GIT update of https://sourceware.org/git/glibc.git/release/2.42/master from glibc-2.42 to 8aaf4b732d7650c2db3beb4dc8bb70eab5b022c3 diff --git a/ADVISORIES b/ADVISORIES new file mode 100644 index 0000000000..d4e33f2df3 --- /dev/null +++ b/ADVISORIES @@ -0,0 +1,2 @@ +For the GNU C Library Security Advisories, see the git master branch: +https://sourceware.org/git/?p=glibc.git;a=tree;f=advisories;hb=HEAD diff --git a/NEWS b/NEWS index f0b0e924a4..ed3c114c7a 100644 --- a/NEWS +++ b/NEWS @@ -5,6 +5,18 @@ See the end for copying conditions. Please send GNU C library bug reports via using `glibc' in the "product" field. +Version 2.42.1 + +The following bugs were resolved with this release: + + [31943] _dl_find_object can fail if ld.so contains gaps between load segments + [32994] stdlib: resolve a double lock init issue after fork + [33234] Use TLS initial-exec model for __libc_tsd_CTYPE_* thread variables + [33245] nptl: nptl: error in internal cancellation syscall handling + [33356] nptl: creating thread stack with guardsize 0 can erroneously + conclude MADV_GUARD_INSTALL is available + [33361] nss: Group merge does not react to ERANGE during merge + Version 2.42 Major new features: diff --git a/advisories/GLIBC-SA-2023-0001 b/advisories/GLIBC-SA-2023-0001 deleted file mode 100644 index 3d19c91b6a..0000000000 --- a/advisories/GLIBC-SA-2023-0001 +++ /dev/null @@ -1,14 +0,0 @@ -printf: incorrect output for integers with thousands separator and width field - -When the printf family of functions is called with a format specifier -that uses an (enable grouping) and a minimum width -specifier, the resulting output could be larger than reasonably expected -by a caller that computed a tight bound on the buffer size. The -resulting larger than expected output could result in a buffer overflow -in the printf family of functions. - -CVE-Id: CVE-2023-25139 -Public-Date: 2023-02-02 -Vulnerable-Commit: e88b9f0e5cc50cab57a299dc7efe1a4eb385161d (2.37) -Fix-Commit: c980549cc6a1c03c23cc2fe3e7b0fe626a0364b0 (2.38) -Fix-Commit: 07b9521fc6369d000216b96562ff7c0ed32a16c4 (2.37-4) diff --git a/advisories/GLIBC-SA-2023-0002 b/advisories/GLIBC-SA-2023-0002 deleted file mode 100644 index 5122669a64..0000000000 --- a/advisories/GLIBC-SA-2023-0002 +++ /dev/null @@ -1,15 +0,0 @@ -getaddrinfo: Stack read overflow in no-aaaa mode - -If the system is configured in no-aaaa mode via /etc/resolv.conf, -getaddrinfo is called for the AF_UNSPEC address family, and a DNS -response is received over TCP that is larger than 2048 bytes, -getaddrinfo may potentially disclose stack contents via the returned -address data, or crash. - -CVE-Id: CVE-2023-4527 -Public-Date: 2023-09-12 -Vulnerable-Commit: f282cdbe7f436c75864e5640a409a10485e9abb2 (2.36) -Fix-Commit: bd77dd7e73e3530203be1c52c8a29d08270cb25d (2.39) -Fix-Commit: 4ea972b7edd7e36610e8cde18bf7a8149d7bac4f (2.36-113) -Fix-Commit: b7529346025a130fee483d42178b5c118da971bb (2.37-38) -Fix-Commit: b25508dd774b617f99419bdc3cf2ace4560cd2d6 (2.38-19) diff --git a/advisories/GLIBC-SA-2023-0003 b/advisories/GLIBC-SA-2023-0003 deleted file mode 100644 index d3aef80348..0000000000 --- a/advisories/GLIBC-SA-2023-0003 +++ /dev/null @@ -1,15 +0,0 @@ -getaddrinfo: Potential use-after-free - -When an NSS plugin only implements the _gethostbyname2_r and -_getcanonname_r callbacks, getaddrinfo could use memory that was freed -during buffer resizing, potentially causing a crash or read or write to -arbitrary memory. - -CVE-Id: CVE-2023-4806 -Public-Date: 2023-09-12 -Fix-Commit: 973fe93a5675c42798b2161c6f29c01b0e243994 (2.39) -Fix-Commit: e09ee267c03e3150c2c9ba28625ab130705a485e (2.34-420) -Fix-Commit: e3ccb230a961b4797510e6a1f5f21fd9021853e7 (2.35-270) -Fix-Commit: a9728f798ec7f05454c95637ee6581afaa9b487d (2.36-115) -Fix-Commit: 6529a7466c935f36e9006b854d6f4e1d4876f942 (2.37-39) -Fix-Commit: 00ae4f10b504bc4564e9f22f00907093f1ab9338 (2.38-20) diff --git a/advisories/GLIBC-SA-2023-0004 b/advisories/GLIBC-SA-2023-0004 deleted file mode 100644 index 5286a7aa54..0000000000 --- a/advisories/GLIBC-SA-2023-0004 +++ /dev/null @@ -1,16 +0,0 @@ -tunables: local privilege escalation through buffer overflow - -If a tunable of the form NAME=NAME=VAL is passed in the environment of a -setuid program and NAME is valid, it may result in a buffer overflow, -which could be exploited to achieve escalated privileges. This flaw was -introduced in glibc 2.34. - -CVE-Id: CVE-2023-4911 -Public-Date: 2023-10-03 -Vulnerable-Commit: 2ed18c5b534d9e92fc006202a5af0df6b72e7aca (2.34) -Fix-Commit: 1056e5b4c3f2d90ed2b4a55f96add28da2f4c8fa (2.39) -Fix-Commit: dcc367f148bc92e7f3778a125f7a416b093964d9 (2.34-423) -Fix-Commit: c84018a05aec80f5ee6f682db0da1130b0196aef (2.35-274) -Fix-Commit: 22955ad85186ee05834e47e665056148ca07699c (2.36-118) -Fix-Commit: b4e23c75aea756b4bddc4abcf27a1c6dca8b6bd3 (2.37-45) -Fix-Commit: 750a45a783906a19591fb8ff6b7841470f1f5701 (2.38-27) diff --git a/advisories/GLIBC-SA-2023-0005 b/advisories/GLIBC-SA-2023-0005 deleted file mode 100644 index cc4eb90b82..0000000000 --- a/advisories/GLIBC-SA-2023-0005 +++ /dev/null @@ -1,18 +0,0 @@ -getaddrinfo: DoS due to memory leak - -The fix for CVE-2023-4806 introduced a memory leak when an application -calls getaddrinfo for AF_INET6 with AI_CANONNAME, AI_ALL and AI_V4MAPPED -flags set. - -CVE-Id: CVE-2023-5156 -Public-Date: 2023-09-25 -Vulnerable-Commit: e09ee267c03e3150c2c9ba28625ab130705a485e (2.34-420) -Vulnerable-Commit: e3ccb230a961b4797510e6a1f5f21fd9021853e7 (2.35-270) -Vulnerable-Commit: a9728f798ec7f05454c95637ee6581afaa9b487d (2.36-115) -Vulnerable-Commit: 6529a7466c935f36e9006b854d6f4e1d4876f942 (2.37-39) -Vulnerable-Commit: 00ae4f10b504bc4564e9f22f00907093f1ab9338 (2.38-20) -Fix-Commit: 8006457ab7e1cd556b919f477348a96fe88f2e49 (2.34-421) -Fix-Commit: 17092c0311f954e6f3c010f73ce3a78c24ac279a (2.35-272) -Fix-Commit: 856bac55f98dc840e7c27cfa82262b933385de90 (2.36-116) -Fix-Commit: 4473d1b87d04b25cdd0e0354814eeaa421328268 (2.37-42) -Fix-Commit: 5ee59ca371b99984232d7584fe2b1a758b4421d3 (2.38-24) diff --git a/advisories/GLIBC-SA-2024-0001 b/advisories/GLIBC-SA-2024-0001 deleted file mode 100644 index 28931c75ae..0000000000 --- a/advisories/GLIBC-SA-2024-0001 +++ /dev/null @@ -1,15 +0,0 @@ -syslog: Heap buffer overflow in __vsyslog_internal - -__vsyslog_internal did not handle a case where printing a SYSLOG_HEADER -containing a long program name failed to update the required buffer -size, leading to the allocation and overflow of a too-small buffer on -the heap. - -CVE-Id: CVE-2023-6246 -Public-Date: 2024-01-30 -Vulnerable-Commit: 52a5be0df411ef3ff45c10c7c308cb92993d15b1 (2.37) -Fix-Commit: 6bd0e4efcc78f3c0115e5ea9739a1642807450da (2.39) -Fix-Commit: 23514c72b780f3da097ecf33a793b7ba9c2070d2 (2.38-42) -Fix-Commit: 97a4292aa4a2642e251472b878d0ec4c46a0e59a (2.37-57) -Vulnerable-Commit: b0e7888d1fa2dbd2d9e1645ec8c796abf78880b9 (2.36-16) -Fix-Commit: d1a83b6767f68b3cb5b4b4ea2617254acd040c82 (2.36-126) diff --git a/advisories/GLIBC-SA-2024-0002 b/advisories/GLIBC-SA-2024-0002 deleted file mode 100644 index 940bfcf2fc..0000000000 --- a/advisories/GLIBC-SA-2024-0002 +++ /dev/null @@ -1,15 +0,0 @@ -syslog: Heap buffer overflow in __vsyslog_internal - -__vsyslog_internal used the return value of snprintf/vsnprintf to -calculate buffer sizes for memory allocation. If these functions (for -any reason) failed and returned -1, the resulting buffer would be too -small to hold output. - -CVE-Id: CVE-2023-6779 -Public-Date: 2024-01-30 -Vulnerable-Commit: 52a5be0df411ef3ff45c10c7c308cb92993d15b1 (2.37) -Fix-Commit: 7e5a0c286da33159d47d0122007aac016f3e02cd (2.39) -Fix-Commit: d0338312aace5bbfef85e03055e1212dd0e49578 (2.38-43) -Fix-Commit: 67062eccd9a65d7fda9976a56aeaaf6c25a80214 (2.37-58) -Vulnerable-Commit: b0e7888d1fa2dbd2d9e1645ec8c796abf78880b9 (2.36-16) -Fix-Commit: 2bc9d7c002bdac38b5c2a3f11b78e309d7765b83 (2.36-127) diff --git a/advisories/GLIBC-SA-2024-0003 b/advisories/GLIBC-SA-2024-0003 deleted file mode 100644 index b43a5150ab..0000000000 --- a/advisories/GLIBC-SA-2024-0003 +++ /dev/null @@ -1,13 +0,0 @@ -syslog: Integer overflow in __vsyslog_internal - -__vsyslog_internal calculated a buffer size by adding two integers, but -did not first check if the addition would overflow. - -CVE-Id: CVE-2023-6780 -Public-Date: 2024-01-30 -Vulnerable-Commit: 52a5be0df411ef3ff45c10c7c308cb92993d15b1 (2.37) -Fix-Commit: ddf542da94caf97ff43cc2875c88749880b7259b (2.39) -Fix-Commit: d37c2b20a4787463d192b32041c3406c2bd91de0 (2.38-44) -Fix-Commit: 2b58cba076e912961ceaa5fa58588e4b10f791c0 (2.37-59) -Vulnerable-Commit: b0e7888d1fa2dbd2d9e1645ec8c796abf78880b9 (2.36-16) -Fix-Commit: b9b7d6a27aa0632f334352fa400771115b3c69b7 (2.36-128) diff --git a/advisories/GLIBC-SA-2024-0004 b/advisories/GLIBC-SA-2024-0004 deleted file mode 100644 index 08df2b3118..0000000000 --- a/advisories/GLIBC-SA-2024-0004 +++ /dev/null @@ -1,28 +0,0 @@ -ISO-2022-CN-EXT: fix out-of-bound writes when writing escape sequence - -The iconv() function in the GNU C Library versions 2.39 and older may -overflow the output buffer passed to it by up to 4 bytes when converting -strings to the ISO-2022-CN-EXT character set, which may be used to -crash an application or overwrite a neighbouring variable. - -ISO-2022-CN-EXT uses escape sequences to indicate character set changes -(as specified by RFC 1922). While the SOdesignation has the expected -bounds checks, neither SS2designation nor SS3designation have its; -allowing a write overflow of 1, 2, or 3 bytes with fixed values: -'$+I', '$+J', '$+K', '$+L', '$+M', or '$*H'. - -CVE-Id: CVE-2024-2961 -Public-Date: 2024-04-17 -Vulnerable-Commit: 755104edc75c53f4a0e7440334e944ad3c6b32fc (2.1.93-169) -Fix-Commit: f9dc609e06b1136bb0408be9605ce7973a767ada (2.40) -Fix-Commit: 31da30f23cddd36db29d5b6a1c7619361b271fb4 (2.39-31) -Fix-Commit: e1135387deded5d73924f6ca20c72a35dc8e1bda (2.38-66) -Fix-Commit: 89ce64b269a897a7780e4c73a7412016381c6ecf (2.37-89) -Fix-Commit: 4ed98540a7fd19f458287e783ae59c41e64df7b5 (2.36-164) -Fix-Commit: 36280d1ce5e245aabefb877fe4d3c6cff95dabfa (2.35-315) -Fix-Commit: a8b0561db4b9847ebfbfec20075697d5492a363c (2.34-459) -Fix-Commit: ed4f16ff6bed3037266f1fa682ebd32a18fce29c (2.33-263) -Fix-Commit: 682ad4c8623e611a971839990ceef00346289cc9 (2.32-140) -Fix-Commit: 3703c32a8d304c1ee12126134ce69be965f38000 (2.31-154) - -Reported-By: Charles Fol diff --git a/advisories/GLIBC-SA-2024-0005 b/advisories/GLIBC-SA-2024-0005 deleted file mode 100644 index a59596610a..0000000000 --- a/advisories/GLIBC-SA-2024-0005 +++ /dev/null @@ -1,22 +0,0 @@ -nscd: Stack-based buffer overflow in netgroup cache - -If the Name Service Cache Daemon's (nscd) fixed size cache is exhausted -by client requests then a subsequent client request for netgroup data -may result in a stack-based buffer overflow. This flaw was introduced -in glibc 2.15 when the cache was added to nscd. - -This vulnerability is only present in the nscd binary. - -CVE-Id: CVE-2024-33599 -Public-Date: 2024-04-23 -Vulnerable-Commit: 684ae515993269277448150a1ca70db3b94aa5bd (2.15) -Fix-Commit: 69c58d5ef9f584ea198bd00f7964d364d0e6b921 (2.31-155) -Fix-Commit: a77064893bfe8a701770e2f53a4d33805bc47a5a (2.32-141) -Fix-Commit: 5c75001a96abcd50cbdb74df24c3f013188d076e (2.33-264) -Fix-Commit: 52f73e5c4e29b14e79167272297977f360ae1e97 (2.34-460) -Fix-Commit: 