SSL_write_ex on stream object sometimes retransmits parts of stream data #924
Assignees
Milestone
Comments
Closed
nhorman
added a commit
to nhorman/openssl
that referenced
this issue
Nov 19, 2024
When writing to a blocking quic stream, we sometimes get duplicate transmitted data. This occurs when a call to quic_write_blocking has to wait for space to become available in the ring buffer. When we do a wait, the call sets *written to the value returned in args.total_written as filled out by the calls to block_until_pred->quic_write_again. However, the value there is based on the amount we requested, which is only the remaining data that we didn't append in xso_sstream_write. So if we call quic_write_blocking with a buffer of length X, and initially append Y bytes, and write the remainig X-Y bytes via a block_until_pred call, then *written will return with the value X-Y, even though we wrote the full X bytes to the ring buffer. Fix it by recording the initial amount appended into *written, and then add the args.total_written value if we have to wait on more space Fixes openssl/project#924
nhorman
added a commit
to nhorman/openssl
that referenced
this issue
Nov 19, 2024
When writing to a blocking quic stream, we sometimes get duplicate transmitted data. This occurs when a call to quic_write_blocking has to wait for space to become available in the ring buffer. When we do a wait, the call sets *written to the value returned in args.total_written as filled out by the calls to block_until_pred->quic_write_again. However, the value there is based on the amount we requested, which is only the remaining data that we didn't append in xso_sstream_write. So if we call quic_write_blocking with a buffer of length X, and initially append Y bytes, and write the remainig X-Y bytes via a block_until_pred call, then *written will return with the value X-Y, even though we wrote the full X bytes to the ring buffer. Fix it by recording the initial amount appended into *written, and then add the args.total_written value if we have to wait on more space Fixes openssl/project#924
nhorman
added a commit
to nhorman/openssl
that referenced
this issue
Nov 19, 2024
When writing to a blocking quic stream, we sometimes get duplicate transmitted data. This occurs when a call to quic_write_blocking has to wait for space to become available in the ring buffer. When we do a wait, the call sets *written to the value returned in args.total_written as filled out by the calls to block_until_pred->quic_write_again. However, the value there is based on the amount we requested, which is only the remaining data that we didn't append in xso_sstream_write. So if we call quic_write_blocking with a buffer of length X, and initially append Y bytes, and write the remainig X-Y bytes via a block_until_pred call, then *written will return with the value X-Y, even though we wrote the full X bytes to the ring buffer. Fix it by recording the initial amount appended into *written, and then add the args.total_written value if we have to wait on more space Fixes openssl/project#924
nhorman
added a commit
to nhorman/openssl
that referenced
this issue
Nov 21, 2024
When writing to a blocking quic stream, we sometimes get duplicate transmitted data. This occurs when a call to quic_write_blocking has to wait for space to become available in the ring buffer. When we do a wait, the call sets *written to the value returned in args.total_written as filled out by the calls to block_until_pred->quic_write_again. However, the value there is based on the amount we requested, which is only the remaining data that we didn't append in xso_sstream_write. So if we call quic_write_blocking with a buffer of length X, and initially append Y bytes, and write the remainig X-Y bytes via a block_until_pred call, then *written will return with the value X-Y, even though we wrote the full X bytes to the ring buffer. Fix it by recording the initial amount appended into *written, and then add the args.total_written value if we have to wait on more space Fixes openssl/project#924
openssl-machine
pushed a commit
to openssl/openssl
that referenced
this issue
Nov 21, 2024
When writing to a blocking quic stream, we sometimes get duplicate transmitted data. This occurs when a call to quic_write_blocking has to wait for space to become available in the ring buffer. When we do a wait, the call sets *written to the value returned in args.total_written as filled out by the calls to block_until_pred->quic_write_again. However, the value there is based on the amount we requested, which is only the remaining data that we didn't append in xso_sstream_write. So if we call quic_write_blocking with a buffer of length X, and initially append Y bytes, and write the remainig X-Y bytes via a block_until_pred call, then *written will return with the value X-Y, even though we wrote the full X bytes to the ring buffer. Fix it by recording the initial amount appended into *written, and then add the args.total_written value if we have to wait on more space Fixes openssl/project#924 Reviewed-by: Matt Caswell <matt@openssl.org> Reviewed-by: Tomas Mraz <tomas@openssl.org> (Merged from #26023) (cherry picked from commit 2de7e1d)
