linux-trace-devel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Yordan Karadzhov <y.karadz@gmail.com>
To: "Tzvetomir Stoyanov (VMware)" <tz.stoyanov@gmail.com>,
	rostedt@goodmis.org
Cc: linux-trace-devel@vger.kernel.org
Subject: Re: [PATCH v2 03/12] libtracefs: New kprobes APIs
Date: Mon, 1 Nov 2021 19:22:26 +0200	[thread overview]
Message-ID: <78703fa1-f811-a50c-6b5d-db68f6dbf21c@gmail.com> (raw)
In-Reply-To: <20211101090904.81454-4-tz.stoyanov@gmail.com>



On 1.11.21 г. 11:08, Tzvetomir Stoyanov (VMware) wrote:
> In order to be consistent with the other APIs, a new set of kprobe APIs
> is introduced:
>   tracefs_kprobe_alloc();
>   tracefs_kretprobe_alloc();
>   tracefs_kprobe_create();
>   tracefs_kprobe_destroy();
>   tracefs_kprobe_free();
> These APIs work with kprobe context, represented by the tracefs_dynevent
> structure.
> 
> Signed-off-by: Tzvetomir Stoyanov (VMware) <tz.stoyanov@gmail.com>
> ---
>   include/tracefs.h     |   8 ++
>   src/tracefs-kprobes.c | 168 ++++++++++++++++++++++++++++++++++++++++++
>   2 files changed, 176 insertions(+)
> 
> diff --git a/include/tracefs.h b/include/tracefs.h
> index 4e721eb..85185ce 100644
> --- a/include/tracefs.h
> +++ b/include/tracefs.h
> @@ -247,6 +247,14 @@ enum tracefs_kprobe_type {
>   	TRACEFS_KRETPROBE,
>   };
>   
> +struct tracefs_dynevent *
> +tracefs_kprobe_alloc(const char *system, const char *event, const char *addr, const char *format);
> +struct tracefs_dynevent *
> +tracefs_kretprobe_alloc(const char *system, const char *event,
> +			const char *addr, const char *format, int max);
> +int tracefs_kprobe_create(struct tracefs_dynevent *kprobe);
> +int tracefs_kprobe_destroy(struct tracefs_dynevent *kprobe, bool force);
> +void tracefs_kprobe_free(struct tracefs_dynevent *kprobe);
>   int tracefs_kprobe_raw(const char *system, const char *event,
>   		       const char *addr, const char *format);
>   int tracefs_kretprobe_raw(const char *system, const char *event,
> diff --git a/src/tracefs-kprobes.c b/src/tracefs-kprobes.c
> index d4c5f9e..906f914 100644
> --- a/src/tracefs-kprobes.c
> +++ b/src/tracefs-kprobes.c
> @@ -20,6 +20,134 @@
>   #define KPROBE_EVENTS "kprobe_events"
>   #define KPROBE_DEFAULT_GROUP "kprobes"
>   
> +static struct tracefs_dynevent *
> +kprobe_alloc(enum trace_dynevent_type type, const char *system, const char *event,
> +	     const char *addr, const char *format)
> +{
> +	struct tracefs_dynevent *kp;
> +	const char *sys = system;
> +	const char *ename = event;
> +	char *tmp;
> +
> +	if (!addr) {
> +		errno = EBADMSG;
> +		return NULL;
> +	}
> +	if (!sys)
> +		sys = KPROBE_DEFAULT_GROUP;
> +
> +	if (!event) {
> +		ename = strdup(addr);
> +		if (!ename)
> +			return NULL;
> +		tmp = strchr(ename, ':');
> +		if (tmp)
> +			*tmp = '\0';
> +	}
> +
> +	kp = dynevent_alloc(type, sys, ename, addr, format);
> +	if (!event)
> +		free((char *)ename);
> +
> +	return kp;
> +}
> +
> +
> +/**
> + * tracefs_kprobe_alloc - Allocate new kprobe context
> + * @system: The system name (NULL for the default kprobes)
> + * @event: The event to create (NULL to use @addr for the event)
> + * @addr: The function and offset (or address) to insert the probe
> + * @format: The format string to define the probe.
> + *
> + * Allocate a kprobe context that will be in the @system group (or kprobes if
> + * @system is NULL). Have the name of @event (or @addr if @event is
> + * NULL). Will be inserted to @addr (function name, with or without
> + * offset, or a address). And the @format will define the format
> + * of the kprobe. See the Linux documentation file under:
> + * Documentation/trace/kprobetrace.rst
> + * The kprobe is not created in the system.
> + *
> + * Return a pointer to a kprobe context on success, or NULL on error.
> + * The returned pointer must be freed with tracefs_kprobe_free()
> + *
> + * errno will be set to EBADMSG if addr is NULL.
> + */
> +struct tracefs_dynevent *
> +tracefs_kprobe_alloc(const char *system, const char *event, const char *addr, const char *format)
> +
> +{
> +	return kprobe_alloc(TRACE_DYNEVENT_KPROBE, system, event, addr, format);
> +}
> +
> +/**
> + * tracefs_kretprobe_alloc - Allocate new kretprobe context
> + * @system: The system name (NULL for the default kprobes)
> + * @event: The event to create (NULL to use @addr for the event)
> + * @addr: The function and offset (or address) to insert the retprobe
> + * @format: The format string to define the retprobe.
> + *

