| Message ID | 20200504062555.2048028-1-yhs@fb.com |
|---|---|
| State | Changes Requested |
| Delegated to: | BPF Maintainers |
| Headers | show |
| Series | bpf: implement bpf iterator for kernel data | expand |
On Sun, May 3, 2020 at 11:28 PM Yonghong Song <yhs@fb.com> wrote: > > Macro DEFINE_BPF_ITER_FUNC is implemented so target > can define an init function to capture the BTF type > which represents the target. > > The bpf_iter_meta is a structure holding meta data, common > to all targets in the bpf program. > > Additional marker functions are called before/after > bpf_seq_read() show() and stop() callback functions > to help calculate precise seq_num and whether call bpf_prog > inside stop(). > > Two functions, bpf_iter_get_info() and bpf_iter_run_prog(), > are implemented so target can get needed information from > bpf_iter infrastructure and can run the program. > > Signed-off-by: Yonghong Song <yhs@fb.com> > --- > include/linux/bpf.h | 11 +++++ > kernel/bpf/bpf_iter.c | 94 ++++++++++++++++++++++++++++++++++++++++--- > 2 files changed, 100 insertions(+), 5 deletions(-) > > diff --git a/include/linux/bpf.h b/include/linux/bpf.h > index 26daf85cba10..70c71c3cd9e8 100644 > --- a/include/linux/bpf.h > +++ b/include/linux/bpf.h > @@ -1129,6 +1129,9 @@ int bpf_obj_pin_user(u32 ufd, const char __user *pathname); > int bpf_obj_get_user(const char __user *pathname, int flags); > > #define BPF_ITER_FUNC_PREFIX "__bpf_iter__" > +#define DEFINE_BPF_ITER_FUNC(target, args...) \ > + extern int __bpf_iter__ ## target(args); \ > + int __init __bpf_iter__ ## target(args) { return 0; } Why is extern declaration needed here? Doesn't the same macro define global function itself? I'm probably missing some C semantics thingy, sorry... > > typedef int (*bpf_iter_init_seq_priv_t)(void *private_data); > typedef void (*bpf_iter_fini_seq_priv_t)(void *private_data); > @@ -1141,11 +1144,19 @@ struct bpf_iter_reg { > u32 seq_priv_size; > }; > > +struct bpf_iter_meta { > + __bpf_md_ptr(struct seq_file *, seq); > + u64 session_id; > + u64 seq_num; > +}; > + [...] > /* bpf_seq_read, a customized and simpler version for bpf iterator. > * no_llseek is assumed for this file. > * The following are differences from seq_read(): > @@ -83,12 +119,15 @@ static ssize_t bpf_seq_read(struct file *file, char __user *buf, size_t size, > if (!p || IS_ERR(p)) > goto Stop; > > + bpf_iter_inc_seq_num(seq); so seq_num is one-based, not zero-based? So on first show() call it will be set to 1, not 0, right? > err = seq->op->show(seq, p); > if (seq_has_overflowed(seq)) { > + bpf_iter_dec_seq_num(seq); > err = -E2BIG; > goto Error_show; > } else if (err) { > /* < 0: go out, > 0: skip */ > + bpf_iter_dec_seq_num(seq); > if (likely(err < 0)) > goto Error_show; > seq->count = 0; [...]
