[v2] Based on linux-next/akpm [3dc4623]. For v4.1 merge window Incorporated comments from Andrew And Kirill [v1] This will allow FS that uses VM_PFNMAP | VM_MIXEDMAP (no page structs) to get notified when access is a write to a read-only PFN. This can happen if we mmap() a file then first mmap-read from it to page-in a read-only PFN, than we mmap-write to the same page. We need this functionality to fix a DAX bug, where in the scenario above we fail to set ctime/mtime though we modified the file. An xfstest is attached to this patchset that shows the failure and the fix. (A DAX patch will follow) This functionality is extra important for us, because upon dirtying of a pmem page we also want to RDMA the page to a remote cluster node. We define a new pfn_mkwrite and do not reuse page_mkwrite because 1 - The name ;-) 2 - But mainly because it would take a very long and tedious audit of all page_mkwrite functions of VM_MIXEDMAP/VM_PFNMAP users. To make sure they do not now CRASH. For example current DAX code (which this is for) would crash. If we would want to reuse page_mkwrite, We will need to first patch all users, so to not-crash-on-no-page. Then enable this patch. But even if I did that I would not sleep so well at night. Adding a new vector is the safest thing to do, and is not that expensive. an extra pointer at a static function vector per driver. Also the new vector is better for performance, because else we Will call all current Kernel vectors, so to: check-ha-no-page-do-nothing and return. No need to call it from do_shared_fault because do_wp_page is called to change pte permissions anyway. CC: Matthew Wilcox <matthew.r.wilcox(a)intel.com&gt; CC: Kirill A. Shutemov <kirill.shutemov(a)linux.intel.com&gt; CC: Jan Kara <jack(a)suse.cz&gt; CC: Andrew Morton <akpm(a)linux-foundation.org&gt; CC: Hugh Dickins <hughd(a)google.com&gt; CC: Mel Gorman <mgorman(a)suse.de&gt; CC: linux-mm(a)kvack.org Signed-off-by: Yigal Korman <yigal(a)plexistor.com&gt; Signed-off-by: Boaz Harrosh <boaz(a)plexistor.com&gt; --- include/linux/mm.h | 3 +++ mm/memory.c | 35 +++++++++++++++++++++++++++++++---- 2 files changed, 34 insertions(+), 4 deletions(-) diff --git a/include/linux/mm.h b/include/linux/mm.h index d584b95..70c47f2 100644 --- a/include/linux/mm.h +++ b/include/linux/mm.h @@ -251,6 +251,9 @@ struct vm_operations_struct { * writable, if an error is returned it will cause a SIGBUS */ int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf); + /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */ + int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf); + /* called by access_process_vm when get_user_pages() fails, typically * for use by special VMAs that can switch between memory and hardware */ diff --git a/mm/memory.c b/mm/memory.c index 59f6268..6e8f3f6 100644 --- a/mm/memory.c +++ b/mm/memory.c @@ -1982,6 +1982,19 @@ static int do_page_mkwrite(struct vm_area_struct *vma, struct page *page, return ret; } +static int do_pfn_mkwrite(struct vm_area_struct *vma, unsigned long address) +{ + struct vm_fault vmf = { + .page = 0, + .pgoff = (((address & PAGE_MASK) - vma->vm_start) + >> PAGE_SHIFT) + vma->vm_pgoff, + .virtual_address = (void __user *)(address & PAGE_MASK), + .flags = FAULT_FLAG_WRITE | FAULT_FLAG_MKWRITE, + }; + + return vma->vm_ops->pfn_mkwrite(vma, &vmf); +} + /* * Handle write page faults for pages that can be reused in the current vma * @@ -2259,14 +2272,28 @@ static int do_wp_page(struct mm_struct *mm, struct vm_area_struct *vma, * VM_PFNMAP VMA. * * We should not cow pages in a shared writeable mapping. - * Just mark the pages writable as we can't do any dirty - * accounting on raw pfn maps. + * Just mark the pages writable and/or call ops->pfn_mkwrite. */ if ((vma->vm_flags & (VM_WRITE|VM_SHARED)) == - (VM_WRITE|VM_SHARED)) + (VM_WRITE|VM_SHARED)) { + if (vma->vm_ops && vma->vm_ops->pfn_mkwrite) { + int ret; + + pte_unmap_unlock(page_table, ptl); + ret = do_pfn_mkwrite(vma, address); + if (ret & VM_FAULT_ERROR) + return ret; + page_table = pte_offset_map_lock(mm, pmd, + address, &ptl); + /* Did pfn_mkwrite already fixed up the pte */ + if (!pte_same(*page_table, orig_pte)) { + pte_unmap_unlock(page_table, ptl); + return ret; + } + } return wp_page_reuse(mm, vma, address, page_table, ptl, orig_pte, old_page, 0, 0); - + } pte_unmap_unlock(page_table, ptl); return wp_page_copy(mm, vma, address, page_table, pmd, orig_pte, old_page); -- 1.9.3

