On Mon, May 16, 2022 at 08:32:22AM -0600, Keith Busch wrote:
>
> > The XFS_IOC_DIOINFO ioctl from xfs is one, although ugly, option.
> > The other is to export the alignment requirements through statx.
> > We had that whole discussion with the inline crypto direct I/O support,
> > and we really need to tackle it rather sooner than later.
>
> I'll try out assigning one of the spares in 'struct statx' for the purpose. I
> like the interface for that, and looks easy enough to pass along the dma
> alignment requirement through it.
A little less easy than I thought... This is what I'm coming up with, and it's
bad enough that I'm sure there has to be a better way. Specifically concerning
is the new "do_dma()" function below, and was the only way I managed to get the
correct 'struct block_device' whether we're stat'ing a filesystem file or raw
block device file.
---
diff --git a/fs/stat.c b/fs/stat.c
index 5c2c94464e8b..72c0b36599c2 100644
--- a/fs/stat.c
+++ b/fs/stat.c
@@ -17,6 +17,7 @@
#include <linux/syscalls.h>
#include <linux/pagemap.h>
#include <linux/compat.h>
+#include <linux/blkdev.h>
#include <linux/uaccess.h>
#include <asm/unistd.h>
@@ -594,6 +595,7 @@ cp_statx(const struct kstat *stat, struct statx __user *buffer)
tmp.stx_uid = from_kuid_munged(current_user_ns(), stat->uid);
tmp.stx_gid = from_kgid_munged(current_user_ns(), stat->gid);
tmp.stx_mode = stat->mode;
+ tmp.stx_dma = stat->dma;
tmp.stx_ino = stat->ino;
tmp.stx_size = stat->size;
tmp.stx_blocks = stat->blocks;
@@ -615,6 +617,35 @@ cp_statx(const struct kstat *stat, struct statx __user *buffer)
return copy_to_user(buffer, &tmp, sizeof(tmp)) ? -EFAULT : 0;
}
+static void do_dma(int dfd, struct filename *filename, struct kstat *stat)
+{
+ struct open_flags op = {};
+ struct block_device *bdev;
+ struct file *f;
+
+ stat->dma = 511;
+
+ if (S_ISBLK(stat->mode)) {
+ bdev = blkdev_get_no_open(stat->rdev);
+
+ if (bdev) {
+ stat->dma = bdev_dma_alignment(bdev);
+ blkdev_put_no_open(bdev);
+ }
+ return;
+ }
+
+ f = do_filp_open(dfd, filename, &op);
+ if (IS_ERR(f))
+ return;
+
+ bdev = f->f_inode->i_sb->s_bdev;
+ if (bdev)
+ stat->dma = bdev_dma_alignment(bdev);
+ fput(f);
+}
+
int do_statx(int dfd, struct filename *filename, unsigned int flags,
unsigned int mask, struct statx __user *buffer)
{
@@ -630,6 +661,7 @@ int do_statx(int dfd, struct filename *filename, unsigned int flags,
if (error)
return error;
+ do_dma(dfd, filename, &stat);
return cp_statx(&stat, buffer);
}
diff --git a/include/linux/stat.h b/include/linux/stat.h
index 7df06931f25d..0a12c7498aa0 100644
--- a/include/linux/stat.h
+++ b/include/linux/stat.h
@@ -50,6 +50,7 @@ struct kstat {
struct timespec64 btime; /* File creation time */
u64 blocks;
u64 mnt_id;
+ u16 dma;
};
#endif
diff --git a/include/uapi/linux/stat.h b/include/uapi/linux/stat.h
index 1500a0f58041..380820c29c35 100644
--- a/include/uapi/linux/stat.h
+++ b/include/uapi/linux/stat.h
@@ -106,7 +106,7 @@ struct statx {
__u32 stx_uid; /* User ID of owner */
__u32 stx_gid; /* Group ID of owner */
__u16 stx_mode; /* File mode */
- __u16 __spare0[1];
+ __u16 stx_dma; /* DMA alignment */
/* 0x20 */
__u64 stx_ino; /* Inode number */
__u64 stx_size; /* File size */
--
