Codestin Search App

OwenSanzas · 2026-03-09T23:51:03Z

Summary

do_evp_cipher() and do_evp_md() call parameter-setting functions (EVP_CIPHER_CTX_set_params / EVP_MD_CTX_set_params) before initializing the algorithm context (EVP_EncryptInit_ex2 / EVP_DigestInit_ex2). Since the context has no algorithm associated at that point, set_params always returns 0 and the function early-returns, making the cipher and digest paths dead code (~20% of all fuzzer inputs).

This PR swaps the call order so the context is initialized before parameters are set.

Additionally, key/iv buffers are now heap-allocated and sized to the cipher's actual key and IV length, since some ciphers (e.g. DES-EDE3-OFB) require buffers larger than the previous fixed 16/8-byte stack arrays.

Verification

PoC (single input, cipher path operation % 10 == 1):

Version	Execution time	Result
Original	0 ms	Dead code — `set_params` fails, function returns immediately
Fixed	19 ms	Cipher actually executes

Same for digest path (operation % 10 == 0): 0 ms → 23 ms.

Fuzzing (60s, ASan, fork mode):

Version	Crashes	Note
Original	0	Cipher path never reached
Fixed	44	Cipher path now active, finds real `stack-use-after-return` in DES OFB

KDF path (unmodified control): identical behavior in both versions (18–19 ms).

Fixes #30281

Both do_evp_cipher() and do_evp_md() call the parameter-setting function (EVP_CIPHER_CTX_set_params / EVP_MD_CTX_set_params) before initializing the algorithm context (EVP_EncryptInit_ex2 / EVP_DigestInit_ex2). Since the context has no algorithm associated at that point, set_params always returns 0 and the function early-returns, making the cipher and digest paths dead code (~20% of all fuzzer inputs). Fix by swapping the call order so the context is initialized first. Additionally, heap-allocate key/iv buffers sized to the cipher's actual key and IV length, since some ciphers (e.g. DES-EDE3-OFB) require buffers larger than the previous fixed 16/8-byte arrays. Fixes openssl#30281

nhorman · 2026-03-09T23:58:17Z

given that this is a non-trivial change, we will need you to submit a CLA prior to reviewing/accepting this change. If you could please fill out the form here:
https://openssl-library.org/policies/cla/

If you could submit the online form on that page, we will get you into our CLA database, and start taking a look at this.

Thank you!

npajkovsky · 2026-03-10T06:54:39Z

+    if (!EVP_CIPHER_CTX_set_params(ctx, param)) {
        EVP_CIPHER_CTX_free(ctx);
+        OPENSSL_free(key);
+        OPENSSL_free(iv);


nitpick: iv, key, and ctx can be set to NULL in the declaration, and then at the end of the function, you can have

err: EVP_CIPHER_CTX_free(ctx); OPENSSL_free(key); OPENSSL_free(iv); return 0; }

that will make the err path slightly nicer.

Thanks for the suggestion! I've updated the code — key, iv, and ctx are now initialized to NULL and all error paths go through a single err: label. Please take a look when you get a chance.

@npajkovsky

Initialize key, iv, and ctx to NULL at declaration and consolidate all cleanup into a single err label, as suggested by @npajkovsky.

kroeckx · 2026-03-10T08:41:39Z

+    iv = OPENSSL_zalloc(iv_len);
+    if (key == NULL || iv == NULL)
+        goto err;
+    for (i = 0; i < key_len && i < 16; i++)


Why the && i < 16?

Drop the `&& i < 16` and `&& i < 8` guards that were carried over from the original fixed-size arrays. The loops now fill the entire heap-allocated buffer, with values wrapping naturally via unsigned char.

t8m · 2026-03-10T10:51:54Z

The check style CI failure is relevant.

nhorman · 2026-03-10T11:51:26Z

@OwenSanzas as @t8m notes, this still needs to be fixed up:
https://github.com/openssl/openssl/actions/runs/22896362493/job/66439333050?pr=30331
That job provides a patch to apply to your PR that will fix the issue

OwenSanzas · 2026-03-10T16:12:49Z

Done, thanks for pointing that out.

openssl-machine · 2026-03-11T19:01:05Z

This pull request is ready to merge

nhorman · 2026-03-11T21:00:51Z

merged to master, thank you!