@max is not documented.

> + * Allocate a kretprobe that will be in the @system group (or kprobes if
> + * @system is NULL). Have the name of @event (or @addr if @event is
> + * NULL). Will be inserted to @addr (function name, with or without
> + * offset, or a address). And the @format will define the raw format
> + * of the kprobe. See the Linux documentation file under:
> + * Documentation/trace/kprobetrace.rst
> + * The kretprobe is not created in the system.
> + *
> + * Return a pointer to a kprobe context on success, or NULL on error.
> + * The returned pointer must be freed with tracefs_kprobe_free()
> + *
> + * errno will be set to EBADMSG if addr is NULL.
> + */
> +struct tracefs_dynevent *
> +tracefs_kretprobe_alloc(const char *system, const char *event,
> +			const char *addr, const char *format, int max)
> +{
> +	struct tracefs_dynevent *kp;
> +	int ret;
> +
> +	kp = kprobe_alloc(TRACE_DYNEVENT_KRETPROBE, system, event, addr, format);
> +	if (!kp)
> +		return NULL;
> +	if (max) {
> +		free(kp->prefix);
> +		kp->prefix = NULL;
> +		ret = asprintf(&kp->prefix, "r%d:", max);
> +		if (ret < 0)
> +			goto error;
> +	}
> +
> +	return kp;
> +error:
> +	dynevent_free(kp);
> +	return NULL;
> +}
> +
> +/**
> + * tracefs_kprobe_create - Create a kprobe or kretprobe in the system
> + * @kprobe: Pointer to a kprobe context, describing the probe
> + *
> + * Return 0 on success, or -1 on error.
> + */
> +int tracefs_kprobe_create(struct tracefs_dynevent *kprobe)
> +{
> +	return dynevent_create(kprobe);
> +}
> +

hmm, what will happen if if you have a code like this:


struct tracefs_dynevent *synth = tracefs_synth_alloc(...)

int ret = tracefs_kprobe_create(synth);


Are you just giving several additional fake names to the same function ('dynevent_create()')?


> +/**
> + * tracefs_kprobe_free - Free a kprobe context
> + * @kprobe: Pointer to a kprobe context
> + *
> + * The kprobe/kretprobe, described by this context, is not
> + * removed from the system by this API. It only frees the memory.
> + */
> +void tracefs_kprobe_free(struct tracefs_dynevent *kprobe)
> +{
> +	dynevent_free(kprobe);
> +}
> +

The same argument applys here.

