| Message ID | 20220802091030.3742334-3-asavkov@redhat.com (mailing list archive) |
|---|---|
| State | Superseded |
| Delegated to: | BPF |
| Headers | show |
| Series | destructive bpf_kfuncs | expand |
On Tue, Aug 02, 2022 at 11:10:29AM +0200, Artem Savkov wrote: > Allow properly marked bpf programs to call crash_kexec(). > > Signed-off-by: Artem Savkov <asavkov@redhat.com> > --- > kernel/kexec_core.c | 21 +++++++++++++++++++++ > 1 file changed, 21 insertions(+) > > diff --git a/kernel/kexec_core.c b/kernel/kexec_core.c > index 4d34c78334ce..9259ea3bd693 100644 > --- a/kernel/kexec_core.c > +++ b/kernel/kexec_core.c > @@ -39,6 +39,8 @@ > #include <linux/hugetlb.h> > #include <linux/objtool.h> > #include <linux/kmsg_dump.h> > +#include <linux/btf.h> > +#include <linux/btf_ids.h> > > #include <asm/page.h> > #include <asm/sections.h> > @@ -1238,3 +1240,22 @@ void __weak arch_kexec_protect_crashkres(void) > > void __weak arch_kexec_unprotect_crashkres(void) > {} > + > +#ifdef CONFIG_DEBUG_INFO_BTF_MODULES > +BTF_SET8_START(kexec_btf_ids) > +BTF_ID_FLAGS(func, crash_kexec, KF_DESTRUCTIVE) > +BTF_SET8_END(kexec_btf_ids) > + > +static const struct btf_kfunc_id_set kexec_kfunc_set = { > + .owner = THIS_MODULE, > + .set = &kexec_btf_ids, > +}; > + > +static int __init crash_kfunc_init(void) > +{ > + register_btf_kfunc_id_set(BPF_PROG_TYPE_TRACING, &kexec_kfunc_set); > + return 0; should we do 'return register_btf_kfunc_id_set(...' in here? jirka > +} > + > +subsys_initcall(crash_kfunc_init); > +#endif > -- > 2.35.3 >
On Tue, Aug 02, 2022 at 12:46:42PM +0200, Jiri Olsa wrote: > On Tue, Aug 02, 2022 at 11:10:29AM +0200, Artem Savkov wrote: > > +static int __init crash_kfunc_init(void) > > +{ > > + register_btf_kfunc_id_set(BPF_PROG_TYPE_TRACING, &kexec_kfunc_set); > > + return 0; > > should we do 'return register_btf_kfunc_id_set(...' in here? Maybe, but as far as I can tell the return value for init calls does absolutely nothing except for showing up in a debug message. So I don't think it will be worth a respin, but if there is one anyway I'll change this.
On Tue, Aug 2, 2022 at 2:10 AM Artem Savkov <asavkov@redhat.com> wrote: > > Allow properly marked bpf programs to call crash_kexec(). > > Signed-off-by: Artem Savkov <asavkov@redhat.com> > --- > kernel/kexec_core.c | 21 +++++++++++++++++++++ > 1 file changed, 21 insertions(+) > > diff --git a/kernel/kexec_core.c b/kernel/kexec_core.c > index 4d34c78334ce..9259ea3bd693 100644 > --- a/kernel/kexec_core.c > +++ b/kernel/kexec_core.c > @@ -39,6 +39,8 @@ > #include <linux/hugetlb.h> > #include <linux/objtool.h> > #include <linux/kmsg_dump.h> > +#include <linux/btf.h> > +#include <linux/btf_ids.h> > > #include <asm/page.h> > #include <asm/sections.h> > @@ -1238,3 +1240,22 @@ void __weak arch_kexec_protect_crashkres(void) > > void __weak arch_kexec_unprotect_crashkres(void) > {} > + > +#ifdef CONFIG_DEBUG_INFO_BTF_MODULES > +BTF_SET8_START(kexec_btf_ids) > +BTF_ID_FLAGS(func, crash_kexec, KF_DESTRUCTIVE) > +BTF_SET8_END(kexec_btf_ids) > + > +static const struct btf_kfunc_id_set kexec_kfunc_set = { > + .owner = THIS_MODULE, > + .set = &kexec_btf_ids, > +}; > + > +static int __init crash_kfunc_init(void) > +{ > + register_btf_kfunc_id_set(BPF_PROG_TYPE_TRACING, &kexec_kfunc_set); > + return 0; > +} > + > +subsys_initcall(crash_kfunc_init); > +#endif It feels there will be a bunch of such boiler plate code in different .c files in many places in the kernel if we go with this approach. Maybe we should do one call: register_btf_kfunc_id_set(BPF_PROG_TYPE_TRACING from kernel/bpf/helper.c to register all tracing kfuncs ? And gate BTF_ID_FLAGS(func, crash_kexec, KF_DESTRUCTIVE) with #ifdef CONFIG_KEXEC_CORE. We have such a pattern in verifier.c already.