Both do_evp_cipher() and do_evp_md() call the parameter-setting function (EVP_CIPHER_CTX_set_params / EVP_MD_CTX_set_params) before initializing the algorithm context (EVP_EncryptInit_ex2 / EVP_DigestInit_ex2). Since the context has no algorithm associated at that point, set_params always returns 0 and the function early-returns, making the cipher and digest paths dead code (~20% of all fuzzer inputs). Fix by swapping the call order so the context is initialized first. Additionally, heap-allocate key/iv buffers sized to the cipher's actual key and IV length, since some ciphers (e.g. DES-EDE3-OFB) require buffers larger than the previous fixed 16/8-byte arrays. Fixes #30281 Reviewed-by: Neil Horman <[email protected]> Reviewed-by: Kurt Roeckx <[email protected]> MergeDate: Wed Mar 11 20:58:44 2026 (Merged from #30331)

@npajkovsky

Initialize key, iv, and ctx to NULL at declaration and consolidate all cleanup into a single err label, as suggested by @npajkovsky. Reviewed-by: Neil Horman <[email protected]> Reviewed-by: Kurt Roeckx <[email protected]> MergeDate: Wed Mar 11 20:58:47 2026 (Merged from #30331)

Drop the `&& i < 16` and `&& i < 8` guards that were carried over from the original fixed-size arrays. The loops now fill the entire heap-allocated buffer, with values wrapping naturally via unsigned char. Reviewed-by: Neil Horman <[email protected]> Reviewed-by: Kurt Roeckx <[email protected]> MergeDate: Wed Mar 11 20:58:49 2026 (Merged from #30331)

Reviewed-by: Neil Horman <[email protected]> Reviewed-by: Kurt Roeckx <[email protected]> MergeDate: Wed Mar 11 20:58:51 2026 (Merged from #30331)

Reviewed-by: Neil Horman <[email protected]> Reviewed-by: Kurt Roeckx <[email protected]> MergeDate: Wed Mar 11 20:58:54 2026 (Merged from #30331)

openssl-machine added the hold: cla required The contributor needs to submit a license agreement label Mar 9, 2026

OwenSanzas mentioned this pull request Mar 9, 2026

fuzz/provider.c: incorrect API call sequence makes cipher and digest paths dead code #30281

Closed

paulidale previously approved these changes Mar 10, 2026

View reviewed changes

openssl-machine added the approval: review pending This pull request needs review by a committer label Mar 10, 2026

npajkovsky previously approved these changes Mar 10, 2026

View reviewed changes

openssl-machine added approval: done This pull request has the required number of approvals and removed approval: review pending This pull request needs review by a committer labels Mar 10, 2026

npajkovsky closed this Mar 10, 2026

npajkovsky reopened this Mar 10, 2026

Refactor do_evp_cipher error paths to use goto err

d5d2057

Initialize key, iv, and ctx to NULL at declaration and consolidate all cleanup into a single err label, as suggested by @npajkovsky.

OwenSanzas dismissed stale reviews from npajkovsky and paulidale via d5d2057 March 10, 2026 07:57

openssl-machine added approval: review pending This pull request needs review by a committer and removed approval: done This pull request has the required number of approvals labels Mar 10, 2026

npajkovsky previously approved these changes Mar 10, 2026

View reviewed changes

kroeckx reviewed Mar 10, 2026

View reviewed changes

Remove unnecessary caps in key/iv initialization loops

98a6cbe

Drop the `&& i < 16` and `&& i < 8` guards that were carried over from the original fixed-size arrays. The loops now fill the entire heap-allocated buffer, with values wrapping naturally via unsigned char.

