strncpy() is deprecated for use on NUL-terminated destination strings
[1] and as such we should prefer more robust and less ambiguous string
interfaces.
@query is already marked as __nonstring and doesn't need to be
NUL-terminated. Due to this, we don't need to use a string API here
(especially a deprecated one). Let's have our stack allocation also
zero-initialize so that we can just perform a standard memcpy. Since the
code now speaks for itself we can drop the comment. A memcpy on a
__nonstring buffer explains everything that this comment talks about.
Link: https://www.kernel.org/doc/html/latest/process/deprecated.html#strncpy-on-nul-terminated-strings [1]
Link: https://manpages.debian.org/testing/linux-manual-4.8/strscpy.9.en.html [2]
Link: https://github.com/KSPP/linux/issues/90
Cc: [email protected]
Signed-off-by: Justin Stitt <[email protected]>
---
Note: build-tested only.
Found with: $ rg "strncpy\("
---
drivers/soc/qcom/cmd-db.c | 9 ++-------
1 file changed, 2 insertions(+), 7 deletions(-)
diff --git a/drivers/soc/qcom/cmd-db.c b/drivers/soc/qcom/cmd-db.c
index a5fd68411bed..512556366a3e 100644
--- a/drivers/soc/qcom/cmd-db.c
+++ b/drivers/soc/qcom/cmd-db.c
@@ -141,18 +141,13 @@ static int cmd_db_get_header(const char *id, const struct entry_header **eh,
const struct rsc_hdr *rsc_hdr;
const struct entry_header *ent;
int ret, i, j;
- u8 query[sizeof(ent->id)] __nonstring;
+ u8 query[sizeof(ent->id)] __nonstring = { 0 };
ret = cmd_db_ready();
if (ret)
return ret;
- /*
- * Pad out query string to same length as in DB. NOTE: the output
- * query string is not necessarily '\0' terminated if it bumps up
- * against the max size. That's OK and expected.
- */
- strncpy(query, id, sizeof(query));
+ memcpy(query, id, sizeof(query));
for (i = 0; i < MAX_SLV_ID; i++) {
rsc_hdr = &cmd_db_header->header[i];
---
base-commit: fe46a7dd189e25604716c03576d05ac8a5209743
change-id: 20240314-strncpy-drivers-soc-qcom-cmd-db-c-284f3abaabb8
Best regards,
--
Justin Stitt <[email protected]>
On Thu, Mar 14, 2024 at 10:29:37PM +0000, Justin Stitt wrote:
> strncpy() is deprecated for use on NUL-terminated destination strings
> [1] and as such we should prefer more robust and less ambiguous string
> interfaces.
>
> @query is already marked as __nonstring and doesn't need to be
> NUL-terminated. Due to this, we don't need to use a string API here
> (especially a deprecated one). Let's have our stack allocation also
> zero-initialize so that we can just perform a standard memcpy. Since the
> code now speaks for itself we can drop the comment. A memcpy on a
> __nonstring buffer explains everything that this comment talks about.
>
> Link: https://www.kernel.org/doc/html/latest/process/deprecated.html#strncpy-on-nul-terminated-strings [1]
> Link: https://manpages.debian.org/testing/linux-manual-4.8/strscpy.9.en.html [2]
> Link: https://github.com/KSPP/linux/issues/90
> Cc: [email protected]
> Signed-off-by: Justin Stitt <[email protected]>
> ---
> Note: build-tested only.
>
> Found with: $ rg "strncpy\("
> ---
> drivers/soc/qcom/cmd-db.c | 9 ++-------
> 1 file changed, 2 insertions(+), 7 deletions(-)
>
> diff --git a/drivers/soc/qcom/cmd-db.c b/drivers/soc/qcom/cmd-db.c
> index a5fd68411bed..512556366a3e 100644
> --- a/drivers/soc/qcom/cmd-db.c
> +++ b/drivers/soc/qcom/cmd-db.c
> @@ -141,18 +141,13 @@ static int cmd_db_get_header(const char *id, const struct entry_header **eh,
> const struct rsc_hdr *rsc_hdr;
> const struct entry_header *ent;
> int ret, i, j;
> - u8 query[sizeof(ent->id)] __nonstring;
> + u8 query[sizeof(ent->id)] __nonstring = { 0 };
>
> ret = cmd_db_ready();
> if (ret)
> return ret;
>
> - /*
> - * Pad out query string to same length as in DB. NOTE: the output
> - * query string is not necessarily '\0' terminated if it bumps up
> - * against the max size. That's OK and expected.
