The users of filemap_write_and_wait_range() and file_write_and_wait_range()
interfaces should set the lend parameter to LLONG_MAX, rather than -1, to
indicate they want to writeback to the very end-of-file, as several kernel
code paths are checking the 'wbc->range_end == LLONG_MAX' conditions.
Signed-off-by: Jiachen Zhang <[email protected]>
---
mm/filemap.c | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/mm/filemap.c b/mm/filemap.c
index 65eee6ec1066..c6d066a39425 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -661,7 +661,8 @@ EXPORT_SYMBOL_GPL(filemap_range_has_writeback);
* Write out and wait upon file offsets lstart->lend, inclusive.
*
* Note that @lend is inclusive (describes the last byte to be written) so
- * that this function can be used to write to the very end-of-file (end = -1).
+ * that this function can be used to write to the very end-of-file (@lend =
+ * LLONG_MAX).
*
* Return: error status of the address space.
*/
@@ -758,7 +759,8 @@ EXPORT_SYMBOL(file_check_and_advance_wb_err);
* Write out and wait upon file offsets lstart->lend, inclusive.
*
* Note that @lend is inclusive (describes the last byte to be written) so
- * that this function can be used to write to the very end-of-file (end = -1).
+ * that this function can be used to write to the very end-of-file (@lend =
+ * LLONG_MAX).
*
* After writing out and waiting on the data, we check and advance the
* f_wb_err cursor to the latest value, and return any errors detected there.
--
2.20.1
On Fri, 25 Nov 2022 15:09:59 +0800 Jiachen Zhang <[email protected]> wrote:
> The users of filemap_write_and_wait_range() and file_write_and_wait_range()
> interfaces should set the lend parameter to LLONG_MAX, rather than -1, to
> indicate they want to writeback to the very end-of-file, as several kernel
> code paths are checking the 'wbc->range_end == LLONG_MAX' conditions.
Unclear. LLONG_MAX differs from -1 on 64-bit and differs differently
on 32-bit.
> --- a/mm/filemap.c
> +++ b/mm/filemap.c
> @@ -661,7 +661,8 @@ EXPORT_SYMBOL_GPL(filemap_range_has_writeback);
> * Write out and wait upon file offsets lstart->lend, inclusive.
> *
> * Note that @lend is inclusive (describes the last byte to be written) so
> - * that this function can be used to write to the very end-of-file (end = -1).
> + * that this function can be used to write to the very end-of-file (@lend =
> + * LLONG_MAX).
> *
The write(2) manpage says "According to POSIX.1, if count is greater
than SSIZE_MAX, the result is implementation-defined; see NOTES for the
upper limit on Linux." And filemap_fdatawrite_wbc() enforces LONG_MAX,
which differs from LLONG_MAX on 32-bit.
I suspect more research is needed here.
On Sat, Nov 26, 2022 at 8:52 AM Andrew Morton <[email protected]> wrote:
>
> On Fri, 25 Nov 2022 15:09:59 +0800 Jiachen Zhang <[email protected]> wrote:
>
> > The users of filemap_write_and_wait_range() and file_write_and_wait_range()
> > interfaces should set the lend parameter to LLONG_MAX, rather than -1, to
> > indicate they want to writeback to the very end-of-file, as several kernel
> > code paths are checking the 'wbc->range_end == LLONG_MAX' conditions.
>
> Unclear. LLONG_MAX differs from -1 on 64-bit and differs differently
> on 32-bit.
>
I think whether using -1 or LLONG_MAX causes no difference if there is
no other code comparing 'wbc->range_end == LLONG_MAX'. There is no
case in the kernel code using -1 for now, but maybe we'd better fix
the misleading comments to prevent future misuse.
> > --- a/mm/filemap.c
> > +++ b/mm/filemap.c
> > @@ -661,7 +661,8 @@ EXPORT_SYMBOL_GPL(filemap_range_has_writeback);
> > * Write out and wait upon file offsets lstart->lend, inclusive.
> > *
> > * Note that @lend is inclusive (describes the last byte to be written) so
> > - * that this function can be used to write to the very end-of-file (end = -1).
> > + * that this function can be used to write to the very end-of-file (@lend =
> > + * LLONG_MAX).
> > *
>
> The write(2) manpage says "According to POSIX.1, if count is greater
> than SSIZE_MAX, the result is implementation-defined; see NOTES for the
> upper limit on Linux." And filemap_fdatawrite_wbc() enforces LONG_MAX,
> which differs from LLONG_MAX on 32-bit.
>
> I suspect more research is needed here.
The reason 'wbc.nr_to_write' might be set to LONG_MAX for
filemap_fdatawrite_wbc() might be because 'nr_to_write' is defined as
the 'long' type. Maybe it should be fine as 'lend' and 'range_end' are
defined as type 'off_t'.
Thanks,
Jiachen