OwenSanzas dismissed npajkovsky’s stale review via 98a6cbe March 10, 2026 09:31

openssl-machine removed the hold: cla required The contributor needs to submit a license agreement label Mar 10, 2026

kroeckx previously approved these changes Mar 10, 2026

View reviewed changes

Remove unnecessary comments and extra blank line

23d6488

OwenSanzas dismissed kroeckx’s stale review via 23d6488 March 10, 2026 09:42

kroeckx previously approved these changes Mar 10, 2026

View reviewed changes

npajkovsky previously approved these changes Mar 10, 2026

View reviewed changes

openssl-machine added approval: done This pull request has the required number of approvals and removed approval: review pending This pull request needs review by a committer labels Mar 10, 2026

t8m added the branch: master Applies to master branch label Mar 10, 2026

t8m added triaged: feature The issue/pr requests/adds a feature tests: exempted The PR is exempt from requirements for testing labels Mar 10, 2026

Fix clang-format: remove leading space from err label

32a30ba

OwenSanzas dismissed stale reviews from npajkovsky and kroeckx via 32a30ba March 10, 2026 16:11

openssl-machine added approval: review pending This pull request needs review by a committer and removed approval: done This pull request has the required number of approvals labels Mar 10, 2026

kroeckx approved these changes Mar 10, 2026

View reviewed changes

nhorman approved these changes Mar 10, 2026

View reviewed changes

nhorman closed this Mar 11, 2026

openssl-machine pushed a commit that referenced this pull request Mar 11, 2026

Remove unnecessary comments and extra blank line

8835a9c

Reviewed-by: Neil Horman <[email protected]> Reviewed-by: Kurt Roeckx <[email protected]> MergeDate: Wed Mar 11 20:58:51 2026 (Merged from #30331)

openssl-machine pushed a commit that referenced this pull request Mar 11, 2026

Fix clang-format: remove leading space from err label

30e3a7a

Reviewed-by: Neil Horman <[email protected]> Reviewed-by: Kurt Roeckx <[email protected]> MergeDate: Wed Mar 11 20:58:54 2026 (Merged from #30331)

t8m mentioned this pull request Mar 12, 2026

Fix stack buffer over-read in DES OFB/CFB64 via unchecked num parameter #30332

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fuzz/provider.c: fix API call order in do_evp_cipher and do_evp_md#30331

fuzz/provider.c: fix API call order in do_evp_cipher and do_evp_md#30331
OwenSanzas wants to merge 5 commits into
openssl:masterfrom
OwenSanzas:fix/provider-fuzzer-api-order

OwenSanzas commented Mar 9, 2026

Uh oh!

nhorman commented Mar 9, 2026

Uh oh!

npajkovsky Mar 10, 2026

Uh oh!

OwenSanzas Mar 10, 2026

Uh oh!

kroeckx Mar 10, 2026

Uh oh!

t8m commented Mar 10, 2026

Uh oh!

nhorman commented Mar 10, 2026

Uh oh!

OwenSanzas commented Mar 10, 2026

Uh oh!

openssl-machine commented Mar 11, 2026

Uh oh!

nhorman commented Mar 11, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants

Uh oh!

Conversation

OwenSanzas commented Mar 9, 2026

Summary

Verification

Uh oh!

nhorman commented Mar 9, 2026

Uh oh!

npajkovsky Mar 10, 2026

Choose a reason for hiding this comment

Uh oh!

OwenSanzas Mar 10, 2026

Choose a reason for hiding this comment

Uh oh!

kroeckx Mar 10, 2026

Choose a reason for hiding this comment

Uh oh!

t8m commented Mar 10, 2026

Uh oh!

nhorman commented Mar 10, 2026

Uh oh!

OwenSanzas commented Mar 10, 2026

Uh oh!

openssl-machine commented Mar 11, 2026

Uh oh!

nhorman commented Mar 11, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants