block: revert runtime dax control of the raw block device
authorDan Williams <dan.j.williams@intel.com>
Fri, 29 Jan 2016 04:13:39 +0000 (20:13 -0800)
committerDan Williams <dan.j.williams@intel.com>
Sat, 30 Jan 2016 21:35:31 +0000 (13:35 -0800)
Dynamically enabling DAX requires that the page cache first be flushed
and invalidated.  This must occur atomically with the change of DAX mode
otherwise we confuse the fsync/msync tracking and violate data
durability guarantees.  Eliminate the possibilty of DAX-disabled to
DAX-enabled transitions for now and revisit this for the next cycle.

Cc: Jan Kara <jack@suse.com>
Cc: Jeff Moyer <jmoyer@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Matthew Wilcox <willy@linux.intel.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
block/ioctl.c
fs/block_dev.c
include/linux/fs.h
include/uapi/linux/fs.h

index 77f5d17779d6875d0595f23410cbd61116491450..d8996bbd7f12805b9c74429813f5b3ee3d744bdc 100644 (file)
@@ -434,42 +434,6 @@ bool blkdev_dax_capable(struct block_device *bdev)
 
        return true;
 }
-
-static int blkdev_daxset(struct block_device *bdev, unsigned long argp)
-{
-       unsigned long arg;
-       int rc = 0;
-
-       if (!capable(CAP_SYS_ADMIN))
-               return -EACCES;
-
-       if (get_user(arg, (int __user *)(argp)))
-               return -EFAULT;
-       arg = !!arg;
-       if (arg == !!(bdev->bd_inode->i_flags & S_DAX))
-               return 0;
-
-       if (arg)
-               arg = S_DAX;
-
-       if (arg && !blkdev_dax_capable(bdev))
-               return -ENOTTY;
-
-       inode_lock(bdev->bd_inode);
-       if (bdev->bd_map_count == 0)
-               inode_set_flags(bdev->bd_inode, arg, S_DAX);
-       else
-               rc = -EBUSY;
-       inode_unlock(bdev->bd_inode);
-       return rc;
-}
-#else
-static int blkdev_daxset(struct block_device *bdev, int arg)
-{
-       if (arg)
-               return -ENOTTY;
-       return 0;
-}
 #endif
 
 static int blkdev_flushbuf(struct block_device *bdev, fmode_t mode,
@@ -634,8 +598,6 @@ int blkdev_ioctl(struct block_device *bdev, fmode_t mode, unsigned cmd,
        case BLKTRACESETUP:
        case BLKTRACETEARDOWN:
                return blk_trace_ioctl(bdev, cmd, argp);
-       case BLKDAXSET:
-               return blkdev_daxset(bdev, arg);
        case BLKDAXGET:
                return put_int(arg, !!(bdev->bd_inode->i_flags & S_DAX));
                break;
index 7b9cd49622b132f5f71e6557f05d0a2591a37c16..afb43748436201e29343a923077976d01a2344fd 100644 (file)
@@ -1736,37 +1736,13 @@ static int blkdev_dax_pmd_fault(struct vm_area_struct *vma, unsigned long addr,
        return __dax_pmd_fault(vma, addr, pmd, flags, blkdev_get_block, NULL);
 }
 
-static void blkdev_vm_open(struct vm_area_struct *vma)
-{
-       struct inode *bd_inode = bdev_file_inode(vma->vm_file);
-       struct block_device *bdev = I_BDEV(bd_inode);
-
-       inode_lock(bd_inode);
-       bdev->bd_map_count++;
-       inode_unlock(bd_inode);
-}
-
-static void blkdev_vm_close(struct vm_area_struct *vma)
-{
-       struct inode *bd_inode = bdev_file_inode(vma->vm_file);
-       struct block_device *bdev = I_BDEV(bd_inode);
-
-       inode_lock(bd_inode);
-       bdev->bd_map_count--;
-       inode_unlock(bd_inode);
-}
-
 static const struct vm_operations_struct blkdev_dax_vm_ops = {
-       .open           = blkdev_vm_open,
-       .close          = blkdev_vm_close,
        .fault          = blkdev_dax_fault,
        .pmd_fault      = blkdev_dax_pmd_fault,
        .pfn_mkwrite    = blkdev_dax_fault,
 };
 
 static const struct vm_operations_struct blkdev_default_vm_ops = {
-       .open           = blkdev_vm_open,
-       .close          = blkdev_vm_close,
        .fault          = filemap_fault,
        .map_pages      = filemap_map_pages,
 };
@@ -1774,18 +1750,14 @@ static const struct vm_operations_struct blkdev_default_vm_ops = {
 static int blkdev_mmap(struct file *file, struct vm_area_struct *vma)
 {
        struct inode *bd_inode = bdev_file_inode(file);
-       struct block_device *bdev = I_BDEV(bd_inode);
 
        file_accessed(file);
-       inode_lock(bd_inode);
-       bdev->bd_map_count++;
        if (IS_DAX(bd_inode)) {
                vma->vm_ops = &blkdev_dax_vm_ops;
                vma->vm_flags |= VM_MIXEDMAP | VM_HUGEPAGE;
        } else {
                vma->vm_ops = &blkdev_default_vm_ops;
        }
-       inode_unlock(bd_inode);
 
        return 0;
 }
index b10002d4a5f568ac4d409bb4bd64225aced57e11..ae681002100a1fb8401e934a99f40dedf039cd7d 100644 (file)
@@ -484,9 +484,6 @@ struct block_device {
        int                     bd_fsfreeze_count;
        /* Mutex for freeze */
        struct mutex            bd_fsfreeze_mutex;
-#ifdef CONFIG_FS_DAX
-       int                     bd_map_count;
-#endif
 };
 
 /*
index 41e0433b4a8398b3dc3da6d2d6aacb8e46287150..149bec83a907515dccea4425acf3b71d8d18410f 100644 (file)
@@ -222,7 +222,6 @@ struct fsxattr {
 #define BLKSECDISCARD _IO(0x12,125)
 #define BLKROTATIONAL _IO(0x12,126)
 #define BLKZEROOUT _IO(0x12,127)
-#define BLKDAXSET _IO(0x12,128)
 #define BLKDAXGET _IO(0x12,129)
 
 #define BMAP_IOCTL 1           /* obsolete - kept for compatibility */