bpf: make unprivileged BPF a compile time choice
Subsystems (3)
PATCH 1/1 bpf: make unprivileged BPF a compile time choice
From: Matteo Croce <[email protected]>
Add a compile time option to permanently disable unprivileged BPF and the corresponding sysctl handler so that there's absolutely no concern about unprivileged BPF being enabled from userspace during runtime. Special purpose kernels can benefit from the build-time assurance that unprivileged eBPF is disabled in all of their kernel builds rather than having to rely on userspace to permanently disable it at boot time.
The default behaviour is left unchanged, which is: unprivileged BPF compiled in but disabled at boot.
--- kernel/bpf/Kconfig | 10 +++++++++- kernel/bpf/syscall.c | 4 +++- kernel/sysctl.c | 4 ++++ 3 files changed, 16 insertions(+), 2 deletions(-)
kernel/bpf/Kconfig +9 -1
@@ -67,10 +67,18 @@ config BPF_JIT_DEFAULT_ON def_bool ARCH_WANT_DEFAULT_BPF_JIT || BPF_JIT_ALWAYS_ON depends on HAVE_EBPF_JIT && BPF_JIT +config BPF_UNPRIV + bool "Unprivileged BPF" + default y + depends on BPF_SYSCALL + help + Enables unprivileged BPF and the corresponding + /proc/sys/kernel/unprivileged_bpf_disabled knob. + config BPF_UNPRIV_DEFAULT_OFF bool "Disable unprivileged BPF by default" default y - depends on BPF_SYSCALL + depends on BPF_UNPRIV help Disables unprivileged BPF by default by setting the corresponding /proc/sys/kernel/unprivileged_bpf_disabled knob to 2. An admin can
kernel/bpf/syscall.c +3 -1
@@ -53,7 +53,9 @@ static DEFINE_IDR(link_idr);static DEFINE_SPINLOCK (link_idr_lock);int sysctl_unprivileged_bpf_disabled__read_mostly = -IS_BUILTIN (CONFIG_BPF_UNPRIV_DEFAULT_OFF ) ?2 :0 ; +IS_BUILTIN (CONFIG_BPF_UNPRIV ) ? + (IS_BUILTIN (CONFIG_BPF_UNPRIV_DEFAULT_OFF ) ?2 :0 ) + :1 ;static const struct bpf_map_ops *const bpf_map_types[] = { #define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type)
kernel/sysctl.c +4 -0
@@ -184,6 +184,7 @@ void __weak unpriv_ebpf_notify(int new_state) { } +#ifdef CONFIG_BPF_UNPRIV static int bpf_unpriv_handler (struct ctl_table *table,int write,void *buffer,size_t *lenp,loff_t *ppos) {
@@ -206,6 +207,7 @@ static int bpf_unpriv_handler(struct ctl_table *table, int write,return ret; } +#endif /* CONFIG_BPF_UNPRIV */ #endif /* CONFIG_BPF_SYSCALL && CONFIG_SYSCTL */ /*
@@ -2300,6 +2302,7 @@ static struct ctl_table kern_table[] = { },#endif #ifdef CONFIG_BPF_SYSCALL +#ifdef CONFIG_BPF_UNPRIV { .procname = "unprivileged_bpf_disabled", .data = &sysctl_unprivileged_bpf_disabled,
@@ -2309,6 +2312,7 @@ static struct ctl_table kern_table[] = { .extra1 = SYSCTL_ZERO, .extra2 = SYSCTL_TWO, }, +#endif { .procname = "bpf_stats_enabled", .data = &bpf_stats_enabled_key.key,
Signature · 1 line
2.35.1
Quoted 12 lines · Matteo Croce
From: Matteo Croce <[email protected]>
Add a compile time option to permanently disable unprivileged BPF and the corresponding sysctl handler so that there's absolutely no concern about unprivileged BPF being enabled from userspace during runtime. Special purpose kernels can benefit from the build-time assurance that unprivileged eBPF is disabled in all of their kernel builds rather than having to rely on userspace to permanently disable it at boot time.
The default behaviour is left unchanged, which is: unprivileged BPF compiled in but disabled at boot.
That is an insane level of "security" paranoia. If you're so concerned about bpf do CONFIG_BPF_SYSCALL=n