On Wed, Jan 31, 2024 at 11:17:23AM +0100, Jiri Olsa wrote:
> On Sun, Jan 28, 2024 at 06:30:19PM -0700, Daniel Xu wrote:
> > This commit teaches pahole to parse symbols in .BTF_ids section in
> > vmlinux and discover exported kfuncs. Pahole then takes the list of
> > kfuncs and injects a BTF_KIND_DECL_TAG for each kfunc.
> >
> > Example of encoding:
> >
> > $ bpftool btf dump file .tmp_vmlinux.btf | rg "DECL_TAG 'bpf_kfunc'" | wc -l
> > 121
> >
> > $ bpftool btf dump file .tmp_vmlinux.btf | rg 56337
> > [56337] FUNC 'bpf_ct_change_timeout' type_id=56336 linkage=static
> > [127861] DECL_TAG 'bpf_kfunc' type_id=56337 component_idx=-1
> >
> > This enables downstream users and tools to dynamically discover which
> > kfuncs are available on a system by parsing vmlinux or module BTF, both
> > available in /sys/kernel/btf.
> >
> > Signed-off-by: Daniel Xu <dxu@xxxxxxxxx>
> >
> > ---
> > Changes from v2:
> > * More reliably detect kfunc membership in set8 by tracking set addr ranges
> > * Rename some variables/functions to be more clear about kfunc vs func
> >
> > Changes from v1:
> > * Fix resource leaks
> > * Fix callee -> caller typo
> > * Rename btf_decl_tag from kfunc -> bpf_kfunc
> > * Only grab btf_id_set funcs tagged kfunc
> > * Presort btf func list
> >
> > btf_encoder.c | 347 ++++++++++++++++++++++++++++++++++++++++++++++++++
> > 1 file changed, 347 insertions(+)
> >
> > diff --git a/btf_encoder.c b/btf_encoder.c
> > index fd04008..4f742b1 100644
> > --- a/btf_encoder.c
> > +++ b/btf_encoder.c
> > @@ -34,6 +34,11 @@
> > #include <pthread.h>
> >
> > #define BTF_ENCODER_MAX_PROTO 512
> > +#define BTF_IDS_SECTION ".BTF_ids"
> > +#define BTF_ID_FUNC_PFX "__BTF_ID__func__"
> > +#define BTF_ID_SET8_PFX "__BTF_ID__set8__"
> > +#define BTF_SET8_KFUNCS (1 << 0)
> > +#define BTF_KFUNC_TYPE_TAG "bpf_kfunc"
> >
> > /* state used to do later encoding of saved functions */
> > struct btf_encoder_state {
> > @@ -79,6 +84,7 @@ struct btf_encoder {
> > gen_floats,
> > is_rel;
> > uint32_t array_index_id;
> > + struct gobuffer btf_funcs;
>
> why does this need to be stored in encoder?
I suppose it doesn't. It's used in multiple functions so I figured it'd
be less verbose than passing it around. Also since it's fairly generic.
I can move it to explicit arg passing if you want.
>
> > struct {
> > struct var_info vars[MAX_PERCPU_VAR_CNT];
> > int var_cnt;
> > @@ -94,6 +100,17 @@ struct btf_encoder {
> > } functions;
> > };
> >
>
> SNIP
>
> > +/* Returns if `sym` points to a kfunc set */
> > +static int is_sym_kfunc_set(GElf_Sym *sym, const char *name, Elf_Data *idlist, size_t idlist_addr)
> > +{
> > + int *ptr = idlist->d_buf;
> > + int idx, flags;
> > + bool is_set8;
> > +
> > + /* kfuncs are only found in BTF_SET8's */
> > + is_set8 = !strncmp(name, BTF_ID_SET8_PFX, sizeof(BTF_ID_SET8_PFX) - 1);
> > + if (!is_set8)
> > + return false;
> > +
> > + idx = sym->st_value - idlist_addr;
> > + if (idx >= idlist->d_size) {
> > + fprintf(stderr, "%s: symbol '%s' out of bounds\n", __func__, name);
> > + return false;
> > + }
> > +
> > + /* Check the set8 flags to see if it was marked as kfunc */
> > + idx = idx / sizeof(int);
> > + flags = ptr[idx + 1];
> > + return flags & BTF_SET8_KFUNCS;
>
> I wonder it'd be easier to read/follow if we bring struct btf_id_set8
> declaration in here and use it to get the flags field
Yeah, it probably would be. Since it looks like resolve_btfids is going
that direction, might as well do it here as well.