On 5/5/20 1:25 PM, Andrii Nakryiko wrote: > On Sun, May 3, 2020 at 11:28 PM Yonghong Song <yhs@fb.com> wrote: >> >> Macro DEFINE_BPF_ITER_FUNC is implemented so target >> can define an init function to capture the BTF type >> which represents the target. >> >> The bpf_iter_meta is a structure holding meta data, common >> to all targets in the bpf program. >> >> Additional marker functions are called before/after >> bpf_seq_read() show() and stop() callback functions >> to help calculate precise seq_num and whether call bpf_prog >> inside stop(). >> >> Two functions, bpf_iter_get_info() and bpf_iter_run_prog(), >> are implemented so target can get needed information from >> bpf_iter infrastructure and can run the program. >> >> Signed-off-by: Yonghong Song <yhs@fb.com> >> --- >> include/linux/bpf.h | 11 +++++ >> kernel/bpf/bpf_iter.c | 94 ++++++++++++++++++++++++++++++++++++++++--- >> 2 files changed, 100 insertions(+), 5 deletions(-) >> >> diff --git a/include/linux/bpf.h b/include/linux/bpf.h >> index 26daf85cba10..70c71c3cd9e8 100644 >> --- a/include/linux/bpf.h >> +++ b/include/linux/bpf.h >> @@ -1129,6 +1129,9 @@ int bpf_obj_pin_user(u32 ufd, const char __user *pathname); >> int bpf_obj_get_user(const char __user *pathname, int flags); >> >> #define BPF_ITER_FUNC_PREFIX "__bpf_iter__" >> +#define DEFINE_BPF_ITER_FUNC(target, args...) \ >> + extern int __bpf_iter__ ## target(args); \ >> + int __init __bpf_iter__ ## target(args) { return 0; } > > Why is extern declaration needed here? Doesn't the same macro define Silence sparse warning. Apparently in kernel, any global function, they want a declaration? > global function itself? I'm probably missing some C semantics thingy, > sorry... > >> >> typedef int (*bpf_iter_init_seq_priv_t)(void *private_data); >> typedef void (*bpf_iter_fini_seq_priv_t)(void *private_data); >> @@ -1141,11 +1144,19 @@ struct bpf_iter_reg { >> u32 seq_priv_size; >> }; >> >> +struct bpf_iter_meta { >> + __bpf_md_ptr(struct seq_file *, seq); >> + u64 session_id; >> + u64 seq_num; >> +}; >> + > > [...] > >> /* bpf_seq_read, a customized and simpler version for bpf iterator. >> * no_llseek is assumed for this file. >> * The following are differences from seq_read(): >> @@ -83,12 +119,15 @@ static ssize_t bpf_seq_read(struct file *file, char __user *buf, size_t size, >> if (!p || IS_ERR(p)) >> goto Stop; >> >> + bpf_iter_inc_seq_num(seq); > > so seq_num is one-based, not zero-based? So on first show() call it > will be set to 1, not 0, right? It is 1 based, we need to document this clearly. I forgot to adjust my bpf program for this. Will adjust them properly in the next revision. > >> err = seq->op->show(seq, p); >> if (seq_has_overflowed(seq)) { >> + bpf_iter_dec_seq_num(seq); >> err = -E2BIG; >> goto Error_show; >> } else if (err) { >> /* < 0: go out, > 0: skip */ >> + bpf_iter_dec_seq_num(seq); >> if (likely(err < 0)) >> goto Error_show; >> seq->count = 0; > > [...] >
On Tue, May 5, 2020 at 1:30 PM Yonghong Song <yhs@fb.com> wrote: > > > > On 5/5/20 1:25 PM, Andrii Nakryiko wrote: > > On Sun, May 3, 2020 at 11:28 PM Yonghong Song <yhs@fb.com> wrote: > >> > >> Macro DEFINE_BPF_ITER_FUNC is implemented so target > >> can define an init function to capture the BTF type > >> which represents the target. > >> > >> The bpf_iter_meta is a structure holding meta data, common > >> to all targets in the bpf program. > >> > >> Additional marker functions are called before/after > >> bpf_seq_read() show() and stop() callback functions > >> to help calculate precise seq_num and whether call bpf_prog > >> inside stop(). > >> > >> Two functions, bpf_iter_get_info() and bpf_iter_run_prog(), > >> are implemented so target can get needed information from > >> bpf_iter infrastructure and can run the program. > >> > >> Signed-off-by: Yonghong Song <yhs@fb.com> > >> --- > >> include/linux/bpf.h | 11 +++++ > >> kernel/bpf/bpf_iter.c | 94 ++++++++++++++++++++++++++++++++++++++++--- > >> 2 files changed, 100 