On Wed, Apr 20, 2022 at 03:30:43PM -0700, Stanislav Fomichev wrote:
> On Wed, Apr 20, 2022 at 3:04 PM Andrii Nakryiko
> <andrii.nakryiko@xxxxxxxxx> wrote:
> >
> > On Tue, Apr 19, 2022 at 3:23 PM Stanislav Fomichev <sdf@xxxxxxxxxx> wrote:
> > >
> > > Rename bpf_prog_run_array_cg_flags to bpf_prog_run_array_cg and
> > > use it everywhere. check_return_code already enforces sane
> > > return ranges for all cgroup types. (only egress and bind hooks have
> > > uncanonical return ranges, the rest is using [0, 1])
> > >
> > > No functional changes.
> > >
> > > Suggested-by: Alexei Starovoitov <ast@xxxxxxxxxx>
> > > Signed-off-by: Stanislav Fomichev <sdf@xxxxxxxxxx>
> > > ---
> > > include/linux/bpf-cgroup.h | 8 ++---
> > > kernel/bpf/cgroup.c | 70 ++++++++++++--------------------------
> > > 2 files changed, 24 insertions(+), 54 deletions(-)
> > >
> > > diff --git a/include/linux/bpf-cgroup.h b/include/linux/bpf-cgroup.h
> > > index 88a51b242adc..669d96d074ad 100644
> > > --- a/include/linux/bpf-cgroup.h
> > > +++ b/include/linux/bpf-cgroup.h
> > > @@ -225,24 +225,20 @@ static inline bool cgroup_bpf_sock_enabled(struct sock *sk,
> > >
> > > #define BPF_CGROUP_RUN_SA_PROG(sk, uaddr, atype) \
> > > ({ \
> > > - u32 __unused_flags; \
> > > int __ret = 0; \
> > > if (cgroup_bpf_enabled(atype)) \
> > > __ret = __cgroup_bpf_run_filter_sock_addr(sk, uaddr, atype, \
> > > - NULL, \
> > > - &__unused_flags); \
> > > + NULL, NULL); \
> > > __ret; \
> > > })
> > >
> > > #define BPF_CGROUP_RUN_SA_PROG_LOCK(sk, uaddr, atype, t_ctx) \
> > > ({ \
> > > - u32 __unused_flags; \
> > > int __ret = 0; \
> > > if (cgroup_bpf_enabled(atype)) { \
> > > lock_sock(sk); \
> > > __ret = __cgroup_bpf_run_filter_sock_addr(sk, uaddr, atype, \
> > > - t_ctx, \
> > > - &__unused_flags); \
> > > + t_ctx, NULL); \
> > > release_sock(sk); \
> > > } \
> > > __ret; \
> > > diff --git a/kernel/bpf/cgroup.c b/kernel/bpf/cgroup.c
> > > index 0cb6211fcb58..f61eca32c747 100644
> > > --- a/kernel/bpf/cgroup.c
> > > +++ b/kernel/bpf/cgroup.c
> > > @@ -25,50 +25,18 @@ EXPORT_SYMBOL(cgroup_bpf_enabled_key);
> > > /* __always_inline is necessary to prevent indirect call through run_prog
> > > * function pointer.
> > > */
> > > -static __always_inline int
> > > -bpf_prog_run_array_cg_flags(const struct cgroup_bpf *cgrp,
> > > - enum cgroup_bpf_attach_type atype,
> > > - const void *ctx, bpf_prog_run_fn run_prog,
> > > - int retval, u32 *ret_flags)
> > > -{
> > > - const struct bpf_prog_array_item *item;
> > > - const struct bpf_prog *prog;
> > > - const struct bpf_prog_array *array;
> > > - struct bpf_run_ctx *old_run_ctx;
> > > - struct bpf_cg_run_ctx run_ctx;
> > > - u32 func_ret;
> > > -
> > > - run_ctx.retval = retval;
> > > - migrate_disable();
> > > - rcu_read_lock();
> > > - array = rcu_dereference(cgrp->effective[atype]);
> > > - item = &array->items[0];
> > > - old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx);
> > > - while ((prog = READ_ONCE(item->prog))) {
> > > - run_ctx.prog_item = item;
> > > - func_ret = run_prog(prog, ctx);
> > > - if (!(func_ret & 1) && !IS_ERR_VALUE((long)run_ctx.retval))
> > > - run_ctx.retval = -EPERM;
> > > - *(ret_flags) |= (func_ret >> 1);
> > > - item++;
> > > - }
> > > - bpf_reset_run_ctx(old_run_ctx);
> > > - rcu_read_unlock();
> > > - migrate_enable();
> > > - return run_ctx.retval;
> > > -}
> > > -
> > > static __always_inline int
> > > bpf_prog_run_array_cg(const struct cgroup_bpf *cgrp,
> > > enum cgroup_bpf_attach_type atype,
> > > const void *ctx, bpf_prog_run_fn run_prog,
> > > - int retval)
> > > + int retval, u32 *ret_flags)
> > > {
> > > const struct bpf_prog_array_item *item;
> > > const struct bpf_prog *prog;
> > > const struct bpf_prog_array *array;
> > > struct bpf_run_ctx *old_run_ctx;
> > > struct bpf_cg_run_ctx run_ctx;
> > > + u32 func_ret;
> > >
> > > run_ctx.retval = retval;
> > > migrate_disable();
> > > @@ -78,8 +46,11 @@ bpf_prog_run_array_cg(const struct cgroup_bpf *cgrp,
> > > old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx);
> > > while ((prog = READ_ONCE(item->prog))) {
> > > run_ctx.prog_item = item;
> > > - if (!run_prog(prog, ctx) && !IS_ERR_VALUE((long)run_ctx.retval))
> > > + func_ret = run_prog(prog, ctx);
> > > + if (!(func_ret & 1) && !IS_ERR_VALUE((long)run_ctx.retval))
> >
> > to be completely true to previous behavior, shouldn't there be
> >
> > if (ret_flags)
> > func_ret &= 1;
> > if (!func_ret && !IS_ERR_VALUE(...))
> >
> > here?
> >
> > This might have been discussed previously and I missed it. If that's
> > so, please ignore.
>
> We are converting the cases where run_prog(prog, ctx) returns 0 or 1,
> so it seems like we don't have to reproduce the existing behavior
> 1-to-1?
> So I'm not sure it matters, or am I missing something?
A nit, how about testing 'if (ret_flags)' first such that
it is obvious which case will use higher bits in the return value.
The compiler may be able to optimize the ret_flags == NULL case also ?
Something like:
func_ret = run_prog(prog, ctx);
/* The cg bpf prog uses the higher bits of the return value */
if (ret_flags) {
*(ret_flags) |= (func_ret >> 1);
func_ret &= 1;
}
if (!func_ret && !IS_ERR_VALUE((long)run_ctx.retval))
run_ctx.retval = -EPERM;
