| Message ID | 1498208859-6099-1-git-send-email-Dave.Martin@arm.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Fri, Jun 23, 2017 at 10:07:39AM +0100, Dave Martin wrote: > ILP32 support uses the same struct sigcontext as the native ABI > (i.e., LP64), but a different layout for the rest of the signal > frame (since siginfo_t and ucontext_t are both ABI-dependent). > > Since the purpose of parse_user_sigframe() is really to parse > sigcontext and not the whole signal frame, the function does not > need to depend on the layout of rt_sigframe -- the only purpose of > the rt_sigframe pointer is for use as a base to measure the signal > frame size. > > So, this patch renames the function to parse_user_sigcontext() and > makes the sigframe base pointer generic. ABI-specific parsers that > share the same sigcontext definition can then call it. > > To minimise churn in this patch, the native LP64 parser is retained > under the old name, but becomes a call to parse_user_sigconext(). > It may make sense instead to fold this into its restore_sigframe(), > depending on how ILP32 support is integrated. > > Suggested-by: Yury Norov <ynorov@caviumnetworks.com> > Signed-off-by: Dave Martin <Dave.Martin@arm.com> > --- > > This patch depends on [1], which does not appear to be applied yet. > > [1] [PATCH] arm64: signal: Allow expansion of the signal frame http://lists.infradead.org/pipermail/linux-arm-kernel/2017-June/514699.html > > arch/arm64/kernel/signal.c | 14 ++++++++++---- > 1 file changed, 10 insertions(+), 4 deletions(-) > > diff --git a/arch/arm64/kernel/signal.c b/arch/arm64/kernel/signal.c > index e47cd3a..a189321 100644 > --- a/arch/arm64/kernel/signal.c > +++ b/arch/arm64/kernel/signal.c > @@ -223,16 +223,16 @@ struct user_ctxs { > struct fpsimd_context __user *fpsimd; > }; > > -static int parse_user_sigframe(struct user_ctxs *user, > - struct rt_sigframe __user *sf) > +static int parse_user_sigcontext(struct user_ctxs *user, > + struct sigcontext __user const *sc, > + void __user const *sigframe_base) > { > - struct sigcontext __user *const sc = &sf->uc.uc_mcontext; > struct _aarch64_ctx __user *head; > char __user *base = (char __user *)&sc->__reserved; > size_t offset = 0; > size_t limit = sizeof(sc->__reserved); > bool have_extra_context = false; > - char const __user *const sfp = (char const __user *)sf; > + char const __user *const sfp = (char const __user *)sigframe_base; > > user->fpsimd = NULL; > > @@ -366,6 +366,12 @@ static int parse_user_sigframe(struct user_ctxs *user, > return -EINVAL; > } > > +static int parse_user_sigframe(struct user_ctxs *user, > + struct rt_sigframe __user const *sf) > +{ > + return parse_user_sigcontext(user, &sf->uc.uc_mcontext, sf); > +} > + Hi Dave, Thank you. Seems this is what I need. Only one thing. We can #define parse_user_sigframe() as macro, and so bypass type control. The macro then will be used both by lp64 and ilp32 without any modification. Yury > static int restore_sigframe(struct pt_regs *regs, > struct rt_sigframe __user *sf) > { > -- > 2.1.4
On Sun, Jun 25, 2017 at 06:05:44PM +0300, Yury Norov wrote: > On Fri, Jun 23, 2017 at 10:07:39AM +0100, Dave Martin wrote: > > ILP32 support uses the same struct sigcontext as the native ABI > > (i.e., LP64), but a different layout for the rest of the signal > > frame (since siginfo_t and ucontext_t are both ABI-dependent). > > > > Since the purpose of parse_user_sigframe() is really to parse > > sigcontext and not the whole signal frame, the function does not > > need to depend on the layout of rt_sigframe -- the only purpose of > > the rt_sigframe pointer is for use as a base to measure the signal > > frame size. > > > > So, this patch renames the function to parse_user_sigcontext() and > > makes the sigframe base pointer generic. ABI-specific parsers that > > share the same sigcontext definition can then call it. > > > > To minimise churn in this patch, the native LP64 parser is retained > > under the old name, but becomes a call to parse_user_sigconext(). > > It may make sense instead to fold this into its restore_sigframe(), > > depending on how ILP32 support is integrated. > > > > Suggested-by: Yury Norov <ynorov@caviumnetworks.com> > > Signed-off-by: Dave Martin <Dave.Martin@arm.com> > > --- > > > > This patch depends on [1], which does not appear to be applied yet. > > > > [1] [PATCH] arm64: signal: Allow expansion of the signal frame http://lists.infradead.org/pipermail/linux-arm-kernel/2017-June/514699.html [...] FYI, [1] above is now merged in arm64 for-next/core, so we shouldn't have dependency issues with this patch. > > +static int parse_user_sigframe(struct user_ctxs *user, > > + struct rt_sigframe __user const *sf) > > +{ > > + return parse_user_sigcontext(user, &sf->uc.uc_mcontext, sf); > > +} > > + > > Hi Dave, > > Thank you. Seems this is what I need. Only one thing. We can #define > parse_user_sigframe() as macro, and so bypass type control. > The macro then will be used both by lp64 and ilp32 without any > modification. For the reasons you already gave, parse_user_sigframe is probably not the right name for such a macro. But it otherwise makes sense as an interface. Does [2] look OK for you? Unless you suggest otherwise, I don't plan to maintain or push this patch -- I think it makes more sense for you to carry it in the ILP32 series, since that's the only thing that needs it. Cheers ---Dave [2] [RFC PATCH v2] arm64: signal: Make parse_user_sigframe() independent of rt_sigframe layout http://lists.infradead.org/pipermail/linux-arm-kernel/2017-June/515746.html
On Mon, Jun 26, 2017 at 10:35:21AM +0100, Dave Martin wrote: > On Sun, Jun 25, 2017 at 06:05:44PM +0300, Yury Norov wrote: > > On Fri, Jun 23, 2017 at 10:07:39AM +0100, Dave Martin wrote: > > > ILP32 support uses the same struct sigcontext as the native ABI > > > (i.e., LP64), but a different layout for the rest of the signal > > > frame (since siginfo_t and ucontext_t are both ABI-dependent). > > > > > > Since the purpose of parse_user_sigframe() is really to parse > > > sigcontext and not the whole signal frame, the function does not > > > need to depend on the layout of rt_sigframe -- the only purpose of > > > the rt_sigframe pointer is for use as a base to measure the signal > > > frame size. > > > > > > So, this patch renames the function to parse_user_sigcontext() and > > > makes the sigframe base pointer generic. ABI-specific parsers that > > > share the same sigcontext definition can then call it. > > > > > > To minimise churn in this patch, the native LP64 parser is retained > > > under the old name, but becomes a call to parse_user_sigconext(). > > > It may make sense instead to fold this into its restore_sigframe(), > > > depending on how ILP32 support is integrated. > > > > > > Suggested-by: Yury Norov <ynorov@caviumnetworks.com> > > > Signed-off-by: Dave Martin <Dave.Martin@arm.com> > > > --- > > > > > > This patch depends on [1], which does not appear to be applied yet. > > > > > > [1] [PATCH] arm64: signal: Allow expansion of the signal frame http://lists.infradead.org/pipermail/linux-arm-kernel/2017-June/514699.html > > [...] > > FYI, [1] above is now merged in arm64 for-next/core, so we shouldn't > have dependency issues with this patch. > > > > +static int parse_user_sigframe(struct user_ctxs *user, > > > + struct rt_sigframe __user const *sf) > > > +{ > > > + return parse_user_sigcontext(user, &sf->uc.uc_mcontext, sf); > > > +} > > > + > > > > Hi Dave, > > > > Thank you. Seems this is what I need. Only one thing. We can #define > > parse_user_sigframe() as macro, and so bypass type control. > > The macro then will be used both by lp64 and ilp32 without any > > modification. > > For the reasons you already gave, parse_user_sigframe is probably not > the right name for such a macro. But it otherwise makes sense as an > interface. Does [2] look OK for you? Yes, looks good, thank you. > Unless you suggest otherwise, I don't plan to maintain or push this > patch -- I think it makes more sense for you to carry it in the ILP32 > series, since that's the only thing that needs it. I'll just add it to ilp32 series. Yury
On Mon, Jun 26, 2017 at 06:38:27PM +0300, Yury Norov wrote: > On Mon, Jun 26, 2017 at 10:35:21AM +0100, Dave Martin wrote: > > On Sun, Jun 25, 2017 at 06:05:44PM +0300, Yury Norov wrote: > > > On Fri, Jun 23, 2017 at 10:07:39AM +0100, Dave Martin wrote: > > > > ILP32 support uses the same struct sigcontext as the native ABI > > > > (i.e., LP64), but a different layout for the rest of the signal > > > > frame (since siginfo_t and ucontext_t are both ABI-dependent). > > > > > > > > Since the purpose of parse_user_sigframe() is really to parse > > > > sigcontext and not the whole signal frame, the function does not > > > > need to depend on the layout of rt_sigframe -- the only purpose of > > > > the rt_sigframe pointer is for use as a base to measure the signal > > > > frame size. > > > > > > > > So, this patch renames