insertions(+), 5 deletions(-) > >> > >> diff --git a/include/linux/bpf.h b/include/linux/bpf.h > >> index 26daf85cba10..70c71c3cd9e8 100644 > >> --- a/include/linux/bpf.h > >> +++ b/include/linux/bpf.h > >> @@ -1129,6 +1129,9 @@ int bpf_obj_pin_user(u32 ufd, const char __user *pathname); > >> int bpf_obj_get_user(const char __user *pathname, int flags); > >> > >> #define BPF_ITER_FUNC_PREFIX "__bpf_iter__" > >> +#define DEFINE_BPF_ITER_FUNC(target, args...) \ > >> + extern int __bpf_iter__ ## target(args); \ > >> + int __init __bpf_iter__ ## target(args) { return 0; } > > > > Why is extern declaration needed here? Doesn't the same macro define > > Silence sparse warning. Apparently in kernel, any global function, they > want a declaration? Ah.. alright :) > > > global function itself? I'm probably missing some C semantics thingy, > > sorry... > > > >> > >> typedef int (*bpf_iter_init_seq_priv_t)(void *private_data); > >> typedef void (*bpf_iter_fini_seq_priv_t)(void *private_data); > >> @@ -1141,11 +1144,19 @@ struct bpf_iter_reg { > >> u32 seq_priv_size; > >> }; > >> > >> +struct bpf_iter_meta { > >> + __bpf_md_ptr(struct seq_file *, seq); > >> + u64 session_id; > >> + u64 seq_num; > >> +}; > >> + > > > > [...] > > > >> /* bpf_seq_read, a customized and simpler version for bpf iterator. > >> * no_llseek is assumed for this file. > >> * The following are differences from seq_read(): > >> @@ -83,12 +119,15 @@ static ssize_t bpf_seq_read(struct file *file, char __user *buf, size_t size, > >> if (!p || IS_ERR(p)) > >> goto Stop; > >> > >> + bpf_iter_inc_seq_num(seq); > > > > so seq_num is one-based, not zero-based? So on first show() call it > > will be set to 1, not 0, right? > > It is 1 based, we need to document this clearly. I forgot to adjust my > bpf program for this. Will adjust them properly in the next revision. I see. IMO, seq_num starting at 0 is more natural, but whichever way is fine with me. > > > >> err = seq->op->show(seq, p); > >> if (seq_has_overflowed(seq)) { > >> + bpf_iter_dec_seq_num(seq); > >> err = -E2BIG; > >> goto Error_show; > >> } else if (err) { > >> /* < 0: go out, > 0: skip */ > >> + bpf_iter_dec_seq_num(seq); > >> if (likely(err < 0)) > >> goto Error_show; > >> seq->count = 0; > > > > [...] > >
diff --git a/include/linux/bpf.h b/include/linux/bpf.h index 26daf85cba10..70c71c3cd9e8 100644 --- a/include/linux/bpf.h +++ b/include/linux/bpf.h @@ -1129,6 +1129,9 @@ int bpf_obj_pin_user(u32 ufd, const char __user *pathname); int bpf_obj_get_user(const char __user *pathname, int flags); #define BPF_ITER_FUNC_PREFIX "__bpf_iter__" +#define DEFINE_BPF_ITER_FUNC(target, args...) \ + extern int __bpf_iter__ ## target(args); \ + int __init __bpf_iter__ ## target(args) { return 0; } typedef int (*bpf_iter_init_seq_priv_t)(void *private_data); typedef void (*bpf_iter_fini_seq_priv_t)(void *private_data); @@ -1141,11 +1144,19 @@ struct bpf_iter_reg { u32 seq_priv_size; }; +struct bpf_iter_meta { + __bpf_md_ptr(struct seq_file *, seq); + u64 session_id; + u64 seq_num; +}; + int bpf_iter_reg_target(struct bpf_iter_reg *reg_info); bool bpf_iter_prog_supported(struct bpf_prog *prog); int bpf_iter_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); int bpf_iter_new_fd(struct bpf_link *link); bool bpf_link_is_iter(struct bpf_link *link); +struct bpf_prog *bpf_iter_get_info(struct bpf_iter_meta *meta, bool in_stop); +int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx); int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value); int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value); diff --git a/kernel/bpf/bpf_iter.c b/kernel/bpf/bpf_iter.c index 8bd787f3db6f..90d58c589816 100644 --- a/kernel/bpf/bpf_iter.c +++ b/kernel/bpf/bpf_iter.c @@ -43,6 +43,42 @@ static atomic64_t session_id; static int prepare_seq_file(struct file *file, struct bpf_iter_link *link); +static void bpf_iter_inc_seq_num(struct seq_file *seq) +{ + struct bpf_iter_priv_data *iter_priv; + + iter_priv = container_of(seq->private, struct bpf_iter_priv_data, + target_private); + iter_priv->seq_num++; +} + +static void bpf_iter_dec_seq_num(struct seq_file *seq) +{ + struct bpf_iter_priv_data *iter_priv; + + iter_priv = container_of(seq->private, struct bpf_iter_priv_data, + target_private); + iter_priv->seq_num--; +} + +static void bpf_iter_set_stop(struct seq_file *seq) +{ + struct bpf_iter_priv_data *iter_priv; + + iter_priv = container_of(seq->private, struct bpf_iter_priv_data, + target_private); + iter_priv->do_stop++; +} + +static void bpf_iter_unset_stop(struct seq_file *seq) +{ + struct bpf_iter_priv_data *iter_priv; + + iter_priv = container_of(seq->private, struct bpf_iter_priv_data, + target_private); + iter_priv->do_stop--; +} + /* bpf_seq_read, a customized and simpler version for bpf iterator. * no_llseek is assumed for this file. * The following are differences from seq_read(): @@ -83,12 +119,15 @@ static ssize_t bpf_seq_read(struct file *file, char __user *buf, size_t size, if (!p || IS_ERR(p)) goto Stop; + bpf_iter_inc_seq_num(seq); err = seq->op->show(seq, p); if (seq_has_overflowed(seq)) { + bpf_iter_dec_seq_num(seq); err = -E2BIG; goto Error_show; } else if (err) { /* < 0: go out, > 0: skip */ + bpf_iter_dec_seq_num(seq); if (likely(err < 0)) goto Error_show; seq->count = 0; @@ -113,8 +152,10 @@ static ssize_t bpf_seq_read(struct file *file, char __user *buf, size_t size, if (seq->count >= size) break; + bpf_iter_inc_seq_num(seq); err = seq->op->show(seq, p); if (seq_has_overflowed(seq)) { + bpf_iter_dec_seq_num(seq); if (offs == 0) { err = -E2BIG; goto Error_show; @@ -122,6 +163,7 @@ static ssize_t bpf_seq_read(struct file *file, char __user *buf, size_t size, seq->count = offs; break; } else if (err) { + bpf_iter_dec_seq_num(seq); /* < 0: go out, > 0: skip */ seq->count = offs; if (likely(err < 0)) { @@ -134,11 +176,17 @@ static ssize_t bpf_seq_read(struct file *file, char __user *buf, size_t size, Stop: offs = seq->count; /* may call bpf program */ - seq->op->stop(seq, p); - if (seq_has_overflowed(seq)) { - if (offs == 0) - goto Error_stop; - seq->count = offs; + if (!p) { + bpf_iter_set_stop(seq); + seq->op->stop(seq, p); + if (seq_has_overflowed(seq)) { + bpf_iter_unset_stop(seq); + if (offs == 0) + goto Error_stop; + seq->count = offs; + } + } else { + seq->op->stop(seq, p); } n = min(seq->count, size); @@ -432,3 +480,39 @@ int bpf_iter_new_fd(struct bpf_link *link) put_unused_fd(fd); return err; } + +struct bpf_prog *bpf_iter_get_info(struct bpf_iter_meta *meta, bool in_stop) +{ + struct bpf_iter_priv_data *iter_priv; + struct seq_file *seq; + void *seq_priv; + + seq = meta->seq; + if (seq->file->f_op != &bpf_iter_fops) + return NULL; + + seq_priv = seq->private; + iter_priv = container_of(seq_priv, struct bpf_iter_priv_data, + target_private); + + if (in_stop && iter_priv->do_stop != 1) + return NULL; + + meta->session_id = iter_priv->session_id; + meta->seq_num = iter_priv->seq_num; + + return iter_priv->prog; +} + +int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx) +{ + int ret; + + rcu_read_lock(); + migrate_disable(); + ret = BPF_PROG_RUN(prog, ctx); + migrate_enable(); + rcu_read_unlock(); + + return ret == 0 ? 0 : -EAGAIN; +}
Macro DEFINE_BPF_ITER_FUNC is implemented so target can define an init function to capture the BTF type which represents the target. The bpf_iter_meta is a structure holding meta data, common to all targets in the bpf program. Additional marker functions are called before/after bpf_seq_read() show() and stop() callback functions to help calculate precise seq_num and whether call bpf_prog inside stop(). Two functions, bpf_iter_get_info() and bpf_iter_run_prog(), are implemented so target can get needed information from bpf_iter infrastructure and can run the program. Signed-off-by: Yonghong Song <yhs@fb.com> --- include/linux/bpf.h | 11 +++++ kernel/bpf/bpf_iter.c | 94 ++++++++++++++++++++++++++++++++++++++++--- 2 files changed, 100 insertions(+), 5 deletions(-)