| Message ID | Zv9aEoDo1k36Hg3H@mail.google.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [next] ARM: Replace snprintf() with the safer scnprintf() variant | expand |
On Fri, Oct 04, 2024 at 03:59:30PM +1300, Paulo Miguel Almeida wrote: > There is a general misunderstanding amongst engineers that {v}snprintf() > returns the length of the data *actually* encoded into the destination > array. However, as per the C99 standard {v}snprintf() really returns > the length of the data that *would have been* written if there were > enough space for it. This misunderstanding has led to buffer-overruns > in the past. It's generally considered safer to use the {v}scnprintf() > variants in their place (or even sprintf() in simple cases). So, basically, it's unsafe to use the result of (v)snprintf(). So why do we need to change locations that do not use the result? This patch is mere noise. Sorry, I won't be applying it.
On Fri, Oct 04, 2024 at 10:09:50AM +0100, Russell King (Oracle) wrote: > On Fri, Oct 04, 2024 at 03:59:30PM +1300, Paulo Miguel Almeida wrote: > > There is a general misunderstanding amongst engineers that {v}snprintf() > > returns the length of the data *actually* encoded into the destination > > array. However, as per the C99 standard {v}snprintf() really returns > > the length of the data that *would have been* written if there were > > enough space for it. This misunderstanding has led to buffer-overruns > > in the past. It's generally considered safer to use the {v}scnprintf() > > variants in their place (or even sprintf() in simple cases). > > So, basically, it's unsafe to use the result of (v)snprintf(). So why > do we need to change locations that do not use the result? > > This patch is mere noise. Sorry, I won't be applying it. > Thanks for taking the time to review this patch. My take on this is that it boils down to nipping it in the bud proactively, so if the result starts being used, no one has to remember to change from scprint() to scnprint(), which can be easy to miss. There have been other instances where the result wasn't being used, yet the patch was still accepted [1] — should that help sway your opinion. :) I understand that each maintainer has different approaches to what they deem trivial patches, but I hope you reconsider accepting this one. That said, if there is no utilization of snprint(), we could eventually deprecate or remove the function altogether. [1] https://lore.kernel.org/lkml/20231213164246.1021885-11-lee@kernel.org/ Paulo A.
diff --git a/arch/arm/kernel/process.c b/arch/arm/kernel/process.c index e16ed102960c..9d768a93fb1c 100644 --- a/arch/arm/kernel/process.c +++ b/arch/arm/kernel/process.c @@ -186,8 +186,8 @@ void __show_regs(struct pt_regs *regs) unsigned int transbase; asm("mrc p15, 0, %0, c2, c0\n\t" : "=r" (transbase)); - snprintf(buf, sizeof(buf), " Table: %08x DAC: %08x", - transbase, domain); + scnprintf(buf, sizeof(buf), " Table: %08x DAC: %08x", + transbase, domain); } #endif asm("mrc p15, 0, %0, c1, c0\n" : "=r" (ctrl));
There is a general misunderstanding amongst engineers that {v}snprintf() returns the length of the data *actually* encoded into the destination array. However, as per the C99 standard {v}snprintf() really returns the length of the data that *would have been* written if there were enough space for it. This misunderstanding has led to buffer-overruns in the past. It's generally considered safer to use the {v}scnprintf() variants in their place (or even sprintf() in simple cases). Link: https://lwn.net/Articles/69419/ Link: https://github.com/KSPP/linux/issues/105 Signed-off-by: Paulo Miguel Almeida <paulo.miguel.almeida.rodenas@gmail.com> --- arch/arm/kernel/process.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)