the function to parse_user_sigcontext() and > > > > makes the sigframe base pointer generic. ABI-specific parsers that > > > > share the same sigcontext definition can then call it. > > > > > > > > To minimise churn in this patch, the native LP64 parser is retained > > > > under the old name, but becomes a call to parse_user_sigconext(). > > > > It may make sense instead to fold this into its restore_sigframe(), > > > > depending on how ILP32 support is integrated. > > > > > > > > Suggested-by: Yury Norov <ynorov@caviumnetworks.com> > > > > Signed-off-by: Dave Martin <Dave.Martin@arm.com> > > > > --- > > > > > > > > This patch depends on [1], which does not appear to be applied yet. > > > > > > > > [1] [PATCH] arm64: signal: Allow expansion of the signal frame http://lists.infradead.org/pipermail/linux-arm-kernel/2017-June/514699.html > > > > [...] > > > > FYI, [1] above is now merged in arm64 for-next/core, so we shouldn't > > have dependency issues with this patch. > > > > > > +static int parse_user_sigframe(struct user_ctxs *user, > > > > + struct rt_sigframe __user const *sf) > > > > +{ > > > > + return parse_user_sigcontext(user, &sf->uc.uc_mcontext, sf); > > > > +} > > > > + > > > > > > Hi Dave, > > > > > > Thank you. Seems this is what I need. Only one thing. We can #define > > > parse_user_sigframe() as macro, and so bypass type control. > > > The macro then will be used both by lp64 and ilp32 without any > > > modification. > > > > For the reasons you already gave, parse_user_sigframe is probably not > > the right name for such a macro. But it otherwise makes sense as an > > interface. Does [2] look OK for you? > > Yes, looks good, thank you. > > > Unless you suggest otherwise, I don't plan to maintain or push this > > patch -- I think it makes more sense for you to carry it in the ILP32 > > series, since that's the only thing that needs it. > > I'll just add it to ilp32 series. OK, thanks Let me know if you have any other questions. Cheers ---Dave
diff --git a/arch/arm64/kernel/signal.c b/arch/arm64/kernel/signal.c index e47cd3a..a189321 100644 --- a/arch/arm64/kernel/signal.c +++ b/arch/arm64/kernel/signal.c @@ -223,16 +223,16 @@ struct user_ctxs { struct fpsimd_context __user *fpsimd; }; -static int parse_user_sigframe(struct user_ctxs *user, - struct rt_sigframe __user *sf) +static int parse_user_sigcontext(struct user_ctxs *user, + struct sigcontext __user const *sc, + void __user const *sigframe_base) { - struct sigcontext __user *const sc = &sf->uc.uc_mcontext; struct _aarch64_ctx __user *head; char __user *base = (char __user *)&sc->__reserved; size_t offset = 0; size_t limit = sizeof(sc->__reserved); bool have_extra_context = false; - char const __user *const sfp = (char const __user *)sf; + char const __user *const sfp = (char const __user *)sigframe_base; user->fpsimd = NULL; @@ -366,6 +366,12 @@ static int parse_user_sigframe(struct user_ctxs *user, return -EINVAL; } +static int parse_user_sigframe(struct user_ctxs *user, + struct rt_sigframe __user const *sf) +{ + return parse_user_sigcontext(user, &sf->uc.uc_mcontext, sf); +} + static int restore_sigframe(struct pt_regs *regs, struct rt_sigframe __user *sf) {
ILP32 support uses the same struct sigcontext as the native ABI (i.e., LP64), but a different layout for the rest of the signal frame (since siginfo_t and ucontext_t are both ABI-dependent). Since the purpose of parse_user_sigframe() is really to parse sigcontext and not the whole signal frame, the function does not need to depend on the layout of rt_sigframe -- the only purpose of the rt_sigframe pointer is for use as a base to measure the signal frame size. So, this patch renames the function to parse_user_sigcontext() and makes the sigframe base pointer generic. ABI-specific parsers that share the same sigcontext definition can then call it. To minimise churn in this patch, the native LP64 parser is retained under the old name, but becomes a call to parse_user_sigconext(). It may make sense instead to fold this into its restore_sigframe(), depending on how ILP32 support is integrated. Suggested-by: Yury Norov <ynorov@caviumnetworks.com> Signed-off-by: Dave Martin <Dave.Martin@arm.com> --- This patch depends on [1], which does not appear to be applied yet. [1] [PATCH] arm64: signal: Allow expansion of the signal frame http://lists.infradead.org/pipermail/linux-arm-kernel/2017-June/514699.html arch/arm64/kernel/signal.c | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-)