All of lore.kernel.org
 help / color / mirror / Atom feed
From: Kris Van Hees <kris.van.hees@oracle.com>
To: Nick Alcock <nick.alcock@oracle.com>
Cc: dtrace@lists.linux.dev, dtrace-devel@oss.oracle.com
Subject: Re: [DTrace-devel] [PATCH v2 05/14] probe: improve dt_probe_lookup2()
Date: Tue, 29 Oct 2024 15:50:38 -0400	[thread overview]
Message-ID: <ZyE8jj3mle7c9NKl@oracle.com> (raw)
In-Reply-To: <20241028211803.458685-6-nick.alcock@oracle.com>

On Mon, Oct 28, 2024 at 09:17:54PM +0000, Nick Alcock via DTrace-devel wrote:
> This function had numerous problems:
> 
>  - it broke _FORTIFY_SOURCE due to using a function that snprintf()ed
>    with a size of INT_MAX, into a much smaller alloca()ed buffer
>  - it claimed to call down to the dtrace kernel module for a probe
>    description (long obsolete)
>  - it invariably leaked the probe description it constructed by calling
>    dtrace_xstr2desc()
>  - it did meaningless errno testing to determine whether to return
>    EDT_NOPROBE on error, which more or less led to EDT_NOPROBE never
>    being returned even though it should always be in this case (there
>    is no other cause for the ID hash lookup failing).
> 
> Fix the lot: while we're at it, implement a dt_desc_destroy() to
> undo the strdup()s done by dtrace_xstr2desc(), and use it additionally
> in dt_probe_destroy() instead of an identical sequence of dt_free()s.
> 
> Bug: https://github.com/oracle/dtrace-utils/issues/78
> Signed-off-by: Nick Alcock <nick.alcock@oracle.com>
> ---
>  libdtrace/dt_impl.h  |  2 ++
>  libdtrace/dt_probe.c | 44 +++++++++++++-------------------------------
>  libdtrace/dt_subr.c  | 10 ++++++++++
>  3 files changed, 25 insertions(+), 31 deletions(-)
> 
> diff --git a/libdtrace/dt_impl.h b/libdtrace/dt_impl.h
> index 340dc1960c5e..d11ad2839f8e 100644
> --- a/libdtrace/dt_impl.h
> +++ b/libdtrace/dt_impl.h
> @@ -759,6 +759,8 @@ extern dtrace_difo_t *dt_difo_copy(dtrace_hdl_t *dtp, const dtrace_difo_t *odp);
>  extern int dt_consume_init(dtrace_hdl_t *);
>  extern void dt_consume_fini(dtrace_hdl_t *);
>  
> +extern void dt_desc_destroy(dtrace_hdl_t *dtp, dtrace_probedesc_t *);

I think it would be better to let dt_desc_destroy() take a third arg, to
indicate whether the desc itself should be free'd as well or whether it
should be left alone.

That avoids calling it and then following it up with a dt_free() on the desc.

> +
>  extern dtrace_datadesc_t *dt_datadesc_hold(dtrace_datadesc_t *ddp);
>  extern void dt_datadesc_release(dtrace_hdl_t *, dtrace_datadesc_t *);
>  extern dtrace_datadesc_t *dt_datadesc_create(dtrace_hdl_t *);
> diff --git a/libdtrace/dt_probe.c b/libdtrace/dt_probe.c
> index 686e2a661253..f84581687caf 100644
> --- a/libdtrace/dt_probe.c
> +++ b/libdtrace/dt_probe.c
> @@ -172,24 +172,9 @@ dt_probe_alloc_args(dt_probe_t *prp, int nargc, int xargc)
>  	}
>  }
>  
> -static size_t
> -dt_probe_keylen(const dtrace_probedesc_t *pdp)
> -{
> -	return strlen(pdp->mod) + 1 + strlen(pdp->fun) + 1 +
> -		strlen(pdp->prb) + 1;
> -}
> -
> -static char *
> -dt_probe_key(const dtrace_probedesc_t *pdp, char *s)
> -{
> -	snprintf(s, INT_MAX, "%s:%s:%s", pdp->mod, pdp->fun, pdp->prb);
> -	return s;
> -}
> -
>  /*
>   * Lookup a probe declaration based on a known provider and full or partially
> - * specified module, function, and name.  If the probe is not known to us yet,
> - * ask dtrace(7D) to match the description and then cache any useful results.
> + * specified module, function, and name.
>   */
>  dt_probe_t *
>  dt_probe_lookup2(dt_provider_t *pvp, const char *s)
> @@ -197,28 +182,28 @@ dt_probe_lookup2(dt_provider_t *pvp, const char *s)
>  	dtrace_hdl_t *dtp = pvp->pv_hdl;
>  	dtrace_probedesc_t pd;
>  	dt_ident_t *idp;
> -	size_t keylen;
>  	char *key;
>  
>  	if (dtrace_str2desc(dtp, DTRACE_PROBESPEC_NAME, s, &pd) != 0)
>  		return NULL; /* dt_errno is set for us */
>  
> -	keylen = dt_probe_keylen(&pd);
> -	key = dt_probe_key(&pd, alloca(keylen));
> +	if (asprintf(&key, "%s:%s:%s", pd.mod, pd.fun, pd.prb) == -1) {
> +		dt_set_errno(dtp, errno);
> +		goto out;
> +	}
>  
>  	/*
>  	 * If the probe is already declared, then return the dt_probe_t from
> -	 * the existing identifier.  This could come from a static declaration
> -	 * or it could have been cached from an earlier call to this function.
> +	 * the existing identifier.
>  	 */
> -	if ((idp = dt_idhash_lookup(pvp->pv_probes, key)) != NULL)
> +	if ((idp = dt_idhash_lookup(pvp->pv_probes, key)) != NULL) {
> +		dt_desc_destroy(dtp, &pd);
>  		return idp->di_data;
> +	}
>  
> -	if (errno == ESRCH || errno == EBADF)
> -		dt_set_errno(dtp, EDT_NOPROBE);
> -	else
> -		dt_set_errno(dtp, errno);
> -
> +	dt_set_errno(dtp, EDT_NOPROBE);
> + out:
> +	dt_desc_destroy(dtp, &pd);
>  	return NULL;
>  }
>  
> @@ -387,10 +372,7 @@ dt_probe_destroy(dt_probe_t *prp)
>  	dt_free(dtp, prp->mapping);
>  	dt_free(dtp, prp->argv);
>  	if (prp->desc) {
> -		dt_free(dtp, (void *)prp->desc->prv);
> -		dt_free(dtp, (void *)prp->desc->mod);
> -		dt_free(dtp, (void *)prp->desc->fun);
> -		dt_free(dtp, (void *)prp->desc->prb);
> +		dt_desc_destroy(dtp, (dtrace_probedesc_t *)prp->desc);
>  		dt_free(dtp, (void *)prp->desc);
>  	}
>  	dt_free(dtp, prp);
> diff --git a/libdtrace/dt_subr.c b/libdtrace/dt_subr.c
> index d6aad7637fb9..72631b33a0ad 100644
> --- a/libdtrace/dt_subr.c
> +++ b/libdtrace/dt_subr.c
> @@ -175,6 +175,16 @@ dtrace_desc2str(const dtrace_probedesc_t *pdp, char *buf, size_t len)
>  	return buf;
>  }
>  
> +/* Only use on probedescs derived from dtrace_xstr2desc above. */