7a95873543ce225376faf13bb71c43dea6d24f86 (2.35-316) -Fix-Commit: caa3151ca460bdd9330adeedd68c3112d97bffe4 (2.36-165) -Fix-Commit: f75c298e747b2b8b41b1c2f551c011a52c41bfd1 (2.37-91) -Fix-Commit: 5968aebb86164034b8f8421b4abab2f837a5bdaf (2.38-72) -Fix-Commit: 1263d583d2e28afb8be53f8d6922f0842036f35d (2.39-35) -Fix-Commit: 87801a8fd06db1d654eea3e4f7626ff476a9bdaa (2.40) diff --git a/advisories/GLIBC-SA-2024-0006 b/advisories/GLIBC-SA-2024-0006 deleted file mode 100644 index d44148d3d9..0000000000 --- a/advisories/GLIBC-SA-2024-0006 +++ /dev/null @@ -1,32 +0,0 @@ -nscd: Null pointer crash after notfound response - -If the Name Service Cache Daemon's (nscd) cache fails to add a not-found -netgroup response to the cache, the client request can result in a null -pointer dereference. This flaw was introduced in glibc 2.15 when the -cache was added to nscd. - -This vulnerability is only present in the nscd binary. - -CVE-Id: CVE-2024-33600 -Public-Date: 2024-04-24 -Vulnerable-Commit: 684ae515993269277448150a1ca70db3b94aa5bd (2.15) -Fix-Commit: b048a482f088e53144d26a61c390bed0210f49f2 (2.40) -Fix-Commit: 7835b00dbce53c3c87bbbb1754a95fb5e58187aa (2.40) -Fix-Commit: c99f886de54446cd4447db6b44be93dabbdc2f8b (2.39-37) -Fix-Commit: 5a508e0b508c8ad53bd0d2fb48fd71b242626341 (2.39-36) -Fix-Commit: 2ae9446c1b7a3064743b4a51c0bbae668ee43e4c (2.38-74) -Fix-Commit: 541ea5172aa658c4bd5c6c6d6fd13903c3d5bb0a (2.38-73) -Fix-Commit: a8070b31043c7585c36ba68a74298c4f7af075c3 (2.37-93) -Fix-Commit: 5eea50c4402e39588de98aa1d4469a79774703d4 (2.37-92) -Fix-Commit: f205b3af56740e3b014915b1bd3b162afe3407ef (2.36-167) -Fix-Commit: c34f470a615b136170abd16142da5dd0c024f7d1 (2.36-166) -Fix-Commit: bafadc589fbe21ae330e8c2af74db9da44a17660 (2.35-318) -Fix-Commit: 4370bef52b0f3f3652c6aa13d7a9bb3ac079746d (2.35-317) -Fix-Commit: 1f94122289a9bf7dba573f5d60327aaa2b85cf2e (2.34-462) -Fix-Commit: 966d6ac9e40222b84bb21674cc4f83c8d72a5a26 (2.34-461) -Fix-Commit: e3eef1b8fbdd3a7917af466ca9c4b7477251ca79 (2.33-266) -Fix-Commit: f20a8d696b13c6261b52a6434899121f8b19d5a7 (2.33-265) -Fix-Commit: be602180146de37582a3da3a0caa4b719645de9c (2.32-143) -Fix-Commit: 394eae338199078b7961b051c191539870742d7b (2.32-142) -Fix-Commit: 8d7949183760170c61e55def723c1d8050187874 (2.31-157) -Fix-Commit: 304ce5fe466c4762b21b36c26926a4657b59b53e (2.31-156) diff --git a/advisories/GLIBC-SA-2024-0007 b/advisories/GLIBC-SA-2024-0007 deleted file mode 100644 index b6928fa27a..0000000000 --- a/advisories/GLIBC-SA-2024-0007 +++ /dev/null @@ -1,28 +0,0 @@ -nscd: netgroup cache may terminate daemon on memory allocation failure - -The Name Service Cache Daemon's (nscd) netgroup cache uses xmalloc or -xrealloc and these functions may terminate the process due to a memory -allocation failure resulting in a denial of service to the clients. The -flaw was introduced in glibc 2.15 when the cache was added to nscd. - -This vulnerability is only present in the nscd binary. - -Subsequent refactoring of the netgroup cache only added more uses of -xmalloc and xrealloc. Uses of xmalloc and xrealloc in other parts of -nscd only occur during startup of the daemon and so are not affected by -client requests that could trigger an out of memory followed by -termination. - -CVE-Id: CVE-2024-33601 -Public-Date: 2024-04-24 -Vulnerable-Commit: 684ae515993269277448150a1ca70db3b94aa5bd (2.15) -Fix-Commit: c04a21e050d64a1193a6daab872bca2528bda44b (2.40) -Fix-Commit: a9a8d3eebb145779a18d90e3966009a1daa63cd8 (2.39-38) -Fix-Commit: 71af8ca864345d39b746d5cee84b94b430fad5db (2.38-75) -Fix-Commit: 6e106dc214d6a033a4e945d1c6cf58061f1c5f1f (2.37-94) -Fix-Commit: b6742463694b1dfdd5120b91ee21cf05d15ec2e2 (2.36-168) -Fix-Commit: 7a5864cac60e06000394128a5a2817b03542f5a3 (2.35-319) -Fix-Commit: 