Re: [PATCH bpf-next v11 1/2] selftests/bpf: trace_helpers.c: optimize kallsyms cache

[Jiri Olsa]

On Tue, Sep 05, 2023 at 10:04:18PM +0800, Rong Tao wrote:
> From: Rong Tao <rongtao@xxxxxxxx>
> 
> Static ksyms often have problems because the number of symbols exceeds the
> MAX_SYMS limit. Like changing the MAX_SYMS from 300000 to 400000 in
> commit e76a014334a6("selftests/bpf: Bump and validate MAX_SYMS") solves
> the problem somewhat, but it's not the perfect way.
> 
> This commit uses dynamic memory allocation, which completely solves the
> problem caused by the limitation of the number of kallsyms. At the same
> time, add APIs:
> 
>     load_kallsyms_local()
>     ksym_search_local()
>     ksym_get_addr_local()
>     free_kallsyms_local()
> 
> There are used to solve the problem of selftests/bpf updating kallsyms
> after attach new symbols during testmod testing.
> 
> Acked-by: Stanislav Fomichev <sdf@xxxxxxxxxx>
> Signed-off-by: Rong Tao <rongtao@xxxxxxxx>

looks good, I added few more coments, with them addressed you can add my

Acked-by: Jiri Olsa <jolsa@xxxxxxxxxx>

thanks,
jirka


> ---
> v11: Remove useless load_kallsyms_refresh() and modify code some format
> v10: https://lore.kernel.org/lkml/tencent_0A73B402B1D440480838ABF7124CE5EA5505@xxxxxx/
>     Keep the original load_kallsyms().
> v9: https://lore.kernel.org/lkml/tencent_254B7015EED7A5D112C45E033DA1822CF107@xxxxxx/
>     Add load_kallsyms_local,ksym_search_local,ksym_get_addr_local functions.
> v8: https://lore.kernel.org/lkml/tencent_6D23FE187408D965E95DFAA858BC7E8C760A@xxxxxx/
>     Resolves inter-thread contention for ksyms global variables.
> v7: https://lore.kernel.org/lkml/tencent_BD6E19C00BF565CD5C36A9A0BD828CFA210A@xxxxxx/
>     Fix __must_check macro.
> v6: https://lore.kernel.org/lkml/tencent_4A09A36F883A06EA428A593497642AF8AF08@xxxxxx/
>     Apply libbpf_ensure_mem()
> v5: https://lore.kernel.org/lkml/tencent_0E9E1A1C0981678D5E7EA9E4BDBA8EE2200A@xxxxxx/
>     Release the allocated memory once the load_kallsyms_refresh() upon error
>     given it's dynamically allocated.
> v4: https://lore.kernel.org/lkml/tencent_59C74613113F0C728524B2A82FE5540A5E09@xxxxxx/
>     Make sure most cases we don't need the realloc() path to begin with,
>     and check strdup() return value.
> v3: https://lore.kernel.org/lkml/tencent_50B4B2622FE7546A5FF9464310650C008509@xxxxxx/
>     Do not use structs and judge ksyms__add_symbol function return value.
> v2: https://lore.kernel.org/lkml/tencent_B655EE5E5D463110D70CD2846AB3262EED09@xxxxxx/
>     Do the usual len/capacity scheme here to amortize the cost of realloc, and
>     don't free symbols.
> v1: https://lore.kernel.org/lkml/tencent_AB461510B10CD484E0B2F62E3754165F2909@xxxxxx/

SNIP

> +static int ksyms__add_symbol(struct ksyms *ksyms, const char *name,
> +			     unsigned long addr)
> +{
> +	void *tmp;
> +
> +	tmp = strdup(name);
> +	if (!tmp)
> +		return -ENOMEM;
> +	ksyms->syms[ksyms->sym_cnt].addr = addr;
> +	ksyms->syms[ksyms->sym_cnt].name = tmp;
> +
> +	ksyms->sym_cnt++;
> +
> +	return 0;