openssl-machine
pushed a commit
to openssl/openssl
that referenced
this issue
Nov 21, 2024
When writing to a blocking quic stream, we sometimes get duplicate transmitted data. This occurs when a call to quic_write_blocking has to wait for space to become available in the ring buffer. When we do a wait, the call sets *written to the value returned in args.total_written as filled out by the calls to block_until_pred->quic_write_again. However, the value there is based on the amount we requested, which is only the remaining data that we didn't append in xso_sstream_write. So if we call quic_write_blocking with a buffer of length X, and initially append Y bytes, and write the remainig X-Y bytes via a block_until_pred call, then *written will return with the value X-Y, even though we wrote the full X bytes to the ring buffer. Fix it by recording the initial amount appended into *written, and then add the args.total_written value if we have to wait on more space Fixes openssl/project#924 Reviewed-by: Matt Caswell <matt@openssl.org> Reviewed-by: Tomas Mraz <tomas@openssl.org> (Merged from #26023) (cherry picked from commit 2de7e1d)
openssl-machine
pushed a commit
to openssl/openssl
that referenced
this issue
Nov 21, 2024
When writing to a blocking quic stream, we sometimes get duplicate transmitted data. This occurs when a call to quic_write_blocking has to wait for space to become available in the ring buffer. When we do a wait, the call sets *written to the value returned in args.total_written as filled out by the calls to block_until_pred->quic_write_again. However, the value there is based on the amount we requested, which is only the remaining data that we didn't append in xso_sstream_write. So if we call quic_write_blocking with a buffer of length X, and initially append Y bytes, and write the remainig X-Y bytes via a block_until_pred call, then *written will return with the value X-Y, even though we wrote the full X bytes to the ring buffer. Fix it by recording the initial amount appended into *written, and then add the args.total_written value if we have to wait on more space Fixes openssl/project#924 Reviewed-by: Matt Caswell <matt@openssl.org> Reviewed-by: Tomas Mraz <tomas@openssl.org> (Merged from #26023) (cherry picked from commit 2de7e1d)
openssl-machine
pushed a commit
to openssl/openssl
that referenced
this issue
Nov 21, 2024
When writing to a blocking quic stream, we sometimes get duplicate transmitted data. This occurs when a call to quic_write_blocking has to wait for space to become available in the ring buffer. When we do a wait, the call sets *written to the value returned in args.total_written as filled out by the calls to block_until_pred->quic_write_again. However, the value there is based on the amount we requested, which is only the remaining data that we didn't append in xso_sstream_write. So if we call quic_write_blocking with a buffer of length X, and initially append Y bytes, and write the remainig X-Y bytes via a block_until_pred call, then *written will return with the value X-Y, even though we wrote the full X bytes to the ring buffer. Fix it by recording the initial amount appended into *written, and then add the args.total_written value if we have to wait on more space Fixes openssl/project#924 Reviewed-by: Matt Caswell <matt@openssl.org> Reviewed-by: Tomas Mraz <tomas@openssl.org> (Merged from #26023)
github-project-automation
bot
moved this from In Progress
to Done
in Project Board 3.5.0 planning
github-project-automation
bot
moved this from Waiting Review
to Done
in Development Board
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
In attempting to build out QUIC server testing using the quic-interop-runner, I have written a demo server here:
https://github.com/nhorman/openssl/blob/quic-server-interop/demos/guide/quic-hq-interop-server.c
When running several tests with openssl specified as a server, a number of them failed indicating:
Some debugging in the server code seems to have confirmed that we in fact wrote 3145729 bytes through the SSL_write_ex call, but the client still received 3147207 bytes.
Some additional debugging, in which I modified the interop runner harness to write sequential data into the file to download (I filled the file with a series of sequential values followed by a newline) led to an identification of the differences in the file that was sent and the file that was received:
From the diff it appears that, periodically, chunks of data within the stream get sent a second time over the wire.
Need to investigate the root cause and fix it. The error might be in the server code itself, but it doesn't appear so to me. I think the issue is a race condition in the reactor consuming bytes from the stream map with updates to the stream map from the application, but I've not looked closely enough yet
The text was updated successfully, but these errors were encountered: