On 23/08/2024 14:05, Darrick J. Wong wrote:

On Fri, Aug 23, 2024 at 09:11:22AM -0700, Bart Van Assche wrote:

quoted

On 8/23/24 7:26 AM, Guilherme G. Piccoli wrote:

quoted

+	bdev_allow_write_mounted=
+			Format: <bool>
+			Control the ability of directly writing to mounted block
+			devices' page cache, i.e., allow / disallow writes that
+			bypasses the FS. This was implemented as a means to
+			prevent fuzzers from crashing the kernel by overwriting
+			the metadata underneath a mounted FS without its awareness.
+			This also prevents destructive formatting of mounted
+			filesystems by naive storage tooling that don't use
+			O_EXCL. Default is Y and can be changed through the
+			Kconfig option CONFIG_BLK_DEV_WRITE_MOUNTED.
+

Does this flag also affect direct I/O? If so, does this mean that the
reference to the page cache should be left out?

I think it does affect directio, since the validation is done at open
time via bdev_may_open, right?

Indeed, good point! It does affect direct I/O, I've just tested in
6.11-rc3 using dd (with and w/o oflags=direct), same result.

I'll resend, dropping the mention to page cache - thanks!

Guilherme