86f1d5f4129c373ac6fb6df5bcf38273838843cb (2.34-463) -Fix-Commit: 4d27d4b9a188786fc6a56745506cec2acfc51f83 (2.33-267) -Fix-Commit: 3ed195a8ec89da281e3c4bf887a13d281b72d8f4 (2.32-144) -Fix-Commit: bbf5a58ccb55679217f94de706164d15372fbbc0 (2.31-158) diff --git a/advisories/GLIBC-SA-2024-0008 b/advisories/GLIBC-SA-2024-0008 deleted file mode 100644 index d93e2a6f0b..0000000000 --- a/advisories/GLIBC-SA-2024-0008 +++ /dev/null @@ -1,26 +0,0 @@ -nscd: netgroup cache assumes NSS callback uses in-buffer strings - -The Name Service Cache Daemon's (nscd) netgroup cache can corrupt memory -when the NSS callback does not store all strings in the provided buffer. -The flaw was introduced in glibc 2.15 when the cache was added to nscd. - -This vulnerability is only present in the nscd binary. - -There is no guarantee from the NSS callback API that the returned -strings are all within the buffer. However, the netgroup cache code -assumes that the NSS callback uses in-buffer strings and if it doesn't -the buffer resizing logic could lead to potential memory corruption. - -CVE-Id: CVE-2024-33602 -Public-Date: 2024-04-24 -Vulnerable-Commit: 684ae515993269277448150a1ca70db3b94aa5bd (2.15) -Fix-Commit: c04a21e050d64a1193a6daab872bca2528bda44b (2.40) -Fix-Commit: a9a8d3eebb145779a18d90e3966009a1daa63cd8 (2.39-38) -Fix-Commit: 71af8ca864345d39b746d5cee84b94b430fad5db (2.38-75) -Fix-Commit: 6e106dc214d6a033a4e945d1c6cf58061f1c5f1f (2.37-94) -Fix-Commit: b6742463694b1dfdd5120b91ee21cf05d15ec2e2 (2.36-168) -Fix-Commit: 7a5864cac60e06000394128a5a2817b03542f5a3 (2.35-319) -Fix-Commit: 86f1d5f4129c373ac6fb6df5bcf38273838843cb (2.34-463) -Fix-Commit: 4d27d4b9a188786fc6a56745506cec2acfc51f83 (2.33-267) -Fix-Commit: 3ed195a8ec89da281e3c4bf887a13d281b72d8f4 (2.32-144) -Fix-Commit: bbf5a58ccb55679217f94de706164d15372fbbc0 (2.31-158) diff --git a/advisories/GLIBC-SA-2025-0001 b/advisories/GLIBC-SA-2025-0001 deleted file mode 100644 index b053d32e91..0000000000 --- a/advisories/GLIBC-SA-2025-0001 +++ /dev/null @@ -1,40 +0,0 @@ -assert: Buffer overflow when printing assertion failure message - -When the assert() function fails, it does not allocate enough space for the -assertion failure message string and size information, which may lead to a -buffer overflow if the message string size aligns to page size. - -This bug can be triggered when an assertion in a program fails. The assertion -failure message is allocated to allow developers to see this failure in core -dumps and it typically includes, in addition to the invariant assertion -string and function name, the name of the program. If the name of the failing -program is user controlled, for example on a local system, this could allow an -attacker to control the assertion failure to trigger this buffer overflow. - -The only viable vector for exploitation of this bug is local, if a setuid -program exists that has an existing bug that results in an assertion failure. -No such program has been discovered at the time of publishing this advisory, -but the presence of custom setuid programs, although strongly discouraged as a -security practice, cannot be discounted. - -CVE-Id: CVE-2025-0395 -Public-Date: 2025-01-22 -Vulnerable-Commit: f8a3b5bf8fa1d0c43d2458e03cc109a04fdef194 (2.13-175) -Fix-Commit: 68ee0f704cb81e9ad0a78c644a83e1e9cd2ee578 (2.41) -Fix-Commit: cdb9ba84191ce72e86346fb8b1d906e7cd930ea2 (2.42) -Fix-Commit: 69fda28279b497bd405fdd442a6d8e4d3d5f681b (2.41-7) -Fix-Commit: 7d4b6bcae91f29d7b4daf15bab06b66cf1d2217c (2.40-66) -Fix-Commit: d6c156c326999f144cb5b73d29982108d549ad8a (2.40-71) -Fix-Commit: 808a84a8b81468b517a4d721fdc62069cb8c211f (2.39-146) -Fix-Commit: f6d48470aef9264d2d56f4c4533eb76db7f9c2e4 (2.39-150) -Fix-Commit: c32fd59314c343db88c3ea4a203870481d33c3d2 (2.38-122) -Fix-Commit: f984e2d7e8299726891a1a497a3c36cd5542a0bf (2.38-124) -Fix-Commit: a3d7865b098a3a67c44f7812208d9ce4718873ba (2.37-143) -Fix-Commit: b989519fe1683c204ac24ec92830e3fe3bfaccad (2.37-146) -Fix-Commit: 7971add7ee4171fdd8dfd17e7c04c4ed77a18845 (2.36-216) -Fix-Commit: 0487893d5c5bc6710d83d7c3152d888a0339559e (2.36-219) -Fix-Commit: 8b5d4be762419c4f6176261c6fea40ac559b88dc (2.35-370) -Fix-Commit: 8b3d09dc0d350191985f9d291cc30ce96f034b49 (2.35-373) -Fix-Commit: df4e1f4a5096b385c9bcc94424cf2eaa227b3761 (2.34-500) -Fix-Commit: 31eb872cb21449832ab47ad5db83281d240e1d03 (2.34-503) -Reported-By: Qualys Security Advisory diff --git a/advisories/GLIBC-SA-2025-0002 b/advisories/GLIBC-SA-2025-0002 deleted file mode 100644 index 161da13dd4..0000000000 --- a/advisories/GLIBC-SA-2025-0002 +++ /dev/null @@ -1,23 +0,0 @@ -elf: static setuid binary dlopen may incorrectly search LD_LIBRARY_PATH - -A statically linked setuid binary that calls dlopen (including internal -dlopen calls after setlocale or calls to NSS functions such as getaddrinfo) -may incorrectly search LD_LIBRARY_PATH to determine which library to load, -leading to the execution of library code that is attacker controlled. - -The only viable vector for exploitation of this bug is local, if a static -setuid program exists, and that program calls dlopen, then it may search -LD_LIBRARY_PATH to locate the SONAME to load. No such program has been -discovered at the time of publishing this advisory, but the presence of -custom setuid programs, although strongly discouraged as a security -practice, cannot be discounted. - -CVE-Id: CVE-2025-4802 -Public-Date: 2025-05-16 -Vulnerable-Commit: 10e93d968716ab82931d593bada121c17c0a4b93 (2.27) -Fix-Commit: 5451fa962cd0a90a0e2ec1d8910a559ace02bba0 (2.39) -Fix-Commit: 3be3728df2f1912c80abd3288bc6e3a25ad679e4 (2.38-132) -Fix-Commit: 7403ede2d7752e59e0c47d5d33d73c2bf850e7be (2.37-154) -Fix-Commit: 2ef7850279b2931caf6d6d6743ebaa91839e1cf7 (2.36-227) -Fix-Commit: 621c65ccf12ddd415ceeb2234423bd1acd0fabb3 (2.35-387) -Fix-Commit: 35018c0fd20eac9ceaf60060fed2745b3177359d (2.34-517) diff --git a/advisories/GLIBC-SA-2025-0003 b/advisories/GLIBC-SA-2025-0003 deleted file mode 100644 index 2adeb3ce00..0000000000 --- a/advisories/GLIBC-SA-2025-0003 +++ /dev/null @@ -1,30 +0,0 @@ -power10: strcmp fails to save and restore nonvolatile vector registers - -The Power 10 implementation of strcmp in -sysdeps/powerpc/powerpc64/le/power10/strcmp.S failed to save/restore -nonvolatile vector registers in the 32-byte aligned loop path. This -results in callers reading content from those registers in a different -context, potentially altering program logic. - -There could be a program context where a user controlled string could -leak through strcmp into program code, thus altering its logic. There -is also a potential for sensitive strings passed into strcmp leaking -through the clobbered registers into parts of the calling program that -should otherwise not have had access to those strings. - -The impact of this flaw is limited to applications running on Power 10 -hardware that use the nonvolatile vector registers, i.e. v20 to v31 -assuming that they have been treated in accordance with the OpenPower -psABI. It is possible to work around the issue for those specific -applications by setting the glibc.cpu.hwcaps tunable to "-arch_3_1" like -so: - - export GLIBC_TUNABLES=glibc.cpu.hwcaps=-arch_3_1 - -CVE-Id: CVE-2025-5702 -Public-Date: 2025-06-04 -Vulnerable-Commit: 3367d8e180848030d1646f088759f02b8dfe0d6f (2.39) -Fix-Commit: 15808c77b35319e67ee0dc8f984a9a1a434701bc (2.42) -Fix-Commit: 0c76c951620f9e12df2a89b2c684878b55bb6795 (2.41-60) -Fix-Commit: 7e12550b8e3a11764a4a9090ce6bd3fc23fc8a8e (2.40-139) -Fix-Commit: 06a70769fd0b2e1f2a3085ad50ab620282bd77b3 (2.39-209) diff --git a/advisories/GLIBC-SA-2025-0004 b/advisories/GLIBC-SA-2025-0004 deleted file mode 100644 index 9409ca27c4..0000000000 --- a/advisories/GLIBC-SA-2025-0004 +++ /dev/null @@ -1,29 +0,0 @@ -power10: strncmp fails to save and restore nonvolatile vector registers - -The Power 10 implementation of strncmp in -sysdeps/powerpc/powerpc64/le/power10/strncmp.S failed to save/restore -nonvolatile vector registers in the 32-byte aligned loop path. This -results in callers reading content from those registers in a different -context, potentially altering program logic. - -There could be a program context where a user controlled string could -leak through strncmp into program code, thus altering its logic. There -is also a potential for sensitive strings passed into strncmp leaking -through the clobbered registers into parts of the calling program that -should otherwise not have had access to those strings. - -The impact of this flaw is limited to applications running on Power 10 -hardware that use the nonvolatile vector registers, i.e. v20 to v31 -assuming that they have been treated in accordance with the OpenPower -psABI. It is possible to work around the issue for those specific -applications by setting the glibc.cpu.hwcaps tunable to "-arch_3_1" like -so: - - export GLIBC_TUNABLES=glibc.cpu.hwcaps=-arch_3_1 - -CVE-Id: CVE-2025-5745 -Public-Date: 2025-06-05 -Vulnerable-Commit: 23f0d81608d0ca6379894ef81670cf30af7fd081 (2.40) -Fix-Commit: 63c60101ce7c5eac42be90f698ba02099b41b965 (2.42) -Fix-Commit: 84bdbf8a6f2fdafd3661489dbb7f79835a52da82 (2.41-57) -Fix-Commit: 42a5a940c974d02540c8da26d6374c744d148cb9 (2.40-136) diff --git a/advisories/GLIBC-SA-2025-0005 b/advisories/GLIBC-SA-2025-0005 deleted file mode 100644 index 8bcccc59a5..0000000000 --- a/advisories/GLIBC-SA-2025-0005 +++ /dev/null @@ -1,14 +0,0 @@ -posix: Fix double-free after allocation failure in regcomp - -The regcomp function in the GNU C library version from 2.4 to 2.41 is -subject to a double free if some previous allocation fails. It can be -accomplished either by a malloc failure or by using an interposed -malloc that injects random malloc failures. The double free can allow -buffer manipulation depending of how the regex is constructed. -This issue affects all architectures and ABIs supported by the GNU C -library. - -CVE-Id: CVE-2025-8058 -Public-Date: 2025-07-22 -Vulnerable-Commit: 963d8d782fc98fb6dc3a66f0068795f9920c269d (2.3.3-1596) -Fix-Commit: 7ea06e994093fa0bcca0d0ee2c1db271d8d7885d (2.42) diff --git a/advisories/README b/advisories/README deleted file mode 100644 index b8f8a829ca..0000000000 --- a/advisories/README +++ /dev/null @@ -1,77 +0,0 @@ -GNU C Library Security Advisory Format -====================================== - -Security advisories in this directory follow a simple git commit log -format, with a heading and free-format description augmented with tags -to allow parsing key information. References to code changes are -specific to the glibc repository and follow a specific format: - - Tag-name: (release-version) - -The indicates a specific commit in the repository. The -release-version indicates the publicly consumable release in which this -commit is known to exist. The release-version is derived from the -git-describe format, (i.e. stripped out from glibc-2.34.NNN-gxxxx) and -is of the form 2.34-NNN. If the -NNN suffix is absent, it means that -the change is in that release tarball, otherwise the change is on the -release/2.YY/master branch and not in any released tarball. - -The following tags are currently being used: - -CVE-Id: -This is the CVE-Id assigned under the CVE Program -(https://www.cve.org/). - -Public-Date: -The date this issue became publicly known. - -Vulnerable-Commit: -The commit that introduced