rw-r--r-- | arch/arm/mm/flush.c | 39 |
1 files changed, 39 insertions, 0 deletions
diff --git a/arch/arm/mm/flush.c b/arch/arm/mm/flush.c index 628348c9f6c5..9df507d36e0b 100644 --- a/arch/arm/mm/flush.c +++ b/arch/arm/mm/flush.c @@ -202,3 +202,42 @@ void flush_dcache_page(struct page *page) } } EXPORT_SYMBOL(flush_dcache_page); + +/* + * Flush an anonymous page so that users of get_user_pages() + * can safely access the data. The expected sequence is: + * + * get_user_pages() + * -> flush_anon_page + * memcpy() to/from page + * if written to page, flush_dcache_page() + */ +void __flush_anon_page(struct vm_area_struct *vma, struct page *page, unsigned long vmaddr) +{ + unsigned long pfn; + + /* VIPT non-aliasing caches need do nothing */ + if (cache_is_vipt_nonaliasing()) + return; + + /* + * Write back and invalidate userspace mapping. + */ + pfn = page_to_pfn(page); + if (cache_is_vivt()) { + flush_cache_page(vma, vmaddr, pfn); + } else { + /* + * For aliasing VIPT, we can flush an alias of the + * userspace address only. + */ + flush_pfn_alias(pfn, vmaddr); + } + + /* + * Invalidate kernel mapping. No data should be contained + * in this mapping of the page. FIXME: this is overkill + * since we actually ask for a write-back and invalidate. + */ + __cpuc_flush_dcache_page(page_address(page)); +} |