summaryrefslogtreecommitdiff
path: root/src/lib (follow)
Commit message (Collapse)AuthorAgeFilesLines
...
* Treat SSL_ERROR_ZERO_RETURN as a success, rather than a failure. Alsojsing2015-04-151-5/+6
| | | | | | | ensure that outlen is set to zero so that tls_read() has read(2) like semantics for EOF. Spotted by doug@
* Make tls_close() more robust - do not rely on a close notify being receivedjsing2015-04-151-13/+17
| | | | | | | | | | | from the other side and only return TLS_READ_AGAIN/TLS_WRITE_AGAIN if we failed to send a close notify on a non-blocking socket. Otherwise be more forceful and always shutdown/close the socket regardless of other failures. Also do not consider ENOTCONN or ECONNRESET to be a shutdown failure, since there are various situations where this can occur. ok doug@ guenther@
* Another couple of commas in the wrong place, ok jmcnicm2015-04-141-3/+3
|
* Remove d2i_X509_PKEY and i2d_X509_PKEY from the SSLeay days.doug2015-04-124-88/+4
| | | | | | | | i2d_X509_PKEY is a "needs to implement" and d2i_X509_PKEY is broken. Removed upstream in commit b1f3442857c1fd76e91941141bf671d19e90a79d. ok deraadt@, jsing@
* Send OPENSSL_issetugid() straight to hell, no final cigarette.deraadt2015-04-117-34/+6
| | | | | | | | | | | | | | | | | | | The issetugid() API is supposed to make a strong promise where "0 means it is safe to look at the environment". Way back in the past someone on the OpenSSL team responded to the environment access danger by creating a wrapper called OPENSSL_issetugid, and went to use it a number of places. However, by default on systems lacking true issetugid(), OPENSSL_issetugid returns 0. 0 indicating safely. False safety. Which means OPENSSL_issetugid() fails to make any sort of promise about safety, in fact it is just the opposite. Can you believe the OpenSSL team? This nastiness was noticed over the years, however noone could gain traction and get it fixed in OpenSSL. Also see a paragraph about this in http://www.tedunangst.com/flak/post/worst-common-denominator-programming ok jsing
* Remove all getenv() calls, especially those wrapped by issetugid().deraadt2015-04-1113-115/+30
| | | | | | | | | getenv()'s wrapped by issetugid() are safe, but issetugid() is correct difficult to impliment on many operating systems. By accident, a grand experiment was run over the last year, where issetugid() returned 1 (the safe value) on a few operating systems. Noone noticed & complained that certain environment variables were not working....... ok doug beck jsing, discussion with others
* Make pthread_atfork() track the DSO that called it like atexit() does,guenther2015-04-073-2/+81
| | | | | | | | | | unregistering callbacks if the DSO is unloaded. Move the callback handling from libpthread to libc, though libpthread still overrides the inner call to handle locking and thread-library reinitialization. Major version bump for both libc and libpthread. verification that this fixes various ports ajacoutot@ asm assistance miod@; ok millert@ deraadt@
* bludgeon DES support out of crypt. long live the bcrypt.tedu2015-04-062-761/+9
|
* improve realloc. when expanding a region, actually use the free page cachetedu2015-04-061-7/+16
| | | | | instead of simply zapping it. this can save many syscalls in a program that repeatedly grows and shrinks a buffer, as observed in the wild.
* readdir() is thread-safe when DIR handles aren't shared, so delete the lock.guenther2015-04-062-6/+2
| | | | | | | (POSIX is fixing its description: readdir_r() was a botch) Patch from Carlos Mart�n Nieto (cmn (at) dwim.me) no -portable concerns bcook@
* comma fix;jmc2015-04-031-3/+3
|
* Handle the case where multiple calls to SSL_shutdown() are required tojsing2015-04-022-8/+14
| | | | | | close the connection. Also correctly handle the error on failure. Diff from cookieandscream via github.
* Correct man page title.jsing2015-04-021-2/+2
| | | | | | Diff from Tim van der Molen. ok jmc@
* Document the fact that the tls_accept_*() functions can returnjsing2015-04-021-7/+11
| | | | | | TLS_READ_AGAIN and TLS_WRITE_AGAIN. Based on a diff from Tim van der Molen.
* Bump libtls minor due to API addition.jsing2015-03-311-1/+1
|
* Provide a tls_accept_fds() function, which allows a TLS connection to bejsing2015-03-314-9/+39
| | | | | | accepted via an existing pair of file descriptors. Based on a diff from Jan Klemkow.