nit, extra new lines above

> +}
> +
> +void free_kallsyms_local(struct ksyms *ksyms)
> +{
> +	unsigned int i;
> +
> +	if (!ksyms)
> +		return;
> +
> +	if (!ksyms->syms) {
> +		free(ksyms);
> +		return;
> +	}
> +
> +	for (i = 0; i < ksyms->sym_cnt; i++)
> +		free(ksyms->syms[i].name);
> +	free(ksyms->syms);
> +	free(ksyms);
> +}
>  
>  static int ksym_cmp(const void *p1, const void *p2)
>  {
>  	return ((struct ksym *)p1)->addr - ((struct ksym *)p2)->addr;
>  }
>  
> -int load_kallsyms_refresh(void)
> +struct ksyms *load_kallsyms_local(struct ksyms *ksyms)
>  {
>  	FILE *f;
>  	char func[256], buf[256];
>  	char symbol;
>  	void *addr;
> -	int i = 0;
> +	int ret;
>  
> -	sym_cnt = 0;
> +	/* flush kallsyms, free the previously allocated dynamic memory */
> +	free_kallsyms_local(ksyms);

with the removal of the refresh function (from last version) there's
no need now for ksyms argument in load_kallsyms_local

all the current users of load_kallsyms_local are passing NULL arg

>  
>  	f = fopen("/proc/kallsyms", "r");
>  	if (!f)
> -		return -ENOENT;
> +		return NULL;
> +
> +	ksyms = calloc(1, sizeof(struct ksyms));
> +	if (!ksyms)

missing fclose(f);

> +		return NULL;
>  
>  	while (fgets(buf, sizeof(buf), f)) {
>  		if (sscanf(buf, "%p %c %s", &addr, &symbol, func) != 3)
>  			break;
>  		if (!addr)
>  			continue;
> -		if (i >= MAX_SYMS)
> -			return -EFBIG;
>  
> -		syms[i].addr = (long) addr;
> -		syms[i].name = strdup(func);
> -		i++;
> +		ret = libbpf_ensure_mem((void **) &ksyms->syms, &ksyms->sym_cap,
> +					sizeof(struct ksym), ksyms->sym_cnt + 1);
> +		if (ret)
> +			goto error;
> +		ret = ksyms__add_symbol(ksyms, func, (unsigned long)addr);
> +		if (ret)
> +			goto error;
>  	}
>  	fclose(f);
> -	sym_cnt = i;
> -	qsort(syms, sym_cnt, sizeof(struct ksym), ksym_cmp);
> -	return 0;
> +	qsort(ksyms->syms, ksyms->sym_cnt, sizeof(struct ksym), ksym_cmp);
> +	return ksyms;
> +
> +error:
> +	free_kallsyms_local(ksyms);

missing fclose(f);

> +	return NULL;
>  }
>  
>  int load_kallsyms(void)
>  {
> -	/*
> -	 * This is called/used from multiplace places,
> -	 * load symbols just once.
> -	 */
> -	if (sym_cnt)
> -		return 0;
> -	return load_kallsyms_refresh();
> +	if (!ksyms)
> +		ksyms = load_kallsyms_local(NULL);
> +	return ksyms ? 0 : 1;
>  }
>  
> -struct ksym *ksym_search(long key)
> +struct ksym *ksym_search_local(struct ksyms *ksyms, long key)
>  {
> -	int start = 0, end = sym_cnt;
> +	int start = 0, end = ksyms->sym_cnt;
>  	int result;
>  
> +	if (!ksyms)
> +		return NULL;

I don't think we need to check !ksyms in here, you don't do
that check in ksym_get_addr_local and I think it's fine

> +
>  	/* kallsyms not loaded. return NULL */
> -	if (sym_cnt <= 0)
> +	if (ksyms->sym_cnt <= 0)
>  		return NULL;
>  
>  	while (start < end) {

SNIP

> diff --git a/tools/testing/selftests/bpf/trace_helpers.h b/tools/testing/selftests/bpf/trace_helpers.h
> index 876f3e711df6..d6eeec85a5e4 100644
> --- a/tools/testing/selftests/bpf/trace_helpers.h
> +++ b/tools/testing/selftests/bpf/trace_helpers.h
> @@ -11,13 +11,18 @@ struct ksym {
>  	long addr;
>  	char *name;
>  };
> +struct ksyms;
>  
>  int load_kallsyms(void);
> -int load_kallsyms_refresh(void);
> -
>  struct ksym *ksym_search(long key);
>  long ksym_get_addr(const char *name);
>  
> +struct ksyms *load_kallsyms_local(struct ksyms *ksyms);
> +struct ksym *ksym_search_local(struct ksyms *ksyms, long key);
> +long ksym_get_addr_local(struct ksyms *ksyms, const char *name);
> +

nit, extra newline

> +void free_kallsyms_local(struct ksyms *ksyms);
> +
>  /* open kallsyms and find addresses on the fly, faster than load + search. */
>  int kallsyms_find(const char *sym, unsigned long long *addr);
>  
> -- 
> 2.41.0
>