On Tue, Apr 07, 2015 at 11:40:06AM +0300, Boaz Harrosh wrote: Not really. You've ignored most of them. See below. Please, document it in Documentation/filesystems/Locking. .page = NULL, .pgoff = linear_page_index(vma, address), Let's move this case in separate function -- wp_pfn_shared(). As we do for wp_page_shared(). -- Kirill A. Shutemov

[v4] Kirill's comments about splitting out a new wp_pfn_shared(). Add Documentation/filesystems/Locking text about pfn_mkwrite. [v3] Kirill's comments about use of linear_page_index() [v2] Based on linux-next/akpm [3dc4623]. For v4.1 merge window Incorporated comments from Andrew And Kirill [v1] This will allow FS that uses VM_PFNMAP | VM_MIXEDMAP (no page structs) to get notified when access is a write to a read-only PFN. This can happen if we mmap() a file then first mmap-read from it to page-in a read-only PFN, than we mmap-write to the same page. We need this functionality to fix a DAX bug, where in the scenario above we fail to set ctime/mtime though we modified the file. An xfstest is attached to this patchset that shows the failure and the fix. (A DAX patch will follow) This functionality is extra important for us, because upon dirtying of a pmem page we also want to RDMA the page to a remote cluster node. We define a new pfn_mkwrite and do not reuse page_mkwrite because 1 - The name ;-) 2 - But mainly because it would take a very long and tedious audit of all page_mkwrite functions of VM_MIXEDMAP/VM_PFNMAP users. To make sure they do not now CRASH. For example current DAX code (which this is for) would crash. If we would want to reuse page_mkwrite, We will need to first patch all users, so to not-crash-on-no-page. Then enable this patch. But even if I did that I would not sleep so well at night. Adding a new vector is the safest thing to do, and is not that expensive. an extra pointer at a static function vector per driver. Also the new vector is better for performance, because else we Will call all current Kernel vectors, so to: check-ha-no-page-do-nothing and return. No need to call it from do_shared_fault because do_wp_page is called to change pte permissions anyway. CC: Kirill A. Shutemov <kirill.shutemov(a)linux.intel.com&gt; CC: Matthew Wilcox <matthew.r.wilcox(a)intel.com&gt; CC: Jan Kara <jack(a)suse.cz&gt; CC: Andrew Morton <akpm(a)linux-foundation.org&gt; CC: Hugh Dickins <hughd(a)google.com&gt; CC: Mel Gorman <mgorman(a)suse.de&gt; CC: linux-mm(a)kvack.org Signed-off-by: Yigal Korman <yigal(a)plexistor.com&gt; Signed-off-by: Boaz Harrosh <boaz(a)plexistor.com&gt; --- Documentation/filesystems/Locking | 8 ++++++++ include/linux/mm.h | 3 +++ mm/memory.c | 40 +++++++++++++++++++++++++++++++++++---- 3 files changed, 47 insertions(+), 4 deletions(-) diff --git a/Documentation/filesystems/Locking b/Documentation/filesystems/Locking index f91926f..25f36e6 100644 --- a/Documentation/filesystems/Locking +++ b/Documentation/filesystems/Locking @@ -525,6 +525,7 @@ prototypes: void (*close)(struct vm_area_struct*); int (*fault)(struct vm_area_struct*, struct vm_fault *); int (*page_mkwrite)(struct vm_area_struct *, struct vm_fault *); + int (*pfn_mkwrite)(struct vm_area_struct *, struct vm_fault *); int (*access)(struct vm_area_struct *, unsigned long, void*, int, int); locking rules: @@ -534,6 +535,7 @@ close: yes fault: yes can return with page locked map_pages: yes page_mkwrite: yes can return with page locked +pfn_mkwrite: yes access: yes ->fault() is called when a previously not present pte is about @@ -560,6 +562,12 @@ the page has been truncated, the filesystem should not look up a new page like the ->fault() handler, but simply return with VM_FAULT_NOPAGE, which will cause the VM to retry the fault. + ->pfn_mkwrite() is the same as page_mkwrite but when the pte is +VM_PFNMAP or VM_MIXEDMAP with a page-less entry. Expected return is +VM_FAULT_NOPAGE. Or one of the VM_FAULT_ERROR types. The default behavior +after this call is to make the pte read-write, unless pfn_mkwrite() +already touched the pte, in