* Nuke the OPENSSL_MAX_TLS1_2_CIPHER_LENGTH hack - this has to be enabled atjsing2015-03-314-44/+4
| | | | | | | compile time, which we do not do and are unlikely to ever do. Additionally, there are two runtime configurable alternatives that exist. ok bcook@ doug@
* Store errors that occur during a tls_accept_socket() call on the contextjsing2015-03-314-11/+12
| | | | | | | for the server, rather than on the context for the connection. This makes more sense than the current behaviour does. Issue reported by Tim van der Molen.
* add initial AIX getentropy/arc4random files. Thanks to Michael Felt.bcook2015-03-304-0/+1002
|
* Factor out the init_buf initialisation code, rather than duplicating itjsing2015-03-2716-206/+112
| | | | | | in four different places. ok doug@ guenther@
* BUF_MEM_free() has its own explicit NULL check.jsing2015-03-272-18/+12
|
* Make rcmdsh(3) not fail if it is passed a non resolvable hostname.halex2015-03-233-20/+16
| | | | | | | Instead, silently ignore the fact and instead let the underlying ssh (or $RSH) command handle it. ok millert@
* differentiate between a failed read, returning -1, and encounteringhalex2015-03-221-2/+4
| | | | | | | end-of-file, returning 0, in order not to print an unrelated strerror(errno) in the latter case ok millert@
* sys/sysctl.h includes linux/sysctl.h, remove the extra unchecked includebcook2015-03-222-4/+2
|
* unmute rcmd hostname lookup failurehalex2015-03-221-3/+2
| | | | ok millert@ jung@
* free the server certificate in tls_connect_fds(); fixes a memory leak withsthen2015-03-211-1/+2
| | | | repeated use of tls_connect. ok jsing
* Fix indentation that incorrectly implies a different control flow.doug2015-03-216-16/+22
| | | | | | | | | | | | | The actual control flow is intentional while the indenting is incorrect. This is intended to be a cosmetic change. Verified that each of these was part of a KNF commit that wasn't intending to change behavior. Also, double checked against the history of changes in OpenSSL and BoringSSL. Addresses Coverity CIDs: 78842, 78859, 78863. ok tedu@
* Fix a memory leak in an error path.doug2015-03-202-2/+6
| | | | | | From OpenSSL commit 5e5d53d341fd9a9b9cc0a58eb3690832ca7a511f. ok guenther@, logan@
* zap #if 0'd code that's been dead since '96halex2015-03-191-7/+0
| | | | ok todd@
* Fix several crash causing defects from OpenSSL.tedu2015-03-1924-126/+324
| | | | | | | | | | | | | These include: CVE-2015-0209 - Use After Free following d2i_ECPrivatekey error CVE-2015-0286 - Segmentation fault in ASN1_TYPE_cmp CVE-2015-0287 - ASN.1 structure reuse memory corruption CVE-2015-0289 - PKCS7 NULL pointer dereferences Several other issues did not apply or were already fixed. Refer to https://www.openssl.org/news/secadv_20150319.txt joint work with beck, doug, guenther, jsing, miod
* In the neverending saga of enabling and disabling assembler code for shamiod2015-03-182-10/+2
| | | | | | | | | | | routines on hppa, the cause for sha512-parisc subtly misbehaving has been found: despite having fallback pa1.1 code when running on a 32-bit cpu, the shift constants used in the sigma computations in sha512 are >= 32 and are silently truncated to 5 bits by the assembler, so there is no chance of getting this code to work on a non-pa2.0 processor. However, the pa1.1 fallback code for sha256 is safe, as it never attempts to shift by more than 31, so reenable it again.
* Avoid a NULL pointer deref when X509_get_pubkey() returns NULL.doug2015-03-152-4/+8
| | | | | | | | | | | | | | A NULL pointer could be dereferenced when X509_REQ_set_pubkey() calls X509_PUBKEY_set() with pktmp. OpenSSL says it's the fix for CVE-2015-0288, but there aren't any public details yet to confirm. Either way, we should fix this. Based on OpenSSL commit 28a00bcd8e318da18031b2ac8778c64147cd54f9 and BoringSSL commit 9d102ddbc0f6ed835ed12272a3d8a627d6a8e728. "looks sane" beck@ ok miod@, bcook@
* remove incorrect statement. the standard *does* require calloc to checktedu2015-03-141-7/+2
| | | | | for overflow. stop talking about old broken systems, there's little use for such info.
* delay EVP_MD_CTX_init so we don't forget to clean it up.tedu2015-03-112-6/+8
| | | | spotted by miod. ok miod.
