Skip to content

openssl: fix the data race when sharing an SSL session between threads#14751

Closed
AkiSakurai wants to merge 3 commits intocurl:masterfrom
AkiSakurai:share_ssl_session
Closed

openssl: fix the data race when sharing an SSL session between threads#14751
AkiSakurai wants to merge 3 commits intocurl:masterfrom
AkiSakurai:share_ssl_session

Conversation

@AkiSakurai
Copy link
Contributor

@AkiSakurai AkiSakurai commented Sep 1, 2024

The SSL_Session object is mutated during connection inside openssl, and it might not be thread-safe. Besides, according to documentation of openssl:

SSL_SESSION objects keep internal link information about the session
cache list, when being inserted into one SSL_CTX object's session
cache. One SSL_SESSION object, regardless of its reference count,
must therefore only be used with one SSL_CTX object (and the SSL
objects created from this SSL_CTX object).

If I understand correctly, it is not safe to share it even in a single thread.

Instead, serialize the SSL_SESSION before adding it to the cache, and deserialize it after retrieving it from the cache, so that concurrent write to the same object is infeasible.

Also

  • add a ci test for thread sanitizer
  • add a test for sharing ssl sessions concurrently
  • avoid redefining memory functions when not building libcurl, but including the source in libtest
  • increase the concurrent connections limit in sws

Notice that there are fix for a global data race for openssl which is not yet release. The fix is cherry pick for the ci test with thread sanitizer.
openssl/openssl@d8def79

The SSL_Session object is mutated during connection inside openssl,
and it might not be thread-safe. Besides, according to documentation
of openssl:

```
SSL_SESSION objects keep internal link information about the session
cache list, when being inserted into one SSL_CTX object's session
cache. One SSL_SESSION object, regardless of its reference count,
must therefore only be used with one SSL_CTX object (and the SSL
objects created from this SSL_CTX object).
```
If I understand correctly, it is not safe to share it even in a
single thread.

Instead, serialize the SSL_SESSION before adding it to the cache,
and deserialize it after retrieving it from the cache, so that no
concurrent write to the same object is infeasible.

Also
 - add a ci test for thread sanitizer
 - add a test for sharing ssl sessions concurrently
 - avoid redefining memory functions when not building libcurl, but
   including the soruce in libtest
 - increase the concurrent connections limit in sws

Notice that there are fix for a global data race for openssl which
is not yet release. The fix is cherry pick for the ci test with
thread sanitizer.
openssl/openssl@d8def79
@github-actions github-actions bot added tests CI Continuous Integration labels Sep 1, 2024
Co-authored-by: Viktor Szakats <vszakats@users.noreply.github.com>
@icing
Copy link
Contributor

icing commented Sep 2, 2024

I believe this PR makes a necessary change. Well done.

@vszakats vszakats added the TLS label Sep 2, 2024
@bagder bagder closed this in a2bcec0 Sep 2, 2024
@bagder
Copy link
Member

bagder commented Sep 2, 2024

Thanks!

vszakats added a commit that referenced this pull request Oct 23, 2024
The patch is now part of the 3.4.0 stable release.
(Turns out it was part of 3.3.2 already.)

Also:
- rename this local build to match the scheme used with wolfssl.
- drop '3' from local openssl build name.
- sync job name with others.
- quote step names where missing.

Follow-up to a2bcec0 #14751
Closes #15379
pps83 pushed a commit to pps83/curl that referenced this pull request Apr 26, 2025
The patch is now part of the 3.4.0 stable release.
(Turns out it was part of 3.3.2 already.)

Also:
- rename this local build to match the scheme used with wolfssl.
- drop '3' from local openssl build name.
- sync job name with others.
- quote step names where missing.

Follow-up to a2bcec0 curl#14751
Closes curl#15379
vszakats added a commit that referenced this pull request Aug 13, 2025
Replace autotools with cmake to avoid libtool wrappers that are changing
`LD_LIBRARY_PATH` in a way incompatible with the thread sanitizer.

To fix the output when the sanitizier is finding something:
```
==51718==WARNING: Can't write to symbolizer at fd 7
 /usr/bin/llvm-symbolizer-18: /home/runner/work/curl/curl/bld/lib/.libs/libcurl.so.4: no version information available (required by /usr/bin/llvm-symbolizer-18)
 /usr/bin/llvm-symbolizer-18: symbol lookup error: /home/runner/openssl/lib/libcrypto.so.3: undefined symbol: __tsan_func_entry
```
Ref: https://github.com/curl/curl/actions/runs/16911402500/job/47913783729#step:39:4466

After:
```
 13:50:04.117885 == Info:ThreadSanitizer: thread T1  finished with ignores enabled, created at:
  closing connection #0
     #0 pthread_create <null> (libtests+0x6bc0f) (BuildId: 4fe889446291259934205ac03931c397aa0210d3)
     #1 Curl_thread_create /home/runner/work/curl/curl/lib/curl_threads.c:73:6 (libcurl.so.4+0x55a76) (BuildId: cb0f14ba2ad68c9cab0c980d9a5d7a53cc0782da)
     #2 async_thrdd_init /home/runner/work/curl/curl/lib/asyn-thrdd.c:500:26 (libcurl.so.4+0x1c153) (BuildId: cb0f14ba2ad68c9cab0c980d9a5d7a53cc0782da)
[...]
```
Ref: https://github.com/curl/curl/actions/runs/16939193922/job/48003405272?pr=18274#step:39:4018

Also:
- disable memory tracker which turned out to be incompatible with
  the thread sanitizer and detaching threads.
  Ref: #18263 and #curl IRC.
- the job is ~30 seconds faster after this patch.

Reported-by: Stefan Eissing
Bug: #18263 (comment)
Follow-up to a2bcec0 #14751
Closes #18274
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

CI Continuous Integration tests TLS

Development

Successfully merging this pull request may close these issues.

4 participants