that case it is untouched. + ->access() is called when get_user_pages() fails in access_process_vm(), typically used to debug a process through /proc/pid/mem or ptrace. This function is needed only for diff --git a/include/linux/mm.h b/include/linux/mm.h index d584b95..70c47f2 100644 --- a/include/linux/mm.h +++ b/include/linux/mm.h @@ -251,6 +251,9 @@ struct vm_operations_struct { * writable, if an error is returned it will cause a SIGBUS */ int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf); + /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */ + int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf); + /* called by access_process_vm when get_user_pages() fails, typically * for use by special VMAs that can switch between memory and hardware */ diff --git a/mm/memory.c b/mm/memory.c index 59f6268..67b1ee8 100644 --- a/mm/memory.c +++ b/mm/memory.c @@ -2181,6 +2181,39 @@ oom: return VM_FAULT_OOM; } +/* + * Handle write page faults for VM_MIXEDMAP or VM_PFNMAP for a VM_SHARED + * mapping + */ +static int wp_pfn_shared(struct mm_struct *mm, + struct vm_area_struct *vma, unsigned long address, + pte_t *page_table, spinlock_t *ptl, pte_t orig_pte, + pmd_t *pmd) +{ + if (vma->vm_ops && vma->vm_ops->pfn_mkwrite) { + struct vm_fault vmf = { + .page = NULL, + .pgoff = linear_page_index(vma, address), + .virtual_address = (void __user *)(address & PAGE_MASK), + .flags = FAULT_FLAG_WRITE | FAULT_FLAG_MKWRITE, + }; + int ret; + + pte_unmap_unlock(page_table, ptl); + ret = vma->vm_ops->pfn_mkwrite(vma, &vmf); + if (ret & VM_FAULT_ERROR) + return ret; + page_table = pte_offset_map_lock(mm, pmd, address, &ptl); + /* Did pfn_mkwrite already fixed up the pte */ + if (!pte_same(*page_table, orig_pte)) { + pte_unmap_unlock(page_table, ptl); + return ret; + } + } + return wp_page_reuse(mm, vma, address, page_table, ptl, orig_pte, + NULL, 0, 0); +} + static int wp_page_shared(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long address, pte_t *page_table, pmd_t *pmd, spinlock_t *ptl, pte_t orig_pte, @@ -2259,13 +2292,12 @@ static int do_wp_page(struct mm_struct *mm, struct vm_area_struct *vma, * VM_PFNMAP VMA. * * We should not cow pages in a shared writeable mapping. - * Just mark the pages writable as we can't do any dirty - * accounting on raw pfn maps. + * Just mark the pages writable and/or call ops->pfn_mkwrite. */ if ((vma->vm_flags & (VM_WRITE|VM_SHARED)) == (VM_WRITE|VM_SHARED)) - return wp_page_reuse(mm, vma, address, page_table, ptl, - orig_pte, old_page, 0, 0); + return wp_pfn_shared(mm, vma, address, page_table, ptl, + orig_pte, pmd); pte_unmap_unlock(page_table, ptl); return wp_page_copy(mm, vma, address, page_table, pmd, -- 1.9.3

[v5] Changed comments about pte_same check after the call to pfn_mkwrite and the return value. [v4] Kirill's comments about splitting out a new wp_pfn_shared(). Add Documentation/filesystems/Locking text about pfn_mkwrite. [v3] Kirill's comments about use of linear_page_index() [v2] Based on linux-next/akpm [3dc4623]. For v4.1 merge window Incorporated comments from Andrew And Kirill [v1] This will allow FS that uses VM_PFNMAP | VM_MIXEDMAP (no page structs) to get notified when access is a write to a read-only PFN. This can happen if we mmap() a file then first mmap-read from it to page-in a read-only PFN, than we mmap-write to the same page. We need this functionality to fix a DAX bug, where in the scenario above we fail to set ctime/mtime though we modified the file. An xfstest is attached to this patchset that shows the failure and the fix. (A DAX patch will follow) This functionality is extra important for us, because upon dirtying of a pmem page we also want to RDMA the page to a remote cluster node. We define a new pfn_mkwrite and do not reuse page_mkwrite because 1 - The name ;-) 2 - But mainly because it would take a very long and tedious audit of all page_mkwrite functions of VM_MIXEDMAP/VM_PFNMAP users. To make sure they do not now CRASH. For example current DAX code (which this is for) would crash. If we would want to reuse page_mkwrite, We will need to first patch all users, so to not-crash-on-no-page. Then enable this patch. But even if I did that I would not sleep so well at night. Adding a new