Why can't it used on other dtrace_probedesc_t that contain references to
memory that needs to be free'd?  I don't think that dtrace_xstr2desc() does
anything to make the desc it populates be special.

> +void
> +dt_desc_destroy(dtrace_hdl_t *dtp, dtrace_probedesc_t *pdp)
> +{
> +	dt_free(dtp, (void *) pdp->prv);
> +	dt_free(dtp, (void *) pdp->mod);
> +	dt_free(dtp, (void *) pdp->fun);
> +	dt_free(dtp, (void *) pdp->prb);
> +}
> +
>  char *
>  dtrace_attr2str(dtrace_attribute_t attr, char *buf, size_t len)
>  {
> -- 
> 2.46.0.278.g36e3a12567
> 
> 
> _______________________________________________
> DTrace-devel mailing list
> DTrace-devel@oss.oracle.com
> https://oss.oracle.com/mailman/listinfo/dtrace-devel

  reply	other threads:[~2024-10-29 19:50 UTC|newest]

Thread overview: 20+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-10-28 21:17 [PATCH v2 00/14] gentoo, manpage, and assorted other small fixes Nick Alcock
2024-10-28 21:17 ` [PATCH v2 01/14] No longer depend on libsystemd Nick Alcock
2024-10-29 18:27   ` Kris Van Hees
2024-10-29 21:51     ` Nick Alcock
2024-10-28 21:17 ` [PATCH v2 02/14] pkgconfig: drop spaces in variable decls Nick Alcock
2024-10-28 21:17 ` [PATCH v2 03/14] configure, build: make valgrind optional Nick Alcock
2024-10-28 21:17 ` [PATCH v2 04/14] build: substitute LIBDIR in pkg-config files Nick Alcock
2024-10-28 21:17 ` [PATCH v2 05/14] probe: improve dt_probe_lookup2() Nick Alcock
2024-10-29 19:50   ` Kris Van Hees [this message]
2024-10-29 21:57     ` [DTrace-devel] " Nick Alcock
2024-10-28 21:17 ` [PATCH v2 06/14] configure: fix dreadful behaviour of MANDIR / --mandir Nick Alcock
2024-10-28 21:17 ` [PATCH v2 07/14] man: the synopsis is ended with .YS, not .SY Nick Alcock
2024-10-28 21:17 ` [PATCH v2 08/14] man: use \- for option dashes, not - Nick Alcock
2024-10-28 21:17 ` [PATCH v2 09/14] man: drop blank lines Nick Alcock
2024-10-28 21:17 ` [PATCH v2 10/14] man: fix blank line in environment variables list Nick Alcock
2024-10-28 21:18 ` [PATCH v2 11/14] dtprobed: fix parser child timeout Nick Alcock
2024-10-28 21:18 ` [PATCH v2 12/14] man: add manpage for dtprobed(8) Nick Alcock
2024-10-29 21:19   ` Kris Van Hees
2024-10-28 21:18 ` [PATCH v2 13/14] man: drop double-\fB at the start of every option line Nick Alcock
2024-10-28 21:18 ` [PATCH v2 14/14] man: \fP-ize Nick Alcock

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=ZyE8jj3mle7c9NKl@oracle.com \
    --to=kris.van.hees@oracle.com \
    --cc=dtrace-devel@oss.oracle.com \
    --cc=dtrace@lists.linux.dev \
    --cc=nick.alcock@oracle.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.