Wip chunkfix #35
Open
Wip chunkfix #35
Conversation
This file contains 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
transfer-encoding = chunked vs. content-length encoding is documented in rfc 2616 3.6.1 and rfc 7230 4.1. In both cases, it is documented that if transfer-encoding == 'chunked', then content-length should not be supplied by a client, and must not be processed by a server. The previous code in civetweb allowed content-length to override transfer-encoding=chunked, which is clearly wrong by the standard. This chnage corrects the order of those checks. Signed-off-by: Marcus Watts <[email protected]>
In rfc 2616 3.6.1 and rfc 7230 4.1, a mechanism is provided to supply extra data in chunk headers, in the of keyword/value pairs. Old versions of civetweb completely ignored this data. The boto2 s3 client, when it uses chunked encoding, sends a per-chunk header consisting of a hex count followed by a single ';'. This is not, according to a strict reading of the rfcs, legal, but it is something that ought to be acceptable to any server that does not expect to interpret any keyword values. This behavior is not likely to change. Additionally, the final chunk header is supposed to be followed by not one, but two crlf pairs. It is desirable to consume both pairs rather than just leaving the 2nd pair to cause trouble later. Older versions of civetweb did no checking of any possible chunk header data, they only checked to see if they could decode a hexadecimal count. This is clearly more generous than necessary, but the change to be more strict broke boto2 clinets, which is not desirable. This change provides slightly more general logic that allows for a ';' followed by optional data (which is unconditionally discarded for now), and also consumes the extra crlf pair. Signed-off-by: Marcus Watts <[email protected]>
In some cases, ceph may detect an error that indicates that further use of this connection is doubtful. For instance, mg_read() fails to return expected data, due to a chunk decoding error. In this case, the connection should be closed, but not until after the error is sent. civetweb provides low-level logic to do this (conn->must_close); this commit provides an API for the application to trigger that behavior. Signed-off-by: Marcus Watts <[email protected]>
|
This PR is a prerequesite for this PR to ceph: ceph/ceph#35350 |
|
The build failures appear to be due ot preexisting conditions unrelated to this fix. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
3 fixes here.
1 content-length vs. transfer-encoding: chunked; according to rfc 2616,
transfer-encoding chunked should supersede content-length. Later rfcs require this be treated as an error - sadly experience has shown s3 clients lag in this respect.
boto2 fixes. Older versions of ceph (up to mid-luminous) worked with boto2; but the newer civetweb adopted in msater then backported to luminous is pickier and hates boto2. Even though boto2 appears to be abandonware it still seems like a good idea (to me) to fix this.
mg_set_must_close, Gives ceph/rgw a way to say this connection is hosed, do not attempt to read another request out of it.
These are all issues from,
https://tracker.ceph.com/issues/45789