vector is the safest thing to do, and is not that expensive. an extra pointer at a static function vector per driver. Also the new vector is better for performance, because else we Will call all current Kernel vectors, so to: check-ha-no-page-do-nothing and return. No need to call it from do_shared_fault because do_wp_page is called to change pte permissions anyway. CC: Kirill A. Shutemov <kirill.shutemov(a)linux.intel.com&gt; CC: Matthew Wilcox <matthew.r.wilcox(a)intel.com&gt; CC: Jan Kara <jack(a)suse.cz&gt; CC: Andrew Morton <akpm(a)linux-foundation.org&gt; CC: Hugh Dickins <hughd(a)google.com&gt; CC: Mel Gorman <mgorman(a)suse.de&gt; CC: linux-mm(a)kvack.org Signed-off-by: Yigal Korman <yigal(a)plexistor.com&gt; Signed-off-by: Boaz Harrosh <boaz(a)plexistor.com&gt; --- Documentation/filesystems/Locking | 8 ++++++++ include/linux/mm.h | 3 +++ mm/memory.c | 43 +++++++++++++++++++++++++++++++++++---- 3 files changed, 50 insertions(+), 4 deletions(-) diff --git a/Documentation/filesystems/Locking b/Documentation/filesystems/Locking index f91926f..8bb8a7e 100644 --- a/Documentation/filesystems/Locking +++ b/Documentation/filesystems/Locking @@ -525,6 +525,7 @@ prototypes: void (*close)(struct vm_area_struct*); int (*fault)(struct vm_area_struct*, struct vm_fault *); int (*page_mkwrite)(struct vm_area_struct *, struct vm_fault *); + int (*pfn_mkwrite)(struct vm_area_struct *, struct vm_fault *); int (*access)(struct vm_area_struct *, unsigned long, void*, int, int); locking rules: @@ -534,6 +535,7 @@ close: yes fault: yes can return with page locked map_pages: yes page_mkwrite: yes can return with page locked +pfn_mkwrite: yes access: yes ->fault() is called when a previously not present pte is about @@ -560,6 +562,12 @@ the page has been truncated, the filesystem should not look up a new page like the ->fault() handler, but simply return with VM_FAULT_NOPAGE, which will cause the VM to retry the fault. + ->pfn_mkwrite() is the same as page_mkwrite but when the pte is +VM_PFNMAP or VM_MIXEDMAP with a page-less entry. Expected return is +VM_FAULT_NOPAGE. Or one of the VM_FAULT_ERROR types. The default behavior +after this call is to make the pte read-write, unless pfn_mkwrite returns +an error. + ->access() is called when get_user_pages() fails in access_process_vm(), typically used to debug a process through /proc/pid/mem or ptrace. This function is needed only for diff --git a/include/linux/mm.h b/include/linux/mm.h index d584b95..70c47f2 100644 --- a/include/linux/mm.h +++ b/include/linux/mm.h @@ -251,6 +251,9 @@ struct vm_operations_struct { * writable, if an error is returned it will cause a SIGBUS */ int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf); + /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */ + int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf); + /* called by access_process_vm when get_user_pages() fails, typically * for use by special VMAs that can switch between memory and hardware */ diff --git a/mm/memory.c b/mm/memory.c index 59f6268..d839cbc 100644 --- a/mm/memory.c +++ b/mm/memory.c @@ -2181,6 +2181,42 @@ oom: return VM_FAULT_OOM; } +/* + * Handle write page faults for VM_MIXEDMAP or VM_PFNMAP for a VM_SHARED + * mapping + */ +static int wp_pfn_shared(struct mm_struct *mm, + struct vm_area_struct *vma, unsigned long address, + pte_t *page_table, spinlock_t *ptl, pte_t orig_pte, + pmd_t *pmd) +{ + if (vma->vm_ops && vma->vm_ops->pfn_mkwrite) { + struct vm_fault vmf = { + .page = NULL, + .pgoff = linear_page_index(vma, address), + .virtual_address = (void __user *)(address & PAGE_MASK), + .flags = FAULT_FLAG_WRITE | FAULT_FLAG_MKWRITE, + }; + int ret; + + pte_unmap_unlock(page_table, ptl); + ret = vma->vm_ops->pfn_mkwrite(vma, &vmf); + if (ret & VM_FAULT_ERROR) + return ret; + page_table = pte_offset_map_lock(mm, pmd, address, &ptl); + /* + * We might have raced with another page fault while we + * released the pte_offset_map_lock. + */ + if (!pte_same(*page_table, orig_pte)) { + pte_unmap_unlock(page_table, ptl); + return 0; + } + } + return wp_page_reuse(mm, vma, address, page_table, ptl, orig_pte, + NULL, 0, 0); +} + static int wp_page_shared(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long address, pte_t *page_table, pmd_t *pmd, spinlock_t *ptl, pte_t orig_pte, @@ -2259,13 +2295,12 @@ static int do_wp_page(struct mm_struct *mm, struct vm_area_struct *vma, * VM_PFNMAP VMA. * * We should not cow pages in a shared writeable mapping. - * Just mark the pages writable as we can't do any dirty - * accounting on raw pfn maps. + * Just mark the pages writable and/or call ops->pfn_mkwrite. */ if ((vma->vm_flags & (VM_WRITE|VM_SHARED)) == (VM_WRITE|VM_SHARED)) - return wp_page_reuse(mm, vma, address, page_table, ptl, - orig_pte, old_page, 0, 0); + return wp_pfn_shared(mm, vma, address, page_table, ptl, + orig_pte, pmd); pte_unmap_unlock(page_table, ptl); return wp_page_copy(mm, vma, address, page_table, pmd, -- 1.9.3

From: Yigal Korman <yigal(a)plexistor.com&gt; [v1] Without this patch, c/mtime is not updated correctly when mmap'ed page is first read from and then written to. A new xfstest is submitted for testing this (generic/080) [v2] Jan Kara has pointed out that if we add the sb_start/end_pagefault pair in the new pfn_mkwrite we are then fixing another bug where: A user could start writing to the page while filesystem is frozen. CC: Jan Kara <jack(a)suse.cz&gt; CC: Matthew Wilcox <matthew.r.wilcox(a)intel.com&gt; CC: Dave Chinner <david(a)fromorbit.com&gt; CC: Stable Tree <stable(a)vger.kernel.org&gt; Signed-off-by: Yigal Korman <yigal(a)plexistor.com&gt; Signed-off-by: Boaz Harrosh <boaz(a)plexistor.com&gt; --- fs/dax.c | 17 +++++++++++++++++ fs/ext2/file.c | 1 + fs/ext4/file.c | 1 + include/linux/fs.h | 1 + 4 files changed, 20 insertions(+) diff --git a/fs/dax.c b/fs/dax.c index ed1619e..d0bd1f4 100644 --- a/fs/dax.c +++ b/fs/dax.c @@ -464,6 +464,23 @@ int dax_fault(struct vm_area_struct *vma, struct vm_fault *vmf, EXPORT_SYMBOL_GPL(dax_fault); /** + * dax_pfn_mkwrite - handle first write to DAX page + * @vma: The virtual memory area where the fault occurred + * @vmf: The description of the fault + * + */ +int dax_pfn_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) +{ + struct super_block *sb = file_inode(vma->vm_file)->i_sb; + + sb_start_pagefault(sb); + file_update_time(vma->vm_file); + sb_end_pagefault(sb); + return VM_FAULT_NOPAGE; +} +EXPORT_SYMBOL_GPL(dax_pfn_mkwrite); + +/** * dax_zero_page_range - zero a range within a page of a DAX file * @inode: The file being truncated * @from: The file offset that is being truncated to diff --git a/fs/ext2/file.c b/fs/ext2/file.c index e317017..866a3ce 100644 --- a/fs/ext2/file.c +++ b/fs/ext2/file.c @@ -39,6 +39,7 @@ static int ext2_dax_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) static const struct vm_operations_struct ext2_dax_vm_ops = { .fault = ext2_dax_fault, .page_mkwrite = ext2_dax_mkwrite, + .pfn_mkwrite = dax_pfn_mkwrite, }; static int ext2_file_mmap(struct file *file, struct vm_area_struct *vma) diff --git a/fs/ext4/file.c b/fs/ext4/file.c index 598abbb..aa78c70 100644 --- a/fs/ext4/file.c +++ b/fs/ext4/file.c @@ -206,6 +206,7 @@ static int ext4_dax_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) static const struct vm_operations_struct ext4_dax_vm_ops = { .fault = ext4_dax_fault, .page_mkwrite = ext4_dax_mkwrite, + .pfn_mkwrite = dax_pfn_mkwrite, }; #else #define ext4_dax_vm_ops ext4_file_vm_ops diff --git a/include/linux/fs.h b/include/linux/fs.h index 368e349..394035f 100644 --- a/include/linux/fs.h +++ b/include/linux/fs.h @@ -2628,6 +2628,7 @@ int dax_clear_blocks(struct inode *, sector_t block, long size); int dax_zero_page_range(struct inode *, loff_t from, unsigned len, get_block_t); int dax_truncate_page(struct inode *, loff_t from, get_block_t); int dax_fault(struct vm_area_struct *, struct vm_fault *, get_block_t); +int dax_pfn_mkwrite(struct vm_area_struct *, struct vm_fault *); #define dax_mkwrite(vma, vmf, gb) dax_fault(vma, vmf, gb) #ifdef CONFIG_BLOCK -- 1.9.3

The original dax patchset split the ext2/4_file_operations because of the two NULL splice_read/splice_write in the dax case. At vfs if splice_read/splice_write are NULL would then call default_splice_read/write. What we do here is make generic_file_splice_read aware of IS_DAX() so the original ext2/4_file_operations can be used as is. For write it appears that iter_file_splice_write is just fine. It uses the regular f_op->write(file,..) or new_sync_write(file, ...). CC: Dave Chinner <dchinner(a)redhat.com&gt; CC: Matthew Wilcox <willy(a)linux.intel.com&gt; Signed-off-by: Boaz Harrosh <boaz(a)plexistor.com&gt; --- fs/ext2/ext2.h | 1 - fs/ext2/file.c | 18 ------------------ fs/ext2/inode.c | 5 +---- fs/ext2/namei.c | 10 ++-------- fs/ext4/ext4.h | 1 - fs/ext4/file.c | 20 -------------------- fs/ext4/inode.c | 5 +---- fs/ext4/namei.c | 10 ++-------- fs/splice.c | 3 +++ 9 files changed, 9 insertions(+), 64 deletions(-) diff --git a/fs/ext2/ext2.h b/fs/ext2/ext2.h index 678f9ab..8d15feb 100644 --- a/fs/ext2/ext2.h +++ b/fs/ext2/ext2.h @@ -793,7 +793,6 @@ extern int ext2_fsync(struct file *file, loff_t start, loff_t end, int datasync); extern const struct inode_operations ext2_file_inode_operations; extern const struct file_operations ext2_file_operations; -extern const struct file_operations ext2_dax_file_operations; /* inode.c */ extern const struct address_space_operations ext2_aops; diff --git a/fs/ext2/file.c b/fs/ext2/file.c index 866a3ce..19cac93 100644 --- a/fs/ext2/file.c +++ b/fs/ext2/file.c @@ -109,24 +109,6 @@ const struct file_operations ext2_file_operations = { .splice_write = iter_file_splice_write, }; -#ifdef CONFIG_FS_DAX -const struct file_operations ext2_dax_file_operations = { - .llseek = generic_file_llseek, - .read = new_sync_read, - .write = new_sync_write, - .read_iter = generic_file_read_iter, - .write_iter = generic_file_write_iter, - .unlocked_ioctl = ext2_ioctl, -#ifdef CONFIG_COMPAT - .compat_ioctl = ext2_compat_ioctl, -#endif - .mmap = ext2_file_mmap, - .open = dquot_file_open, - .release = ext2_release_file, - .fsync = ext2_fsync, -}; -#endif - const struct inode_operations ext2_file_inode_operations = { #ifdef CONFIG_EXT2_FS_XATTR .setxattr = generic_setxattr, diff --git a/fs/ext2/inode.c b/fs/ext2/inode.c index df9d6af..b29eb67 100644 --- a/fs/ext2/inode.c +++ b/fs/ext2/inode.c @@ -1388,10 +1388,7 @@ struct inode *ext2_iget (struct super_block *sb, unsigned long ino) if (S_ISREG(inode->i_mode)) { inode->i_op = &ext2_file_inode_operations; - if (test_opt(inode->i_sb, DAX)) { - inode->i_mapping->a_ops = &ext2_aops; - inode->i_fop = &ext2_dax_file_operations; - } else if (test_opt(inode->i_sb, NOBH)) { + if (test_opt(inode->i_sb, NOBH)) { inode->i_mapping->a_ops = &ext2_nobh_aops; inode->i_fop = &ext2_file_operations; } else { diff --git a/fs/ext2/namei.c b/fs/ext2/namei.c index 148f6e3..ce42293 100644 --- a/fs/ext2/namei.c +++ b/fs/ext2/namei.c @@ -104,10 +104,7 @@ static int ext2_create (struct inode * dir, struct dentry * dentry, umode_t mode return PTR_ERR(inode); inode->i_op = &ext2_file_inode_operations; - if (test_opt(inode->i_sb, DAX)) { - inode->i_mapping->a_ops = &ext2_aops; - inode->i_fop = &ext2_dax_file_operations; - } else if (test_opt(inode->i_sb, NOBH)) { + if (test_opt(inode->i_sb, NOBH)) { inode->i_mapping->a_ops = &ext2_nobh_aops; inode->i_fop = &ext2_file_operations; } else { @@ -125,10 +122,7 @@ static int ext2_tmpfile(struct inode *dir, struct dentry *dentry, umode_t mode) return PTR_ERR(inode); inode->i_op = &ext2_file_inode_operations; - if (test_opt(inode->i_sb, DAX)) { - inode->i_mapping->a_ops = &ext2_aops; - inode->i_fop = &ext2_dax_file_operations; - } else if (test_opt(inode->i_sb, NOBH)) { + if (test_opt(inode->i_sb, NOBH)) { inode->i_mapping->a_ops = &ext2_nobh_aops; inode->i_fop = &ext2_file_operations; } else { diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h index f63c3d5..8a3981e 100644 --- a/fs/ext4/ext4.h +++ b/fs/ext4/ext4.h @@ -2593,7 +2593,6 @@ extern const struct file_operations ext4_dir_operations; /* file.c */ extern const struct inode_operations ext4_file_inode_operations; extern const struct file_operations ext4_file_operations; -extern const struct file_operations ext4_dax_file_operations; extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin); /* inline.c */ diff --git a/fs/ext4/file.c b/fs/ext4/file.c index aa78c70..e6d4280 100644 --- a/fs/ext4/file.c +++ b/fs/ext4/file.c @@ -625,26 +625,6 @@ const struct file_operations ext4_file_operations = { .fallocate = ext4_fallocate, }; -#ifdef CONFIG_FS_DAX -const struct file_operations ext4_dax_file_operations = { - .llseek = ext4_llseek, - .read = new_sync_read, - .write = new_sync_write, - .read_iter = generic_file_read_iter, - .write_iter = ext4_file_write_iter, - .unlocked_ioctl = ext4_ioctl, -#ifdef CONFIG_COMPAT - .compat_ioctl = ext4_compat_ioctl, -#endif - .mmap = ext4_file_mmap, - .open = ext4_file_open, - .release = ext4_release_file, - .fsync = ext4_sync_file, - /* Splice not yet supported with DAX */ - .fallocate = ext4_fallocate, -}; -#endif - const struct inode_operations ext4_file_inode_operations = { .setattr = ext4_setattr, .getattr = ext4_getattr, diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index a3f4513..035b7a0 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -4090,10 +4090,7 @@ struct inode *ext4_iget(struct super_block *sb, unsigned long ino) if (S_ISREG(inode->i_mode)) { inode->i_op = &ext4_file_inode_operations; - if (test_opt(inode->i_sb, DAX)) - inode->i_fop = &ext4_dax_file_operations; - else - inode->i_fop = &ext4_file_operations; + inode->i_fop = &ext4_file_operations; ext4_set_aops(inode); } else if (S_ISDIR(inode->i_mode)) { inode->i_op = &ext4_dir_inode_operations; diff --git a/fs/ext4/namei.c b/fs/ext4/namei.c index 28fe71a..2291923 100644 --- a/fs/ext4/namei.c +++ b/fs/ext4/namei.c @@ -2235,10 +2235,7 @@ retry: err = PTR_ERR(inode); if (!IS_ERR(inode)) { inode->i_op = &ext4_file_inode_operations; - if (test_opt(inode->i_sb, DAX)) - inode->i_fop = &ext4_dax_file_operations; - else - inode->i_fop = &ext4_file_operations; + inode->i_fop = &ext4_file_operations; ext4_set_aops(inode); err = ext4_add_nondir(handle, dentry, inode); if (!err && IS_DIRSYNC(dir)) @@ -2302,10 +2299,7 @@ retry: err = PTR_ERR(inode); if (!IS_ERR(inode)) { inode->i_op = &ext4_file_inode_operations; - if (test_opt(inode->i_sb, DAX)) - inode->i_fop = &ext4_dax_file_operations; - else - inode->i_fop = &ext4_file_operations; + inode->i_fop = &ext4_file_operations; ext4_set_aops(inode); d_tmpfile(dentry, inode); err = ext4_orphan_add(handle, inode); diff --git a/fs/splice.c b/fs/splice.c index 41cbb16..476024b 100644 --- a/fs/splice.c +++ b/fs/splice.c @@ -523,6 +523,9 @@ ssize_t generic_file_splice_read(struct file *in, loff_t *ppos, loff_t isize, left; int ret; + if (IS_DAX(in->f_mapping->host)) + return default_file_splice_read(in, ppos, pipe, len, flags); + isize = i_size_read(in->f_mapping->host); if (unlikely(*ppos >= isize)) return 0; -- 1.9.3

From: Yigal Korman <yigal(a)plexistor.com&gt; [For Stable 4.0.X] The parallel patch at 4.1-rc1 to this patch is: Subject: mm: new pfn_mkwrite same as page_mkwrite for VM_PFNMAP We need this patch for the 4.0.X stable tree if the patch Subject: dax: use pfn_mkwrite to update c/mtime + freeze protection Was decided to be pulled into stable since it is a dependency of this patch. The file mm/memory.c was heavily changed in 4.1 hence this here. [v3] In the case of !pte_same when we lost the race better return 0 instead of FAULT_NO_PAGE [v2] Fixed according to Kirill's comments [v1] This will allow FS that uses VM_PFNMAP | VM_MIXEDMAP (no page structs) to get notified when access is a write to a read-only PFN. This can happen if we mmap() a file then first mmap-read from it to page-in a read-only PFN, than we mmap-write to the same page. We need this functionality to fix a DAX bug, where in the scenario above we fail to set ctime/mtime though we modified the file. An xfstest is attached to this patchset that shows the failure and the fix. (A DAX patch will follow) This functionality is extra important for us, because upon dirtying of a pmem page we also want to RDMA the page to a remote cluster node. We define a new pfn_mkwrite and do not reuse page_mkwrite because 1 - The name ;-) 2 - But mainly because it would take a very long and tedious audit of all page_mkwrite functions of VM_MIXEDMAP/VM_PFNMAP users. To make sure they do not now CRASH. For example current DAX code (which this is for) would crash. If we would want to reuse page_mkwrite, We will need to first patch all users, so to not-crash-on-no-page. Then enable this patch. But even if I did that I would not sleep so well at night. Adding a new vector is the safest thing to do, and is not that expensive. an extra pointer at a static function vector per driver. Also