| Message ID | 20230831162227.2307863-1-smostafa@google.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | Revert "arm64/sysreg: refactor deprecated strncpy" | expand |
Hi Mostafa, On Thu, 31 Aug 2023 17:22:27 +0100, Mostafa Saleh <smostafa@google.com> wrote: > > This reverts commit d232606773a0b09ec7f1ffc25f63abe801d011fd. > > Using strscpy is not correct in this context and the commit > assumption is not right "strncpy is deprecated for use on > NUL-terminated destination strings". > > strncpy is used here to copy parts of the string(cmdline) separated > by spaces into the buffer and not a NULL terminated string. > > This breaks the arm options "kvm-arm.mode=protected, arm64.nobti ..." > > Signed-off-by: Mostafa Saleh <smostafa@google.com> > --- > arch/arm64/kernel/idreg-override.c | 6 +++--- > 1 file changed, 3 insertions(+), 3 deletions(-) > > diff --git a/arch/arm64/kernel/idreg-override.c b/arch/arm64/kernel/idreg-override.c > index aee12c75b738..2fe2491b692c 100644 > --- a/arch/arm64/kernel/idreg-override.c > +++ b/arch/arm64/kernel/idreg-override.c > @@ -262,9 +262,9 @@ static __init void __parse_cmdline(const char *cmdline, bool parse_aliases) > if (!len) > return; > > - len = strscpy(buf, cmdline, ARRAY_SIZE(buf)); > - if (len == -E2BIG) > - len = ARRAY_SIZE(buf) - 1; > + len = min(len, ARRAY_SIZE(buf) - 1); > + strncpy(buf, cmdline, len); > + buf[len] = 0; > > if (strcmp(buf, "--") == 0) > return; Instead of completely reverting the patch, maybe something like the hack below (completely untested), so that we can still get rid of another instance of strncpy(), and yet bring back some sanity in the logic? Thanks, M. diff --git a/arch/arm64/kernel/idreg-override.c b/arch/arm64/kernel/idreg-override.c index aee12c75b738..be5c778a3f14 100644 --- a/arch/arm64/kernel/idreg-override.c +++ b/arch/arm64/kernel/idreg-override.c @@ -262,9 +262,8 @@ static __init void __parse_cmdline(const char *cmdline, bool parse_aliases) if (!len) return; - len = strscpy(buf, cmdline, ARRAY_SIZE(buf)); - if (len == -E2BIG) - len = ARRAY_SIZE(buf) - 1; + len = min(len, ARRAY_SIZE(buf) - 1); + strscpy(buf, cmdline, len); if (strcmp(buf, "--") == 0) return;
Hi Marc, On Fri, Sep 01, 2023 at 12:24:45PM +0100, Marc Zyngier wrote: > Hi Mostafa, > > On Thu, 31 Aug 2023 17:22:27 +0100, > Mostafa Saleh <smostafa@google.com> wrote: > > > > This reverts commit d232606773a0b09ec7f1ffc25f63abe801d011fd. > > > > Using strscpy is not correct in this context and the commit > > assumption is not right "strncpy is deprecated for use on > > NUL-terminated destination strings". > > > > strncpy is used here to copy parts of the string(cmdline) separated > > by spaces into the buffer and not a NULL terminated string. > > > > This breaks the arm options "kvm-arm.mode=protected, arm64.nobti ..." > > > > Signed-off-by: Mostafa Saleh <smostafa@google.com> > > --- > > arch/arm64/kernel/idreg-override.c | 6 +++--- > > 1 file changed, 3 insertions(+), 3 deletions(-) > > > > diff --git a/arch/arm64/kernel/idreg-override.c b/arch/arm64/kernel/idreg-override.c > > index aee12c75b738..2fe2491b692c 100644 > > --- a/arch/arm64/kernel/idreg-override.c > > +++ b/arch/arm64/kernel/idreg-override.c > > @@ -262,9 +262,9 @@ static __init void __parse_cmdline(const char *cmdline, bool parse_aliases) > > if (!len) > > return; > > > > - len = strscpy(buf, cmdline, ARRAY_SIZE(buf)); > > - if (len == -E2BIG) > > - len = ARRAY_SIZE(buf) - 1; > > + len = min(len, ARRAY_SIZE(buf) - 1); > > + strncpy(buf, cmdline, len); > > + buf[len] = 0; > > > > if (strcmp(buf, "--") == 0) > > return; > > Instead of completely reverting the patch, maybe something like the > hack below (completely untested), so that we can still get rid of > another instance of strncpy(), and yet bring back some sanity in the > logic? I was thinking of something similar, but I see we limit the len anyway by the buffer size - 1 and force the NUL at the end so it should be safe, unless the goal is to get rid of strncpy all the way, in this case we can do it as you proposed. There is also a V3 of the original patch which uses memcpy instead. https://lore.kernel.org/all/20230831-strncpy-arch-arm64-v3-1-cdbb1e7ea5e1@google.com/ > Thanks, > > M. > > diff --git a/arch/arm64/kernel/idreg-override.c b/arch/arm64/kernel/idreg-override.c > index aee12c75b738..be5c778a3f14 100644 > --- a/arch/arm64/kernel/idreg-override.c > +++ b/arch/arm64/kernel/idreg-override.c > @@ -262,9 +262,8 @@ static __init void __parse_cmdline(const char *cmdline, bool parse_aliases) > if (!len) > return; > > - len = strscpy(buf, cmdline, ARRAY_SIZE(buf)); > - if (len == -E2BIG) > - len = ARRAY_SIZE(buf) - 1; > + len = min(len, ARRAY_SIZE(buf) - 1); > + strscpy(buf, cmdline, len); > > if (strcmp(buf, "--") == 0) > return; > > > -- > Tested-by: Mostafa Saleh <smostafa@google.com> > Without deviation from the norm, progress is not possible. > > Thanks, Mostafa
