[
https://issues.apache.org/jira/browse/PROTON-2954?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18106697#comment-18106697
]
ASF subversion and git services commented on PROTON-2954:
---------------------------------------------------------
Commit 717d1bde2dc95fc1b259b9b5eaaec1647e73b4fd in qpid-proton's branch
refs/heads/main from Andrew Stitcher
[ https://gitbox.apache.org/repos/asf?p=qpid-proton.git;h=717d1bde2 ]
PROTON-2954: Further reduce pn_data_t memory use
Reduced the per node memory use to 16 bytes from 24 bytes. This is a
4.5 times reductions over the original 72 bytes. This now mitigates
significantly the maximum node storage that could occur to 1MiB.
Practically the intern buffer storage should always be limited by the
size of the raw protocol bytes that are being decoded.
Achieved by:
- Moving navigation links into the specific node type that uses them.
- Removing members that could be computed.
- Using spare packing space for the base type of an AMQP array.
- Splitting arrays into regular and arrays of described types.
Surprisingly this hasn't made the code significantly more complex and
has simplified a few things.
> Decoding AMQP value into pn_data_t instances can use a lot of memory
> --------------------------------------------------------------------
>
> Key: PROTON-2954
> URL: https://issues.apache.org/jira/browse/PROTON-2954
> Project: Qpid Proton
> Issue Type: Bug
> Components: proton-c
> Affects Versions: proton-c-0.40.0
> Reporter: Andrew Stitcher
> Assignee: Andrew Stitcher
> Priority: Major
> Fix For: proton-c-0.41.0
>
>
> Proton has limited ways to bound the amount of memory used when decoding peer
> controlled AMQP values into a pn_data_t structure.
> There are 2 parts to the structure that use space: A list of nodes which
> represent the structure of the data and a backing buffer for some of the
> types of value which are variable length or long like strings or binary data.
> There *is* a limit to the number of nodes which is 65535 in the current code,
> but that still equates to a lot of memory used for he nodes; but there
> currently is no limit to the memory used for the backing data.
> This could cause problems with malformed or malicious input data.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