the new vector is better for performance, because else we Will call all current Kernel vectors, so to: check-ha-no-page-do-nothing and return. No need to call it from do_shared_fault because do_wp_page is called to change pte permissions anyway. Signed-off-by: Yigal Korman <yigal(a)plexistor.com&gt; Signed-off-by: Boaz Harrosh <boaz(a)plexistor.com&gt; CC: Kirill A. Shutemov <kirill.shutemov(a)linux.intel.com&gt; CC: Matthew Wilcox <matthew.r.wilcox(a)intel.com&gt; CC: Jan Kara <jack(a)suse.cz&gt; CC: Andrew Morton <akpm(a)linux-foundation.org&gt; CC: Hugh Dickins <hughd(a)google.com&gt; CC: Mel Gorman <mgorman(a)suse.de&gt; CC: Konstantin Khlebnikov <koct9i(a)gmail.com&gt; CC: linux-mm(a)kvack.org CC: Stable Tree <stable(a)vger.kernel.org&gt; --- Documentation/filesystems/Locking | 8 ++++++++ include/linux/mm.h | 3 +++ mm/memory.c | 27 ++++++++++++++++++++++++++- 3 files changed, 37 insertions(+), 1 deletion(-) diff --git a/Documentation/filesystems/Locking b/Documentation/filesystems/Locking index f91926f..25f36e6 100644 --- a/Documentation/filesystems/Locking +++ b/Documentation/filesystems/Locking @@ -525,6 +525,7 @@ prototypes: void (*close)(struct vm_area_struct*); int (*fault)(struct vm_area_struct*, struct vm_fault *); int (*page_mkwrite)(struct vm_area_struct *, struct vm_fault *); + int (*pfn_mkwrite)(struct vm_area_struct *, struct vm_fault *); int (*access)(struct vm_area_struct *, unsigned long, void*, int, int); locking rules: @@ -534,6 +535,7 @@ close: yes fault: yes can return with page locked map_pages: yes page_mkwrite: yes can return with page locked +pfn_mkwrite: yes access: yes ->fault() is called when a previously not present pte is about @@ -560,6 +562,12 @@ the page has been truncated, the filesystem should not look up a new page like the ->fault() handler, but simply return with VM_FAULT_NOPAGE, which will cause the VM to retry the fault. + ->pfn_mkwrite() is the same as page_mkwrite but when the pte is +VM_PFNMAP or VM_MIXEDMAP with a page-less entry. Expected return is +VM_FAULT_NOPAGE. Or one of the VM_FAULT_ERROR types. The default behavior +after this call is to make the pte read-write, unless pfn_mkwrite() +already touched the pte, in that case it is untouched. + ->access() is called when get_user_pages() fails in access_process_vm(), typically used to debug a process through /proc/pid/mem or ptrace. This function is needed only for diff --git a/include/linux/mm.h b/include/linux/mm.h index 47a9392..85ba9c2 100644 --- a/include/linux/mm.h +++ b/include/linux/mm.h @@ -251,6 +251,9 @@ struct vm_operations_struct { * writable, if an error is returned it will cause a SIGBUS */ int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf); + /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */ + int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf); + /* called by access_process_vm when get_user_pages() fails, typically * for use by special VMAs that can switch between memory and hardware */ diff --git a/mm/memory.c b/mm/memory.c index 97839f5..6029777 100644 --- a/mm/memory.c +++ b/mm/memory.c @@ -1982,6 +1982,18 @@ static int do_page_mkwrite(struct vm_area_struct *vma, struct page *page, return ret; } +static int do_pfn_mkwrite(struct vm_area_struct *vma, unsigned long address) +{ + struct vm_fault vmf = { + .page = NULL, + .pgoff = linear_page_index(vma, address), + .virtual_address = (void __user *)(address & PAGE_MASK), + .flags = FAULT_FLAG_WRITE | FAULT_FLAG_MKWRITE, + }; + + return vma->vm_ops->pfn_mkwrite(vma, &vmf); +} + /* * This routine handles present pages, when users try to write * to a shared page. It is done by copying the page to a new address @@ -2025,8 +2037,21 @@ static int do_wp_page(struct mm_struct *mm, struct vm_area_struct *vma, * accounting on raw pfn maps. */ if ((vma->vm_flags & (VM_WRITE|VM_SHARED)) == - (VM_WRITE|VM_SHARED)) + (VM_WRITE|VM_SHARED)) { + if (vma->vm_ops && vma->vm_ops->pfn_mkwrite) { + pte_unmap_unlock(page_table, ptl); + ret = do_pfn_mkwrite(vma, address); + if (ret & VM_FAULT_ERROR) + return ret; + page_table = pte_offset_map_lock(mm, pmd, + address, &ptl); + if (!pte_same(*page_table, orig_pte)) { + ret = 0; + goto unlock; + } + } goto reuse; + } goto gotten; } -- 1.9.3