| Message ID | 20230726-asoc-intel-skylake-remove-deprecated-strncpy-v1-1-020e04184c7d@google.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | ASoC: Intel: Skylake: replace deprecated strncpy with strscpy | expand |
On Wed, Jul 26, 2023 at 09:12:18PM +0000, justinstitt@google.com wrote: > `strncpy` is deprecated for use on NUL-terminated destination strings [1]. > > A suitable replacement is `strscpy` [2] due to the fact that it > guarantees NUL-termination on its destination buffer argument which is > _not_ the case for `strncpy`! > > It was pretty difficult, in this case, to try and figure out whether or > not the destination buffer was zero-initialized. If it is and this > behavior is relied on then perhaps `strscpy_pad` is the preferred > option here. > > Kees was able to help me out and identify the following code snippet > which seems to show that the destination buffer is zero-initialized. > > | skl = devm_kzalloc(&pci->dev, sizeof(*skl), GFP_KERNEL); > > With this information, I opted for `strscpy` since padding is seemingly > not required. We did notice that str_elem->string is 44 bytes, but skl->lib_info[ref_count].name is 128 bytes. If str_elem->string isn't NUL-terminated, this can still hit an over-read condition (though CONFIG_FORTIFY_SOURCE would have caught it both before with strncpy() and now with strscpy()). So I assume it is expected to be NUL-terminated? > Also within this patch is a change to an instance of `x > y - 1` to `x >= y` > which tends to be more robust and readable. Consider, for instance, if `y` was > somehow `INT_MIN`. I'd split this change into a separate patch -- it's logically unrelated (but seems a reasonable cleanup). > > [1]: www.kernel.org/doc/html/latest/process/deprecated.html#strncpy-on-nul-terminated-strings > [2]: manpages.debian.org/testing/linux-manual-4.8/strscpy.9.en.html > > Link: https://github.com/KSPP/linux/issues/90 > Suggested-by: Kees Cook <keescook@chromium.org> > Signed-off-by: Justin Stitt <justinstitt@google.com> > --- > sound/soc/intel/skylake/skl-topology.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/sound/soc/intel/skylake/skl-topology.c b/sound/soc/intel/skylake/skl-topology.c > index 96cfebded072..67f08ec3a2ea 100644 > --- a/sound/soc/intel/skylake/skl-topology.c > +++ b/sound/soc/intel/skylake/skl-topology.c > @@ -3154,12 +3154,12 @@ static int skl_tplg_fill_str_mfest_tkn(struct device *dev, > > switch (str_elem->token) { > case SKL_TKN_STR_LIB_NAME: > - if (ref_count > skl->lib_count - 1) { > + if (ref_count >= skl->lib_count) { > ref_count = 0; > return -EINVAL; > } > > - strncpy(skl->lib_info[ref_count].name, > + strscpy(skl->lib_info[ref_count].name, > str_elem->string, > ARRAY_SIZE(skl->lib_info[ref_count].name)); > ref_count++; > > --- > base-commit: 0b4a9fdc9317440a71d4d4c264a5650bf4a90f3c > change-id: 20230726-asoc-intel-skylake-remove-deprecated-strncpy-9dbcfc26040c > > Best regards, > -- > Justin Stitt <justinstitt@google.com> >
On 7/27/2023 12:34 AM, Kees Cook wrote: > On Wed, Jul 26, 2023 at 09:12:18PM +0000, justinstitt@google.com wrote: >> `strncpy` is deprecated for use on NUL-terminated destination strings [1]. >> >> A suitable replacement is `strscpy` [2] due to the fact that it >> guarantees NUL-termination on its destination buffer argument which is >> _not_ the case for `strncpy`! >> >> It was pretty difficult, in this case, to try and figure out whether or >> not the destination buffer was zero-initialized. If it is and this >> behavior is relied on then perhaps `strscpy_pad` is the preferred >> option here. >> >> Kees was able to help me out and identify the following code snippet >> which seems to show that the destination buffer is zero-initialized. >> >> | skl = devm_kzalloc(&pci->dev, sizeof(*skl), GFP_KERNEL); >> >> With this information, I opted for `strscpy` since padding is seemingly >> not required. > > We did notice that str_elem->string is 44 bytes, but > skl->lib_info[ref_count].name is 128 bytes. If str_elem->string isn't > NUL-terminated, this can still hit an over-read condition (though > CONFIG_FORTIFY_SOURCE would have caught it both before with strncpy() > and now with strscpy()). So I assume it is expected to be > NUL-terminated? > Yes it is a filename of additional library which can be loaded, topology UAPI only allows for passing 44 bytes long strings per string token (see snd_soc_tplg_vendor_array -> union -> string flex array -> snd_soc_tplg_vendor_string_elem -> SNDRV_CTL_ELEM_ID_NAME_MAXLEN), so we could also change length of skl->lib_info[ref_count].name and potentially save few bytes. And looking at it again I also think that we should not copy destination size number of bytes, by which I mean ARRAY_SIZE(skl->lib_info[ref_count].name), which is 128 in this case... so either need to change destination buffer size to be same as topology field or calculate it differently.
