From: Rich Felker <dalias@libc.org>
To: musl@lists.openwall.com
Subject: Re: Reviving planned ldso changes
Date: Wed, 4 Jan 2017 14:36:27 -0500 [thread overview]
Message-ID: <20170104193627.GO1555@brightrain.aerifal.cx> (raw)
In-Reply-To: <20170104062203.GN1555@brightrain.aerifal.cx>
[-- Attachment #1: Type: text/plain, Size: 2062 bytes --]
On Wed, Jan 04, 2017 at 01:22:03AM -0500, Rich Felker wrote:
> On Wed, Jan 04, 2017 at 01:06:40AM -0500, Rich Felker wrote:
> > diff --git a/ldso/dynlink.c b/ldso/dynlink.c
> > index c689084..cb82b2c 100644
> > --- a/ldso/dynlink.c
> > +++ b/ldso/dynlink.c
> > @@ -67,6 +67,7 @@ struct dso {
> > char constructed;
> > char kernel_mapped;
> > struct dso **deps, *needed_by;
> > + size_t next_dep;
> > char *rpath_orig, *rpath;
> > struct tls_module tls;
> > size_t tls_id;
> > @@ -1211,13 +1212,14 @@ void __libc_exit_fini()
> > static void do_init_fini(struct dso *p)
> > {
> > size_t dyn[DYN_CNT];
> > - int need_locking = libc.threads_minus_1;
> > - /* Allow recursive calls that arise when a library calls
> > - * dlopen from one of its constructors, but block any
> > - * other threads until all ctors have finished. */
> > - if (need_locking) pthread_mutex_lock(&init_fini_lock);
> > - for (; p; p=p->prev) {
> > - if (p->constructed) continue;
> > + pthread_mutex_lock(&init_fini_lock);
> > + while (!p->constructed) {
> > + while (p->deps[p->next_dep] && p->deps[p->next_dep]->constructed)
> > + p->next_dep++;
> > + if (p->deps[p->next_dep]) {
> > + p = p->deps[p->next_dep++];
> > + continue;
> > + }
>
> I think this logic is probably broken in the case of circular
> dependencies, and will end up skipping ctors for some deps due to the
> increment in this line:
>
> + p = p->deps[p->next_dep++];
>
> which happens before the new p's ctors have actually run. Omitting the
> increment here, however, would turn it into an infinite loop. We need
> some way to detect this case and let the dso with an indirect
> dependency on itself run its ctors once all non-circular deps have
> been satisfied. I don't now the right algorithm for this right off,
> though; suggestions would be welcome.
Here's a v2 of the patch with the above issues fixed, and some
comments that hopefully make it make sense. I still think there's more
logic needed to allow concurrent ctors from unrelated dlopen in
multiple threads, though.
Rich
[-- Attachment #2: ctor_dep_order_v2.diff --]
[-- Type: text/plain, Size: 2681 bytes --]
diff --git a/ldso/dynlink.c b/ldso/dynlink.c
index c689084..fd59389 100644
--- a/ldso/dynlink.c
+++ b/ldso/dynlink.c
@@ -67,6 +67,7 @@ struct dso {
char constructed;
char kernel_mapped;
struct dso **deps, *needed_by;
+ size_t next_dep;
char *rpath_orig, *rpath;
struct tls_module tls;
size_t tls_id;
@@ -1090,9 +1091,12 @@ static void load_deps(struct dso *p)
if (runtime) longjmp(*rtld_fail, 1);
continue;
}
- if (runtime) {
- tmp = realloc(*deps, sizeof(*tmp)*(ndeps+2));
- if (!tmp) longjmp(*rtld_fail, 1);
+ tmp = realloc(*deps, sizeof(*tmp)*(ndeps+2));
+ if (!tmp) {
+ error("Error allocating dependency data for %s: %m",
+ p->name);
+ if (runtime) longjmp(*rtld_fail, 1);
+ } else {
tmp[ndeps++] = dep;
tmp[ndeps] = 0;
*deps = tmp;
@@ -1211,13 +1215,21 @@ void __libc_exit_fini()
static void do_init_fini(struct dso *p)
{
size_t dyn[DYN_CNT];
- int need_locking = libc.threads_minus_1;
- /* Allow recursive calls that arise when a library calls
- * dlopen from one of its constructors, but block any
- * other threads until all ctors have finished. */
- if (need_locking) pthread_mutex_lock(&init_fini_lock);
- for (; p; p=p->prev) {
- if (p->constructed) continue;
+ pthread_mutex_lock(&init_fini_lock);
+ /* Construct in dependency order without any recursive state. */
+ while (p && !p->constructed) {
+ /* The following loop descends into the first dependency
+ * that is neither alredy constructed nor pending
+ * construction due to circular deps, stopping only
+ * when it reaches a dso with no remaining dependencies
+ * to descend into. */
+ while (p->deps && p->deps[p->next_dep]) {
+ if (!p->deps[p->next_dep]->constructed &&
+ !p->deps[p->next_dep]->next_dep)
+ p = p->deps[p->next_dep++];
+ else
+ p->next_dep++;
+ }
p->constructed = 1;
decode_vec(p->dynv, dyn, DYN_CNT);
if (dyn[0] & ((1<<DT_FINI) | (1<<DT_FINI_ARRAY))) {
@@ -1233,12 +1245,14 @@ static void do_init_fini(struct dso *p)
size_t *fn = laddr(p, dyn[DT_INIT_ARRAY]);
while (n--) ((void (*)(void))*fn++)();
}
- if (!need_locking && libc.threads_minus_1) {
- need_locking = 1;
- pthread_mutex_lock(&init_fini_lock);
- }
- }
- if (need_locking) pthread_mutex_unlock(&init_fini_lock);
+ /* Revisit "parent" dso which caused the just-constructed
+ * dso to be pulled in as a dependency. On the next loop
+ * iteration we will either descend to construct a sibling
+ * of the just-constructed dso, or finish constructing the
+ * parent if no unfinished deps remain. */
+ p = p->needed_by;
+ }
+ pthread_mutex_unlock(&init_fini_lock);
}
void __libc_start_init(void)
next prev parent reply other threads:[~2017-01-04 19:36 UTC|newest]
Thread overview: 21+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-01-03 5:43 Rich Felker
2017-01-04 6:06 ` Rich Felker
2017-01-04 6:22 ` Rich Felker
2017-01-04 19:36 ` Rich Felker [this message]
2017-01-14 21:30 ` A. Wilcox
2017-01-15 17:44 ` Rich Felker
2017-02-26 1:04 ` Szabolcs Nagy
2017-02-26 1:39 ` Rich Felker
2017-02-26 10:28 ` Szabolcs Nagy
2017-02-26 15:20 ` Rich Felker
2017-02-26 15:34 ` Szabolcs Nagy
2017-02-26 21:39 ` Rich Felker
2017-03-03 1:30 ` Rich Felker
2017-03-04 10:58 ` Szabolcs Nagy
2017-03-06 1:11 ` Rich Felker
2017-03-07 22:02 ` Rich Felker
2017-03-08 18:55 ` Rich Felker
2017-03-06 16:25 ` Rich Felker
2017-01-04 10:51 ` Szabolcs Nagy
2017-02-16 1:58 ` Szabolcs Nagy
2017-02-16 2:39 ` Rich Felker
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=20170104193627.GO1555@brightrain.aerifal.cx \
--to=dalias@libc.org \
--cc=musl@lists.openwall.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.
Code repositories for project(s) associated with this public inbox
https://git.vuxu.org/mirror/musl/
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).