On Thu, Aug 04, 2022 at 01:41:53PM -0700, Alexei Starovoitov wrote: > On Tue, Aug 2, 2022 at 2:10 AM Artem Savkov <asavkov@redhat.com> wrote: > > > > Allow properly marked bpf programs to call crash_kexec(). > > > > Signed-off-by: Artem Savkov <asavkov@redhat.com> > > --- > > kernel/kexec_core.c | 21 +++++++++++++++++++++ > > 1 file changed, 21 insertions(+) > > > > diff --git a/kernel/kexec_core.c b/kernel/kexec_core.c > > index 4d34c78334ce..9259ea3bd693 100644 > > --- a/kernel/kexec_core.c > > +++ b/kernel/kexec_core.c > > @@ -39,6 +39,8 @@ > > #include <linux/hugetlb.h> > > #include <linux/objtool.h> > > #include <linux/kmsg_dump.h> > > +#include <linux/btf.h> > > +#include <linux/btf_ids.h> > > > > #include <asm/page.h> > > #include <asm/sections.h> > > @@ -1238,3 +1240,22 @@ void __weak arch_kexec_protect_crashkres(void) > > > > void __weak arch_kexec_unprotect_crashkres(void) > > {} > > + > > +#ifdef CONFIG_DEBUG_INFO_BTF_MODULES > > +BTF_SET8_START(kexec_btf_ids) > > +BTF_ID_FLAGS(func, crash_kexec, KF_DESTRUCTIVE) > > +BTF_SET8_END(kexec_btf_ids) > > + > > +static const struct btf_kfunc_id_set kexec_kfunc_set = { > > + .owner = THIS_MODULE, > > + .set = &kexec_btf_ids, > > +}; > > + > > +static int __init crash_kfunc_init(void) > > +{ > > + register_btf_kfunc_id_set(BPF_PROG_TYPE_TRACING, &kexec_kfunc_set); > > + return 0; > > +} > > + > > +subsys_initcall(crash_kfunc_init); > > +#endif > > It feels there will be a bunch of such boiler plate code > in different .c files in many places in the kernel > if we go with this approach. > > Maybe we should do one call: > register_btf_kfunc_id_set(BPF_PROG_TYPE_TRACING > from kernel/bpf/helper.c > to register all tracing kfuncs ? > > And gate > BTF_ID_FLAGS(func, crash_kexec, KF_DESTRUCTIVE) > with #ifdef CONFIG_KEXEC_CORE. > > We have such a pattern in verifier.c already. Good idea, thanks for the pointers. I'll do that in next version.
diff --git a/kernel/kexec_core.c b/kernel/kexec_core.c index 4d34c78334ce..9259ea3bd693 100644 --- a/kernel/kexec_core.c +++ b/kernel/kexec_core.c @@ -39,6 +39,8 @@ #include <linux/hugetlb.h> #include <linux/objtool.h> #include <linux/kmsg_dump.h> +#include <linux/btf.h> +#include <linux/btf_ids.h> #include <asm/page.h> #include <asm/sections.h> @@ -1238,3 +1240,22 @@ void __weak arch_kexec_protect_crashkres(void) void __weak arch_kexec_unprotect_crashkres(void) {} + +#ifdef CONFIG_DEBUG_INFO_BTF_MODULES +BTF_SET8_START(kexec_btf_ids) +BTF_ID_FLAGS(func, crash_kexec, KF_DESTRUCTIVE) +BTF_SET8_END(kexec_btf_ids) + +static const struct btf_kfunc_id_set kexec_kfunc_set = { + .owner = THIS_MODULE, + .set = &kexec_btf_ids, +}; + +static int __init crash_kfunc_init(void) +{ + register_btf_kfunc_id_set(BPF_PROG_TYPE_TRACING, &kexec_kfunc_set); + return 0; +} + +subsys_initcall(crash_kfunc_init); +#endif
Allow properly marked bpf programs to call crash_kexec(). Signed-off-by: Artem Savkov <asavkov@redhat.com> --- kernel/kexec_core.c | 21 +++++++++++++++++++++ 1 file changed, 21 insertions(+)