When reference updating ObjectSpace.trace_object_allocations, we need to
check whether the object is valid or not because it does not mark the
object so the object may be dead. This can cause a segmentation fault
if the object is on a free heap page.
For example, the following script crashes:
require "objspace"
objs = []
ObjectSpace.trace_object_allocations do
1_000_000.times do
objs << Object.new
end
end
objs = nil
# Free pages that the objs were on
GC.start
# Run compaction and check that it doesn't crash
GC.compact
We need to reinsert into the ST table when an object moves because it is
a numtable that hashes on the object address, so when an object moves we
need to reinsert it rather than just updating the key.
This change includes the following updates:
- Added an environment variable `RUBY_TCP_NO_FAST_FALLBACK` to control enabling/disabling fast_fallback
- Updated documentation and man pages
- Revised the implementation of Socket.tcp_fast_fallback= and Socket.tcp_fast_fallback, which previously performed dynamic name resolution of constants and variables. As a result, the following performance improvements were achieved:
(Case of 1000 executions of `TCPSocket.new` to the local host)
Rehearsal -----------------------------------------
before 0.031462 0.147946 0.179408 ( 0.249279)
after 0.031164 0.146839 0.178003 ( 0.346935)
-------------------------------- total: 0.178003sec
user system total real
before 0.027584 0.138712 0.166296 ( 0.233356)
after 0.025953 0.127608 0.153561 ( 0.237971)
Any memory allocated with xmalloc needs to be matched with xfree rather
than plain free.
Ruby unfortunately redefines strdup to be ruby_strdup, which uses
xmalloc so needs to be xfreed. Previously these were mismatched.
This commit changes the copy to be an explicit ruby_strdup (to avoid
confusion) and the free to be xfree.
* Use `rb_thread_fd_select` instead of select(2)
For fixing https://bugs.ruby-lang.org/issues/20932 .
`TCPSocket.new`, which internally uses select(2) for HEv2, can cause SEGV if the number of file descriptors exceeds `FD_SETSIZE`.
This change avoids that issue by replacing select(2) with `rb_thread_fd_select`, which is provided as part of Ruby's internal API.
---
This includes the following changes.
* rb_thread_fd_select does not need common pipe
We observed crashes from rb_io_bufwrite() thread switching (through
rb_thread_check_ints()) in the middle of rb_execution_context_mark(). By
the time rb_execution_context_mark() gets a timeslice again, it read
garbage from a frame that was already popped in another thread, crashing
the process in SEGV. Other mark functions probably have their own ways
of breaking, but clearly, the usual IO code do too much for this
perilous pseudo GC context.
Use `FILE*` like before 5001cc4716
("Optimize ObjectSpace.dump_all"). Also, add type checking for
the private _dump methods.
Co-authored-by: Peter Zhu <peter@peterzhu.ca>
The following two commits fix the proper clearing of the Connection Attempt Delay in `TCPSocket.new`.
- b2f610b0ed
- 6f4efaec53
The same fix will be applied to `Socket.tcp`.
Apply SSL options set in DEFAULT_PARAMS without clearing existing
options.
It currently clears options in order to avoid setting one of the
options included in OpenSSL::SSL::OP_ALL unless explicitly specified,
namely OpenSSL::SSL::OP_DONT_INSERT_EMPTY_FRAGMENTS. Now that
OpenSSL::SSL::OP_ALL has been removed from SSLContext#initialize, it is
no longer necessary.
77c3db2d65
Respect the SSL options set by default by SSL_CTX() and by the
system-wide OpenSSL configuration file.
OpenSSL::SSL::SSLContext#initialize currently adds OpenSSL::SSL::OP_ALL
on top of the default SSL options. Let's stop doing it.
OpenSSL::SSL::OP_ALL is a set of options that changes OpenSSL's behavior
to workaround various TLS implementation bugs. Using it is considered
usually safe, but is not completely harmless.
00bec0d905
get_asn1obj() is used by several methods in OpenSSL::Timestamp to get
the string representation of an OID. On an error, such as memory
allocation failure, it can raise OpenSSL::X509::AttributeError. It
should be OpenSSL::Timestamp::TimestampError instead.
a424aad1df
Internals of OpenSSL::PKCS7 should be kept within ossl_pkcs7.c.
Add a new ossl_pkcs7_new() function for duplicating and wrapping an
OpenSSL PKCS7 object in OpenSSL::PKCS7. This follows the convention
used by other ossl_*_new() functions.
b5f79f771e
Follow-up commit 0789643d73 (openssl: clear OpenSSL error
queue before return to Ruby, 2016-05-18). It should raise
OpenSSL::X509::StoreError instead of OpenSSL::X509::CertificateError.
0201f23ad6
It should raise OpenSSL::Netscape::SPKIError instead of
OpenSSL::X509::CertificateError.
No test cases covered this because it only occurs in exceptional
cases, such as memory allocation failure.
527b6101d1
Fix a copy-and-paste error introduced in commit 74f6c61756 (pkey:
allocate EVP_PKEY on #initialize, 2021-04-12).
It should raise OpenSSL::PKey::ECError instead of
OpenSSL::PKey::DSAError.
b1f6a04abf
Check the ID_callback_state ivar after SSL_read() or SSL_write()
returns, similar to what ossl_start_ssl() does.
Previously, callbacks that can raise a Ruby exception were only called
from ossl_start_ssl(). This has changed in OpenSSL 1.1.1. Particularly,
the session_new_cb will be called whenever a client receives a
NewSessionTicket message, which can happen at any time during a TLS 1.3
connection.
aac9ce1304
ssl_servername_cb() is a callback function called from OpenSSL and Ruby
exceptions must not be raised from it. Allocate the Array within
rb_protect().
3a2bf74d35
The evaluation order of C arguments is unspecified.
`RSTRING_LEN(value)` would fail if the conversion to a String by
`StringValuePtr(value)` is not done yet.
Coverity Scan found this issue.
d1e6bf323a
The evaluation order of C arguments is unspecified.
`RSTRING_LEN(str)` would fails if the conversion to a String by
`StringValuePtr` is not done yet.
Coverity Scan found this issue.