On Fri, Jul 28, 2023 at 09:25:24AM +0200, Amadeusz Sławiński wrote: > On 7/27/2023 12:34 AM, Kees Cook wrote: > > On Wed, Jul 26, 2023 at 09:12:18PM +0000, justinstitt@google.com wrote: > > > `strncpy` is deprecated for use on NUL-terminated destination strings [1]. > > > > > > A suitable replacement is `strscpy` [2] due to the fact that it > > > guarantees NUL-termination on its destination buffer argument which is > > > _not_ the case for `strncpy`! > > > > > > It was pretty difficult, in this case, to try and figure out whether or > > > not the destination buffer was zero-initialized. If it is and this > > > behavior is relied on then perhaps `strscpy_pad` is the preferred > > > option here. > > > > > > Kees was able to help me out and identify the following code snippet > > > which seems to show that the destination buffer is zero-initialized. > > > > > > | skl = devm_kzalloc(&pci->dev, sizeof(*skl), GFP_KERNEL); > > > > > > With this information, I opted for `strscpy` since padding is seemingly > > > not required. > > > > We did notice that str_elem->string is 44 bytes, but > > skl->lib_info[ref_count].name is 128 bytes. If str_elem->string isn't > > NUL-terminated, this can still hit an over-read condition (though > > CONFIG_FORTIFY_SOURCE would have caught it both before with strncpy() > > and now with strscpy()). So I assume it is expected to be > > NUL-terminated? > > > > Yes it is a filename of additional library which can be loaded, topology > UAPI only allows for passing 44 bytes long strings per string token (see > snd_soc_tplg_vendor_array -> union -> string flex array -> > snd_soc_tplg_vendor_string_elem -> SNDRV_CTL_ELEM_ID_NAME_MAXLEN), so we Thanks for the details! And just to confirm, these are (expected to be) NUL-terminated? > could also change length of > skl->lib_info[ref_count].name and potentially save few bytes. And looking at > it again I also think that we should not copy destination size number of > bytes, by which I mean ARRAY_SIZE(skl->lib_info[ref_count].name), which is > 128 in this case... so either need to change destination buffer size to be > same as topology field or calculate it differently. If the source is NUL-terminated, it's fine as-is. (And CONFIG_FORTIFY_SOURCE will catch problems if not.)
diff --git a/sound/soc/intel/skylake/skl-topology.c b/sound/soc/intel/skylake/skl-topology.c index 96cfebded072..67f08ec3a2ea 100644 --- a/sound/soc/intel/skylake/skl-topology.c +++ b/sound/soc/intel/skylake/skl-topology.c @@ -3154,12 +3154,12 @@ static int skl_tplg_fill_str_mfest_tkn(struct device *dev, switch (str_elem->token) { case SKL_TKN_STR_LIB_NAME: - if (ref_count > skl->lib_count - 1) { + if (ref_count >= skl->lib_count) { ref_count = 0; return -EINVAL; } - strncpy(skl->lib_info[ref_count].name, + strscpy(skl->lib_info[ref_count].name, str_elem->string, ARRAY_SIZE(skl->lib_info[ref_count].name)); ref_count++;
`strncpy` is deprecated for use on NUL-terminated destination strings [1]. A suitable replacement is `strscpy` [2] due to the fact that it guarantees NUL-termination on its destination buffer argument which is _not_ the case for `strncpy`! It was pretty difficult, in this case, to try and figure out whether or not the destination buffer was zero-initialized. If it is and this behavior is relied on then perhaps `strscpy_pad` is the preferred option here. Kees was able to help me out and identify the following code snippet which seems to show that the destination buffer is zero-initialized. | skl = devm_kzalloc(&pci->dev, sizeof(*skl), GFP_KERNEL); With this information, I opted for `strscpy` since padding is seemingly not required. Also within this patch is a change to an instance of `x > y - 1` to `x >= y` which tends to be more robust and readable. Consider, for instance, if `y` was somehow `INT_MIN`. [1]: www.kernel.org/doc/html/latest/process/deprecated.html#strncpy-on-nul-terminated-strings [2]: manpages.debian.org/testing/linux-manual-4.8/strscpy.9.en.html Link: https://github.com/KSPP/linux/issues/90 Suggested-by: Kees Cook <keescook@chromium.org> Signed-off-by: Justin Stitt <justinstitt@google.com> --- sound/soc/intel/skylake/skl-topology.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) --- base-commit: 0b4a9fdc9317440a71d4d4c264a5650bf4a90f3c change-id: 20230726-asoc-intel-skylake-remove-deprecated-strncpy-9dbcfc26040c Best regards,