Thanks!
Yordan

> +
>   static int insert_kprobe(const char *type, const char *system,
>   			 const char *event, const char *addr,
>   			 const char *format)
> @@ -474,3 +602,43 @@ int tracefs_kprobe_clear_probe(const char *system, const char *event, bool force
>   
>   	return ret < 0 ? -1 : 0;
>   }
> +
> +/**
> + * tracefs_kprobe_destroy - Remove a kprobe or kretprobe from the system
> + * @kprobe: A kprobe context, describing the kprobe that will be deleted.
> + *	    If NULL, all kprobes and kretprobes in the system will be deleted
> + * @force: Will attempt to disable all kprobe events and clear them
> + *
> + * The kprobe/kretprobe context is not freed by this API.
> + * It only removes the probe from the system.
> + *
> + * Return 0 on success, or -1 on error.
> + */
> +int tracefs_kprobe_destroy(struct tracefs_dynevent *kprobe, bool force)
> +{
> +	char **instance_list;
> +	int ret;
> +
> +	if (!kprobe) {
> +		if (tracefs_instance_file_clear(NULL, KPROBE_EVENTS) == 0)
> +			return 0;
> +		if (!force)
> +			return -1;
> +		/* Attempt to disable all kprobe events */
> +		return kprobe_clear_probes(NULL, force);
> +	}
> +
> +	/*
> +	 * Since we know we are disabling a specific event, try
> +	 * to disable it first before clearing it.
> +	 */
> +	if (force) {
> +		instance_list = tracefs_instances(NULL);
> +		disable_events(kprobe->system, kprobe->event, instance_list);
> +		tracefs_list_free(instance_list);
> +	}
> +
> +	ret = dynevent_destroy(kprobe);
> +
> +	return ret < 0 ? -1 : 0;
> +}
> 

  reply	other threads:[~2021-11-01 17:22 UTC|newest]

Thread overview: 21+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-11-01  9:08 [PATCH v2 00/12] libtracefs dynamic events support Tzvetomir Stoyanov (VMware)
2021-11-01  9:08 ` [PATCH v2 01/12] libtracefs: Add new internal APIs for dynamic events Tzvetomir Stoyanov (VMware)
2021-11-01 17:06   ` Yordan Karadzhov
2021-11-02  4:33     ` Tzvetomir Stoyanov
2021-11-02 13:25       ` Steven Rostedt
2021-11-01  9:08 ` [PATCH v2 02/12] libtracefs: Rename tracefs_get_kprobes API Tzvetomir Stoyanov (VMware)
2021-11-01  9:08 ` [PATCH v2 03/12] libtracefs: New kprobes APIs Tzvetomir Stoyanov (VMware)
2021-11-01 17:22   ` Yordan Karadzhov [this message]
2021-11-02  4:58     ` Tzvetomir Stoyanov
2021-11-02  7:43       ` Yordan Karadzhov
2021-11-02 13:49         ` Steven Rostedt
2021-11-01  9:08 ` [PATCH v2 04/12] libtracefs: Remove redundant " Tzvetomir Stoyanov (VMware)
2021-11-01  9:08 ` [PATCH v2 05/12] libtracefs: Reimplement tracefs_kprobes_get API Tzvetomir Stoyanov (VMware)
2021-11-01  9:08 ` [PATCH v2 06/12] libtracefs: Change tracefs_kprobe_info API Tzvetomir Stoyanov (VMware)
2021-11-01  9:08 ` [PATCH v2 07/12] libtracefs: Reimplement kprobe raw APIs Tzvetomir Stoyanov (VMware)
2021-11-01  9:09 ` [PATCH v2 08/12] libtracefs: Extend kprobes unit test Tzvetomir Stoyanov (VMware)
2021-11-01  9:09 ` [PATCH v2 09/12] libtracefs: Update kprobes man pages Tzvetomir Stoyanov (VMware)
2021-11-01  9:09 ` [PATCH v2 10/12] libtracefs: Rename tracefs_synth_init API Tzvetomir Stoyanov (VMware)
2021-11-01  9:09 ` [PATCH v2 11/12] libtracefs: Use the internal dynamic events API when creating synthetic events Tzvetomir Stoyanov (VMware)
2021-11-01  9:09 ` [PATCH v2 12/12] libtracefs: Document tracefs_dynevent_list_free() API Tzvetomir Stoyanov (VMware)
2021-11-01 14:21 ` [PATCH v2 00/12] libtracefs dynamic events support Tzvetomir Stoyanov

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=78703fa1-f811-a50c-6b5d-db68f6dbf21c@gmail.com \
    --to=y.karadz@gmail.com \
    --cc=linux-trace-devel@vger.kernel.org \
    --cc=rostedt@goodmis.org \
    --cc=tz.stoyanov@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).