The printf KUnit suite exercises valid %ptR, %ptT, and %ptS inputs,
but it does not cover bad pointers for the time64_t and timespec64
paths.

Add NULL and low-address pointer cases for %ptT and %ptS. The new
checks verify that time_and_date() rejects bad pointers before
dereferencing them and formats them as "(null)" or "(efault)".

Validated with the printf KUnit suite on arm64 QEMU and an
incremental W=1 build of lib/tests/printf_kunit.o.

Signed-off-by: Shuvam Pandey <shuvampandey1@gmail.com>
---
 lib/tests/printf_kunit.c | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/lib/tests/printf_kunit.c b/lib/tests/printf_kunit.c
index f6f21b445ece..c64bfe79652e 100644
--- a/lib/tests/printf_kunit.c
+++ b/lib/tests/printf_kunit.c
@@ -509,6 +509,13 @@ time_and_date(struct kunit *kunittest)
 	struct timespec64 ts = { .tv_sec = t, .tv_nsec = 11235813 };
 
 	test("(%pt?)", "%pt", &tm);
+
+	/* %ptT and %ptS reject bad pointers before dereference. */
+	test("(null)", "%ptT", NULL);
+	test("(efault)", "%ptT", PTR_INVALID);
+	test("(null)", "%ptS", NULL);
+	test("(efault)", "%ptS", PTR_INVALID);
+
 	test("2018-11-26T05:35:43", "%ptR", &tm);
 	test("0118-10-26T05:35:43", "%ptRr", &tm);
 	test("05:35:43|2018-11-26", "%ptRt|%ptRd", &tm, &tm);
-- 
2.50.1 (Apple Git-155)

On Mon, Mar 16, 2026 at 04:43:34PM +0545, Shuvam Pandey wrote:
> The printf KUnit suite exercises valid %ptR, %ptT, and %ptS inputs,
> but it does not cover bad pointers for the time64_t and timespec64
> paths.
> 
> Add NULL and low-address pointer cases for %ptT and %ptS. The new
> checks verify that time_and_date() rejects bad pointers before
> dereferencing them and formats them as "(null)" or "(efault)".
> 
> Validated with the printf KUnit suite on arm64 QEMU and an
> incremental W=1 build of lib/tests/printf_kunit.o.

NAK.

It has nothing to do with %pt.

-- 
With Best Regards,
Andy Shevchenko

On Mon 2026-03-16 16:55:36, Andy Shevchenko wrote:
> On Mon, Mar 16, 2026 at 04:43:34PM +0545, Shuvam Pandey wrote:
> > The printf KUnit suite exercises valid %ptR, %ptT, and %ptS inputs,
> > but it does not cover bad pointers for the time64_t and timespec64
> > paths.
> > 
> > Add NULL and low-address pointer cases for %ptT and %ptS. The new
> > checks verify that time_and_date() rejects bad pointers before
> > dereferencing them and formats them as "(null)" or "(efault)".
> > 
> > Validated with the printf KUnit suite on arm64 QEMU and an
> > incremental W=1 build of lib/tests/printf_kunit.o.
> 
> NAK.
> 
> It has nothing to do with %pt.

Let me play the devil advocate.

There is no single check which would catch bad pointers for
the various %p? format modifiers. It is because some of them
handle them differently, for example, %pK, %pe, or plain %p.

I want to say that wrong pointers passed to %pt? are caught only
because of the explicit check in:

static noinline_for_stack
char *time_and_date(char *buf, char *end, void *ptr, struct printf_spec spec,
		    const char *fmt)
{
	if (check_pointer(&buf, end, ptr, spec))    
		return buf;
[...]
}

So, pointer-modifier-specific checks of wrong input might make sense.

Of course, it would be nice to create some generic solution for
all affected pointer modifiers and not just for "%pt?".

Best Regards,
Petr

On Fri, Mar 20, 2026 at 09:54:17AM +0100, Petr Mladek wrote:
> On Mon 2026-03-16 16:55:36, Andy Shevchenko wrote:
> > On Mon, Mar 16, 2026 at 04:43:34PM +0545, Shuvam Pandey wrote:
> > > The printf KUnit suite exercises valid %ptR, %ptT, and %ptS inputs,
> > > but it does not cover bad pointers for the time64_t and timespec64
> > > paths.
> > > 
> > > Add NULL and low-address pointer cases for %ptT and %ptS. The new
> > > checks verify that time_and_date() rejects bad pointers before
> > > dereferencing them and formats them as "(null)" or "(efault)".
> > > 
> > > Validated with the printf KUnit suite on arm64 QEMU and an
> > > incremental W=1 build of lib/tests/printf_kunit.o.
> > 
> > NAK.
> > 
> > It has nothing to do with %pt.
> 
> Let me play the devil advocate.
> 
> There is no single check which would catch bad pointers for
> the various %p? format modifiers. It is because some of them
> handle them differently, for example, %pK, %pe, or plain %p.
> 
> I want to say that wrong pointers passed to %pt? are caught only
> because of the explicit check in:
> 
> static noinline_for_stack
> char *time_and_date(char *buf, char *end, void *ptr, struct printf_spec spec,
> 		    const char *fmt)
> {
> 	if (check_pointer(&buf, end, ptr, spec))    
> 		return buf;
> [...]
> }
> 
> So, pointer-modifier-specific checks of wrong input might make sense.
> 
> Of course, it would be nice to create some generic solution for
> all affected pointer modifiers and not just for "%pt?".

We should have a check for check_pointer() then.

-- 
With Best Regards,
Andy Shevchenko