> - */
> - strncpy(query, id, sizeof(query));
> + memcpy(query, id, sizeof(query));
Hm, no, this isn't right. We do want to stop copying at the first NUL
character, but we don't care about truncation. e.g. imagine if "id" was
a 3 character string followed by other bytes in memory. We'd copy beyond
the end of "id" into query, and the later memcmp()s would start failing.
I think what you want here is:
strtomem(query, id);
-Kees
>
> for (i = 0; i < MAX_SLV_ID; i++) {
> rsc_hdr = &cmd_db_header->header[i];
>
> ---
> base-commit: fe46a7dd189e25604716c03576d05ac8a5209743
> change-id: 20240314-strncpy-drivers-soc-qcom-cmd-db-c-284f3abaabb8
>
> Best regards,
> --
> Justin Stitt <[email protected]>
>
>
--
Kees Cook
On Mon, Mar 18, 2024 at 03:47:38PM -0700, Justin Stitt wrote:
> Gotcha, I was operating under the assumption that we needed to know
> the size of id at compile time. Apparently __builtin_object_size(_, 1)
> will return SIZE_T_MAX if we don't know the size of something. Sending
> a v2.
Yeah, it only requires to know the compile-time size of the destination
buffer.
--
Kees Cook
On Mon, Mar 18, 2024 at 2:52 PM Kees Cook <[email protected]> wrote:
>
> On Thu, Mar 14, 2024 at 10:29:37PM +0000, Justin Stitt wrote:
> > strncpy() is deprecated for use on NUL-terminated destination strings
> > [1] and as such we should prefer more robust and less ambiguous string
> > interfaces.
> >
> > @query is already marked as __nonstring and doesn't need to be
> > NUL-terminated. Due to this, we don't need to use a string API here
> > (especially a deprecated one). Let's have our stack allocation also
> > zero-initialize so that we can just perform a standard memcpy. Since the
> > code now speaks for itself we can drop the comment. A memcpy on a
> > __nonstring buffer explains everything that this comment talks about.
> >
> > Link: https://www.kernel.org/doc/html/latest/process/deprecated.html#strncpy-on-nul-terminated-strings [1]
> > Link: https://manpages.debian.org/testing/linux-manual-4.8/strscpy.9.enhtml [2]
> > Link: https://github.com/KSPP/linux/issues/90
> > Cc: [email protected]
> > Signed-off-by: Justin Stitt <[email protected]>
> > ---
> > Note: build-tested only.
> >
> > Found with: $ rg "strncpy\("
> > ---
> > drivers/soc/qcom/cmd-db.c | 9 ++-------
> > 1 file changed, 2 insertions(+), 7 deletions(-)
> >
> > diff --git a/drivers/soc/qcom/cmd-db.c b/drivers/soc/qcom/cmd-db.c
> > index a5fd68411bed..512556366a3e 100644
> > --- a/drivers/soc/qcom/cmd-db.c
> > +++ b/drivers/soc/qcom/cmd-db.c
> > @@ -141,18 +141,13 @@ static int cmd_db_get_header(const char *id, const struct entry_header **eh,
> > const struct rsc_hdr *rsc_hdr;
> > const struct entry_header *ent;
> > int ret, i, j;
> > - u8 query[sizeof(ent->id)] __nonstring;
> > + u8 query[sizeof(ent->id)] __nonstring = { 0 };
> >
> > ret = cmd_db_ready();
> > if (ret)
> > return ret;
> >
> > - /*
> > - * Pad out query string to same length as in DB. NOTE: the output
> > - * query string is not necessarily '\0' terminated if it bumps up
> > - * against the max size. That's OK and expected.
> > - */
> > - strncpy(query, id, sizeof(query));
> > + memcpy(query, id, sizeof(query));
>
> Hm, no, this isn't right. We do want to stop copying at the first NUL
> character, but we don't care about truncation. e.g. imagine if "id" was
> a 3 character string followed by other bytes in memory. We'd copy beyond
> the end of "id" into query, and the later memcmp()s would start failing.
> I think what you want here is:
>
> strtomem(query, id);
Gotcha, I was operating under the assumption that we needed to know
the size of id at compile time. Apparently __builtin_object_size(_, 1)
will return SIZE_T_MAX if we don't know the size of something. Sending
a v2.
>
> -Kees
>
> >
> > for (i = 0; i < MAX_SLV_ID; i++) {
> > rsc_hdr = &cmd_db_header->header[i];
> >
> > ---
> > base-commit: fe46a7dd189e25604716c03576d05ac8a5209743
> > change-id: 20240314-strncpy-drivers-soc-qcom-cmd-db-c-284f3abaabb8
> >
> > Best regards,
> > --
> > Justin Stitt <[email protected]>
> >
> >
>
> --
> Kees Cook