On Fri, 01 Sep 2023 12:57:10 +0100, Mostafa Saleh <smostafa@google.com> wrote: > > Hi Marc, > > On Fri, Sep 01, 2023 at 12:24:45PM +0100, Marc Zyngier wrote: > > Hi Mostafa, > > > > On Thu, 31 Aug 2023 17:22:27 +0100, > > Mostafa Saleh <smostafa@google.com> wrote: > > > > > > This reverts commit d232606773a0b09ec7f1ffc25f63abe801d011fd. > > > > > > Using strscpy is not correct in this context and the commit > > > assumption is not right "strncpy is deprecated for use on > > > NUL-terminated destination strings". > > > > > > strncpy is used here to copy parts of the string(cmdline) separated > > > by spaces into the buffer and not a NULL terminated string. > > > > > > This breaks the arm options "kvm-arm.mode=protected, arm64.nobti ..." > > > > > > Signed-off-by: Mostafa Saleh <smostafa@google.com> > > > --- > > > arch/arm64/kernel/idreg-override.c | 6 +++--- > > > 1 file changed, 3 insertions(+), 3 deletions(-) > > > > > > diff --git a/arch/arm64/kernel/idreg-override.c b/arch/arm64/kernel/idreg-override.c > > > index aee12c75b738..2fe2491b692c 100644 > > > --- a/arch/arm64/kernel/idreg-override.c > > > +++ b/arch/arm64/kernel/idreg-override.c > > > @@ -262,9 +262,9 @@ static __init void __parse_cmdline(const char *cmdline, bool parse_aliases) > > > if (!len) > > > return; > > > > > > - len = strscpy(buf, cmdline, ARRAY_SIZE(buf)); > > > - if (len == -E2BIG) > > > - len = ARRAY_SIZE(buf) - 1; > > > + len = min(len, ARRAY_SIZE(buf) - 1); > > > + strncpy(buf, cmdline, len); > > > + buf[len] = 0; > > > > > > if (strcmp(buf, "--") == 0) > > > return; > > > > Instead of completely reverting the patch, maybe something like the > > hack below (completely untested), so that we can still get rid of > > another instance of strncpy(), and yet bring back some sanity in the > > logic? > I was thinking of something similar, but I see we limit the len anyway > by the buffer size - 1 and force the NUL at the end so it should be > safe, unless the goal is to get rid of strncpy all the way, in this > case we can do it as you proposed. > > There is also a V3 of the original patch which uses memcpy instead. > https://lore.kernel.org/all/20230831-strncpy-arch-arm64-v3-1-cdbb1e7ea5e1@google.com/ Ah, that one would actually get my vote, because it does exactly what I actually intended this piece of code to do. Justin, do you mind rebasing your v3 on top of to restore the currently broken functionality? > > > Thanks, > > > > M. > > > > diff --git a/arch/arm64/kernel/idreg-override.c b/arch/arm64/kernel/idreg-override.c > > index aee12c75b738..be5c778a3f14 100644 > > --- a/arch/arm64/kernel/idreg-override.c > > +++ b/arch/arm64/kernel/idreg-override.c > > @@ -262,9 +262,8 @@ static __init void __parse_cmdline(const char *cmdline, bool parse_aliases) > > if (!len) > > return; > > > > - len = strscpy(buf, cmdline, ARRAY_SIZE(buf)); > > - if (len == -E2BIG) > > - len = ARRAY_SIZE(buf) - 1; > > + len = min(len, ARRAY_SIZE(buf) - 1); > > + strscpy(buf, cmdline, len); > > > > if (strcmp(buf, "--") == 0) > > return; > > > > > > -- > > > > Tested-by: Mostafa Saleh <smostafa@google.com> Cheers Mostafa. You could also just pick it, and repost it as a v2 (I really don't mind). It should give Will and Catalin a choice of implementations (and an opportunity for some additional bike-shedding ;-)). Thanks, M.
diff --git a/arch/arm64/kernel/idreg-override.c b/arch/arm64/kernel/idreg-override.c index aee12c75b738..2fe2491b692c 100644 --- a/arch/arm64/kernel/idreg-override.c +++ b/arch/arm64/kernel/idreg-override.c @@ -262,9 +262,9 @@ static __init void __parse_cmdline(const char *cmdline, bool parse_aliases) if (!len) return; - len = strscpy(buf, cmdline, ARRAY_SIZE(buf)); - if (len == -E2BIG) - len = ARRAY_SIZE(buf) - 1; + len = min(len, ARRAY_SIZE(buf) - 1); + strncpy(buf, cmdline, len); + buf[len] = 0; if (strcmp(buf, "--") == 0) return;
This reverts commit d232606773a0b09ec7f1ffc25f63abe801d011fd. Using strscpy is not correct in this context and the commit assumption is not right "strncpy is deprecated for use on NUL-terminated destination strings". strncpy is used here to copy parts of the string(cmdline) separated by spaces into the buffer and not a NULL terminated string. This breaks the arm options "kvm-arm.mode=protected, arm64.nobti ..." Signed-off-by: Mostafa Saleh <smostafa@google.com> --- arch/arm64/kernel/idreg-override.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)