summaryrefslogtreecommitdiff
path: root/src/lib (follow)
Commit message (Collapse)AuthorAgeFilesLines
...
* Assume that the size of a pointer will not change at runtime.bcook2015-01-222-56/+30
| | | | | | | Change the runtime check for whether a long is smaller than a pointer to a compile-time check. Replace the silly hash for LLP64 platforms. ok tedu@
* Fix DTLS memory leak (CVE-2015-0206).libressl-v2.1.3doug2015-01-212-18/+46
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | There were four bugs fixed by this patch: * dtls1_buffer_record() now frees rdata->rbuf.buf on error. Since s->s3->rbuf was memset, rdata->rbuf is the only pointer left which points to the old rbuf. On error, rdata is freed so there will not be any way of freeing this memory unless we do it here. * Changed the return code of dtls1_buffer_record() to differentiate between queue full (0) and error (-1). See below as this differs from upstream. * Handle errors if calls to dtls1_buffer_record() fail with -1. Previously, it did not check the return value. * Changed the way receipts are recorded. Previously, it was recorded when processed successfully (whether buffered or not) in dtls1_process_record(). Now, it records when it is handled in dtls1_get_record(): either when it is entered into the queue to buffer for the next epoch or when it is processed directly. Processing buffered records does not add a receipt because it needed one in order to get into the queue. The above bugs combined contributed to an eventual DoS through memory exhaustion. The memory leak came from dtls1_buffer_record()'s error handling. The error handling can be triggered by a duplicate record or malloc failure. It was possible to add duplicate records because they were not being dropped. The faulty receipts logic did not detect replays when dealing with records for the next epoch. Additionally, dtls1_buffer_record()'s return value was not checked so an attacker could send repeated replay records for the next epoch. Reported to OpenSSL by Chris Mueller. Patch based on OpenSSL commit 103b171d8fc282ef435f8de9afbf7782e312961f and BoringSSL commit 44e2709cd65fbd2172b9516c79e56f1875f60300. Our patch matches BoringSSL's commit. OpenSSL returns 0 when the queue is full or when malloc() or pitem_new() fails. They return -1 on error including !ssl3_setup_buffers() which is another failure to allocate memory. BoringSSL and LibreSSL changed the return code for dtls1_buffer_record() to be 1 on success, 0 when the queue is full and -1 on error. input + ok bcook@, jsing@
* Add arc4random/getentropy shims for NetBSD.bcook2015-01-194-0/+300
| | | | | | | | | | The latest NetBSD (6.1.5) arc4random does not appear to reseed the CRNG state after a fork, so provide an override until the fork-safe version in CVS appears in a release. These are the same as the FreeBSD shims. ok deraadt@
* Improve wording in alloca.tedu2015-01-171-7/+5
| | | | | | | 1. it's not a bug; it's a caveat. 2. "slightly unsafe" gives me the willies. 3. one .Xr to malloc should suffice ok deraadt jmc
* Use ">", not ">=" when comparing length to HOST_NAME_MAX sincemillert2015-01-161-3/+3
| | | | | otherwise we end up needlessly replacing a NUL with a NUL. OK deraadt@
* Replace HOST_NAME_MAX+1-1 with HOST_NAME_MAX. OK deraad@millert2015-01-162-8/+8
|
* Replace check for ">= HOST_NAME_MAX+1" with "> HOST_NAME_MAX".millert2015-01-161-1/+1
| | | | OK deraadt@
* Move to the <limits.h> universe.deraadt2015-01-1623-57/+40
| | | | review by millert, binary checking process with doug, concept with guenther
* The SSL/TLS session Id context is limited to 32 bytes. Instead ofreyk2015-01-161-1/+13
| | | | | | | | using the name of relayd relay or smtpd pki, use a 32 byte arc4random buffer that should be unique for the context. This fixes an issue in OpenSMTPD when a long pki name could break the configuration. OK gilles@ benno@
* Delete the MANLINT variable and the related SUFFIXES rules becauseschwarze2015-01-162-4/+4
| | | | | | | | since yesterday, "mandoc -Tlint -Wfatal" can no longer fail. Instead, as suggested by deraadt@, provide a manlint target that is *not* run during make build, but can be run whenever you want to check syntax of manuals. "nice stuff" deraadt@
* remove unused variablechl2015-01-151-2/+1
| | | | ok tedu@
* back in september I did the large abstraction refactoring to allow thesederaadt2015-01-1514-14/+28
| | | | other systems to fit into the same mold, so add copyright
* Make strlcpy/strlcat slightly easier to read.millert2015-01-154-86/+82
|
* For non-blocking sockets tls_connect_fds() could fail with EAGAIN.bluhm2015-01-132-7/+22
| | | | | | | | Use the same logic from the read, write, accept functions to inform the caller wether a readable or writable socket is needed. After that event, the connect function must be called again. All the checks before connecting are done only once. OK tedu@
* rename blocks to words. bcrypt "blocks" are unrelated to blowfish blocks,tedu2015-01-121-9/+9
| | | | nor are they the same size.
* Fix a memory leak in bss_dgram.doug2015-01-122-6/+26
| | | | | | | | Free data->saved_message.data. Based on OpenSSL commit: 41cd41c4416f545a18ead37e09e437c75fa07c95 except this version sets a->ptr to NULL to avoid accidental reuse and handles malloc failing. ok beck@, input + ok miod@
* missing , found by Dongsheng Songderaadt2015-01-081-3/+3
|
* Avoid a double-free in an error path.doug2015-01-082-2/+4
| | | | ok jsing@ beck@
* stupid me. need errno.htedu2015-01-071-1/+2
|
* set errno = EINVAL for invalid salts and hashes in most functions.tedu2015-01-071-12/+20
| | | | | | | remember to set EACCES in bcrypt_checkpass for hash differences. the higher level crypt_checkpass function will reset errno to EACCES in all cases, which is probably the right behavior, but this change gives code working with the lower level functions the correct errno if they care.
* mix in more virtual memory and process informationbcook2015-01-072-4/+8
|
* add initial HP-UX getentropy/arc4random support.bcook2015-01-064-0/+992
| | | | | | patch from Kinichiro Inoguchi, tested on HP-UX 11.31 ok deraadt@
* rename kern enter/exit macros to malloc enter/leave to better reflecttedu2015-01-051-7/+7
| | | | what's going on.
* convert clock() to clock_gettime() for improved precision (and accuracy?)tedu2015-01-051-7/+9
| | | | | guenther suggested using thread time, which actually may improve accuracy if somebody puts this in a threaded program.
* Fix incorrect OPENSSL_assert() usage.doug2015-01-032-48/+106
| | | | | | | | Instead of asserting, return an error code for I/O errors. This is based on OpenSSL commit 2521fcd8527008ceb3e4748f95b0ed4e2d70cfef. Added checks for two calloc()s while I'm here. ok miod@
* Rename the tls_connect_socket() parameter 'socket' to 's' to avoidbluhm2015-01-021-4/+4
| | | | | a compiler warning about shadowing a global declaration. OK jsing@
* Revert previous; tls_accept_socket() was intentionally undocumentedschwarze2015-01-011-7/+1
| | | | | because the API design isn't fully settled. Requested by jsing@ and tedu@.
* minimally document tls_accept_socket();schwarze2015-01-011-2/+8
| | | | patch from Sunil at Nimmagadda dot net
* copy bcrypt autotune from encrypt(1) and expose via crypt_newhashtedu2014-12-302-5/+43
| | | | ok deraadt miod
* don't leak timing info about padding errors by generating a fake keytedu2014-12-292-10/+20
| | | | | | | afterwards. openssl has a more complicated fix, but it's less intrusive for now to simply hoist the expensive part (fake key generation) up without sweating a branch or two. ok bcook jsing
* include netinet/in.h to define struct in6_addr.bcook2014-12-271-1/+2
| | | | Noticed while testing libtls on FreeBSD.
* simplify crypt_checkpass. The API promise is that this function doesn'ttedu2014-12-241-11/+5
| | | | | | | | | use global data. The simplest fix is to only check blowfish passwords, and implicitly lock out DES passwords. crypt_checkpass is currently only used in one place, passwd, to verify the local user's password, so this is probably acceptable. Gives people a little more time to migrate away from DES before introduing checkpass into more places.
* Add size_t to int checks for SSL functions.doug2014-12-173-8/+39
| | | | | | | | | | libtls accepts size_t for lengths but libssl accepts int. This verifies that the input does not exceed INT_MAX. It also avoids truncating size_t when comparing with int and adds printf-style attributes for tls_set_error(). with input from deraadt@ and tedu@ ok tedu@
* typo; ok deraadtsthen2014-12-162-3/+3
|
* warn for correct symbolderaadt2014-12-161-2/+2
|
* Now that we have Camellia support in libcrypto, bring in the SHA256 flavour ofmiod2014-12-164-4/+328
| | | | the Camellia ciphersuites for TLS 1.2 introduced in RFC 5932. From OpenSSL HEAD.
* Add error handling for EVP_DigestInit_ex().doug2014-12-1515-56/+98
| | | | | | | | | | | | | A few EVP_DigestInit_ex() calls were left alone since reporting an error would change the public API. Changed internal ssl3_cbc_digest_record() to return a value due to the above change. It will also now set md_out_size=0 on failure. This is based on part of BoringSSL's commit to fix malloc crashes: https://boringssl.googlesource.com/boringssl/+/69a01608f33ab6fe2c3485d94aef1fe9eacf5364 ok miod@
* unconditionally align SSL payloadsbcook2014-12-146-78/+36
| | | | | | | | Remove support for conditional payload alignment, since we would never want to turn it off. Also, consistently use size_t for calculating the alignment. ok miod@
* Convert all of the straight forward client handshake handling code to usejsing2014-12-146-212/+134
| | | | | | the new handshake functions. ok miod@
* Provide functions for starting, finishing and writing SSL handshakejsing2014-12-146-8/+120
| | | | | | | | | | | | | | | messages. This will allow for removal of repeated/duplicated code. Additionally, DTLS was written by wholesale copying of the SSL/TLS code, with some DTLS specifics being added to the duplicated code. Since these SSL handshake message functions know how to handle both SSL/TLS and DTLS, upon conversion the duplicate versions will become identical (or close to), at which point the DTLS versions can be removed and the SSL/TLS versions used for both protocols. Partially based on similar changes in OpenSSL. ok miod@
* Remove trailing whitespace.jsing2014-12-1459-559/+559
|
* unifdef OPENSSL_NO_NEXTPROTONEG, which is one of the last standing #ifndefjsing2014-12-1412-130/+12
| | | | | | | mazes in libssl. NPN is being replaced by ALPN, however it is still going to be around for a while yet. ok miod@
* unifdef NETSCAPE_HANG_BUG from the DTLS code. The code is not currentlyjsing2014-12-142-46/+2
| | | | | | | enabled and I would hope that no one is using client certificates with DTLS and Netscape, assuming it even supported it... ok bcook@ miod@
* catch up with swab.c rev. 1.9:schwarze2014-12-121-5/+23
| | | | update SYNOPSIS and DESCRIPTION and add STANDARDS
* update swab() to match the current posix definition. "rationale: none."tedu2014-12-111-50/+24
| | | | | | rewrite the function to be simpler as well. the compiler can unroll the loop for us if necessary. ok schwarze
* ssl3_init_finished_mac() calls BIO_new() which can fail since it in turnjsing2014-12-1017-42/+115
| | | | | | | | calls malloc(). Instead of silently continuing on failure, check the return value of BIO_new() and propagate failure back to the caller for appropriate handling. ok bcook@
* Remove support for GOST R 34.10-94 signature authentication, along withjsing2014-12-1018-154/+30
| | | | | | | the two ciphersuites that use it. GOST94 public/private keys have been long obsoleted and libcrypto does not have support for them anyway. Discussed with Dmitry Eremin-Solenikov.
* libssl major++jsing2014-12-102-2/+2
|
* Add support for ALPN.jsing2014-12-1010-16/+594
| | | | | | Based on OpenSSL and BoringSSL. ok bcook@
* add stdint.h to ssl.h.bcook2014-12-102-2/+6
| | | | ok jsing@
generated by cgit v1.2.3-55-g6feb (git 2.39.1) at 2025-05-21 06:33:34 +0000