>
> > +}
> > +
> > +/*
> > + * Parse BTF_ID symbol and return the func name.
> > + *
> > + * Returns:
> > + * Caller-owned string containing func name if successful.
> > + * NULL if !func or on error.
> > + */
> > +
>
> SNIP
>
> > + /* Cannot resolve symbol or .BTF_ids sections. Nothing to do. */
> > + if (symbols_shndx == -1 || idlist_shndx == -1) {
> > + err = 0;
> > + goto out;
> > + }
> > +
> > + if (!gelf_getshdr(symscn, &shdr)) {
> > + elf_error("Failed to get ELF symbol table header");
> > + goto out;
> > + }
> > + nr_syms = shdr.sh_size / shdr.sh_entsize;
> > +
> > + err = btf_encoder__collect_btf_funcs(encoder);
> > + if (err) {
> > + fprintf(stderr, "%s: failed to collect BTF funcs\n", __func__);
> > + goto out;
> > + }
> > +
> > + /* First collect all kfunc set ranges.
> > + *
> > + * Note we choose not to sort these ranges and accept a linear
> > + * search when doing lookups. Reasoning is that the number of
> > + * sets is ~O(100) and not worth the additional code to optimize.
> > + */
>
> I think we could event add gobuffer interface/support to sort and search
> quickly the data (we use it in other place), but that can be done as follow
> up when it will become a problem as you pointed out
>
> > + for (i = 0; i < nr_syms; i++) {
> > + struct btf_kfunc_set_range range = {};
> > + const char *name;
> > + GElf_Sym sym;
> > +
> > + if (!gelf_getsym(symbols, i, &sym)) {
> > + elf_error("Failed to get ELF symbol(%d)", i);
> > + goto out;
> > + }
> > +
> > + if (sym.st_shndx != idlist_shndx)
> > + continue;
> > +
> > + name = elf_strptr(elf, strtabidx, sym.st_name);
> > + if (!is_sym_kfunc_set(&sym, name, idlist, idlist_addr))
> > + continue;
> > +
> > + range.start = sym.st_value;
> > + range.end = sym.st_value + sym.st_size;
> > + gobuffer__add(&btf_kfunc_ranges, &range, sizeof(range));
> > + }
> > +
> > + /* Now inject BTF with kfunc decl tag for detected kfuncs */
> > + for (i = 0; i < nr_syms; i++) {
> > + const struct btf_kfunc_set_range *ranges;
> > + unsigned int ranges_cnt;
> > + char *func, *name;
> > + GElf_Sym sym;
> > + bool found;
> > + int err;
> > + int j;
> > +
> > + if (!gelf_getsym(symbols, i, &sym)) {
> > + elf_error("Failed to get ELF symbol(%d)", i);
> > + goto out;
> > + }
> > +
> > + if (sym.st_shndx != idlist_shndx)
> > + continue;
> > +
> > + name = elf_strptr(elf, strtabidx, sym.st_name);
> > + func = get_func_name(name);
> > + if (!func)
> > + continue;
> > +
> > + /* Check if function belongs to a kfunc set */
> > + ranges = gobuffer__entries(&btf_kfunc_ranges);
> > + ranges_cnt = gobuffer__nr_entries(&btf_kfunc_ranges);
> > + found = false;
> > + for (j = 0; j < ranges_cnt; j++) {
> > + size_t addr = sym.st_value;
>
> missing newline after declaration
>
> > + if (ranges[j].start <= addr && addr < ranges[j].end) {
> > + found = true;
> > + break;
> > + }
> > + }
> > + if (!found)
>
> leaking func
Ack, nice catch.
[..]
Thanks,
Daniel
