[PDO] Fix leak of driver_params argument of bindParam() and bindColumn() - #282
Closed
iliaal wants to merge 5 commits into
Closed
[PDO] Fix leak of driver_params argument of bindParam() and bindColumn()#282iliaal wants to merge 5 commits into
iliaal wants to merge 5 commits into
Conversation
sapi_cli_server_send_headers() sent the header buffer through php_cli_server_client_send_through() but dropped the return value, so it always reported SAPI_HEADER_SENT_SUCCESSFULLY, even when the send failed and php_handle_aborted_connection() did not bail out (ignore_user_abort=1). It now compares the send result against the buffer length and returns SAPI_HEADER_SEND_FAILED when it comes up short. php_cli_server_client_send_through() itself returned the number of bytes left on failure, which equals str_len when nothing was sent: the same value it returns on success. It now returns the number of bytes actually sent on both paths. Since sapi_send_headers() resets headers_sent on SAPI_HEADER_SEND_FAILED, a headers_written flag on the client guards against rebuilding and resending the whole header block on every subsequent output write. Close phpGH-23428
…ams (php#23282) Keep the native archive state alive independently of the ZipArchive object, allowing streams to remain valid without retaining the object itself.
uloc_getKeywordValue failure destroyed the result array and returned without uenum_close(). Close the enumeration on that path. The success path already closes it. Audited the other uenum_close site in this file (acceptLanguage). Closes phpGH-23351
register_bound_param() took a reference on the driver_params argument but never released it: really_register_bound_param() ADDREFs another reference for the bound-params hash, yet its early failure paths after that point (rewrite_name_to_position(), PDO_PARAM_EVT_NORMALIZE hook) and every return path of register_bound_param() itself dropped only param.parameter, leaking one or two references per call. 200k failing bindParam() calls grow memory by ~128MB. The transient copy is now released on both failure and success, and really_register_bound_param() releases its hash-bound reference on early failures; the PDO_PARAM_EVT_ALLOC hook failure path already released it via the hash dtor. Sibling audit: bindValue() and execute()'s input_params loop leave driver_params undefined so the new releases are no-ops there.
iliaal
force-pushed
the
fix/pdo-bindparam-driverparams-leak-84
branch
from
August 25, 2026 23:10
b8a1a66 to
fffa83c
Compare
Owner
Author
|
Promoted as php#23463. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
PDOStatement::bindParam() and PDOStatement::bindColumn() leak every reference taken on the driver_params argument: register_bound_param() never releases its transient copy on return, and really_register_bound_param() ADDREFs a second reference for the bound-parameters hash but drops it nowhere when rewrite_name_to_position() or the PDO_PARAM_EVT_NORMALIZE param_hook fails, so 200k failing bindParam() calls grow memory by roughly 128MB and even successful calls leak one reference each. With this change really_register_bound_param() releases the hash-bound reference before those two early returns and register_bound_param() releases the transient copy on both outcomes; the PDO_PARAM_EVT_ALLOC hook failure needs no new code since deleting the hash entry frees it through param_dtor().