* Reject DH keys sent by a server if they are considered too small; inspiredmiod2015-03-082-2/+24
| | | | | by a similar BoringSSL change, but raising the limit to 1024 bits. ok jsing@ markus@ guenther@ deraadt@
* Do not use sha512-parisc for now, as it is subtly bugged - passes the shamiod2015-03-052-2/+6
| | | | | | | | | | | regress tests but causes tls ciphersuite using sha386 to fail; found the hard way by henning@. I can't see anything wrong in the generated assembly code yet, but building a libcrypto with no assembler code but sha512_block_data_order() is enough to trigger Henning's issue, so the bug lies there. No ABI change; ok deraadt@
* Update comment to match code; Caspar Schutijserlibressl-v2.1.4millert2015-03-021-2/+2
|
* Fix a minor information leak that was introduced in t1_lib.c r1.71, wherebyjsing2015-03-022-4/+4
| | | | | | | | | an additional 28 bytes of .rodata (or .data) is provided to the network. In most cases this is a non-issue since the memory content is already public. Issue found and reported by Felix Groebert of the Google Security Team. ok bcook@ beck@
* Prefix function parameter names with underscores in tls.h, since this makesjsing2015-02-261-42/+44
| | | | | | them guaranteed to not conflict per POSIX. ok espie@ guenther@
* Fix CVE-2014-3570: properly calculate the square of a BIGNUM value.bcook2015-02-256-1308/+606
| | | | | | | | | | | See https://www.openssl.org/news/secadv_20150108.txt for a more detailed discussion. Original OpenSSL patch here: https://github.com/openssl/openssl/commit/a7a44ba55cb4f884c6bc9ceac90072dea38e66d0 The regression test is modified a little for KNF. ok miod@
* Fix CVE-2015-0205: Do not accept client authentication with Diffie-Hellmanbcook2015-02-252-4/+4
| | | | | | | | | | certificates without requiring a CertificateVerify message. From OpenSSL commit: https://github.com/openssl/openssl/commit/1421e0c584ae9120ca1b88098f13d6d2e90b83a3 Thanks to Karthikeyan Bhargavan for reporting this. ok miod@
* we don't let strtonum errors bleed through now.tedu2015-02-241-4/+2
|
* Set errno to EINVAL, instead of letting ERANGE escape out.tedu2015-02-241-2/+4
| | | | | | Printing strerror() in that case will say result too large, even if rounds is actually too small. invalid is less specific, but less incorrect. ok millert
* fourth batch of perlpod(1) to mdoc(7) conversionschwarze2015-02-2337-1276/+2521
|
* Bump libcrypto and libssl majors, due to various recent churn.jsing2015-02-224-4/+4
| | | | Discussed with/requested by deraadt@ at the conclusion of s2k15.
* Reluctantly add server-side support for TLS_FALLBACK_SCSV.jsing2015-02-2215-25/+159
| | | | | | | | | | | | | | | | | | | | | | | | | | This allows for clients that willingly choose to perform a downgrade and attempt to establish a second connection at a lower protocol after the previous attempt unexpectedly failed, to be notified and have the second connection aborted, if the server does in fact support a higher protocol. TLS has perfectly good version negotiation and client-side fallback is dangerous. Despite this, in order to maintain maximum compatability with broken web servers, most mainstream browsers implement this. Furthermore, TLS_FALLBACK_SCSV only works if both the client and server support it and there is effectively no way to tell if this is the case, unless you control both ends. Unfortunately, various auditors and vulnerability scanners (including certain online assessment websites) consider the presence of a not yet standardised feature to be important for security, even if the clients do not perform client-side downgrade or the server only supports current TLS protocols. Diff is loosely based on OpenSSL with some inspiration from BoringSSL. Discussed with beck@ and miod@. ok bcook@
* There is not much point constructing an SSL_CIPHER, then callingjsing2015-02-222-14/+6
| | | | | ssl3_cipher_get_value() to get the cipher suite value that we just put in the struct - use the cipher suite value directly.
* Remove IMPLEMENT_STACK_OF noops.jsing2015-02-224-8/+4
|
* Bump libtls major due to symbol removal.jsing2015-02-221-3/+2
|
* Rename tls_config_insecure_noverifyhost() tojsing2015-02-224-21/+20
| | | | | | | tls_config_insecure_noverifyname(), so that it is more accurate and keeps inline with the distinction between DNS hostname and server name. Requested by tedu@ during s2k15.
generated by cgit v1.2.3-55-g6feb (git 2.39.1) at 2025-04-03 04:03:19 +0000