this vulnerability. There could be multiple -entries, one for each release branch in the glibc repository; the -release-version portion of this tag should tell you which branch this is -on. - -Fix-Commit: -The commit that fixed this vulnerability. There could be multiple -entries for each release branch in the glibc repository, indicating that -all of those commits contributed to fixing that issue in each of those -branches. - -Reported-By: -The entity that reported this issue. There could be multiple entries, one for -each reporter. - -Adding an Advisory ------------------- - -An advisory for a CVE needs to be added on the master branch in two steps: - -1. Add the text of the advisory without any Fix-Commit tags along with - the fix for the CVE. Add the Vulnerable-Commit tag, if applicable. - The advisories directory does not exist in release branches, so keep - the advisory text commit distinct from the code changes, to ease - backports. Ask for the GLIBC-SA advisory number from the security - team. - -2. Finish all backports on release branches and then back on the msater - branch, add all commit refs to the advisory using the Fix-Commit - tags. Don't bother adding the release-version subscript since the - next step will overwrite it. - -3. Run the process-advisories.sh script in the scripts directory on the - advisory: - - scripts/process-advisories.sh update GLIBC-SA-YYYY-NNNN - - (replace YYYY-NNNN with the actual advisory number). - -4. Verify the updated advisory and push the result. - -Getting a NEWS snippet from advisories --------------------------------------- - -Run: - - scripts/process-advisories.sh news - -and copy the content into the NEWS file. diff --git a/ctype/ctype-info.c b/ctype/ctype-info.c index b7d3422726..fb5acf9419 100644 --- a/ctype/ctype-info.c +++ b/ctype/ctype-info.c @@ -24,11 +24,11 @@ __ctype_init before user code runs, but this does not happen for threads in secondary namespaces. With the initializers, secondary namespaces at least get locale data from the C locale. */ -__thread const uint16_t * __libc_tsd_CTYPE_B +__thread const uint16_t * __libc_tsd_CTYPE_B attribute_tls_model_ie = (const uint16_t *) _nl_C_LC_CTYPE_class + 128; -__thread const int32_t * __libc_tsd_CTYPE_TOLOWER +__thread const int32_t * __libc_tsd_CTYPE_TOLOWER attribute_tls_model_ie = (const int32_t *) _nl_C_LC_CTYPE_tolower + 128; -__thread const int32_t * __libc_tsd_CTYPE_TOUPPER +__thread const int32_t * __libc_tsd_CTYPE_TOUPPER attribute_tls_model_ie = (const int32_t *) _nl_C_LC_CTYPE_toupper + 128; diff --git a/elf/Makefile b/elf/Makefile index 48aa0b57e5..3a5596e2bb 100644 --- a/elf/Makefile +++ b/elf/Makefile @@ -543,6 +543,8 @@ tests-internal += \ tst-dl_find_object-threads \ tst-dlmopen2 \ tst-hash-collision3 \ + tst-link-map-contiguous-ldso \ + tst-link-map-contiguous-libc \ tst-ptrguard1 \ tst-stackguard1 \ tst-tls-surplus \ @@ -554,6 +556,10 @@ tests-internal += \ unload2 \ # tests-internal +ifeq ($(build-hardcoded-path-in-tests),yes) +tests-internal += tst-link-map-contiguous-main +endif + tests-container += \ tst-dlopen-self-container \ tst-dlopen-tlsmodid-container \ diff --git a/elf/dl-find_object.c b/elf/dl-find_object.c index 1e76373292..c9f4c1c8d1 100644 --- a/elf/dl-find_object.c +++ b/elf/dl-find_object.c @@ -465,6 +465,37 @@ _dl_find_object (void *pc1, struct dl_find_object *result) } rtld_hidden_def (_dl_find_object) +/* Subroutine of _dlfo_process_initial to split out noncontigous link + maps. NODELETE is the number of used _dlfo_nodelete_mappings + elements. It is incremented as needed, and the new NODELETE value + is returned. */ +static size_t +_dlfo_process_initial_noncontiguous_map (struct link_map *map, + size_t nodelete) +{ + struct dl_find_object_internal dlfo; + _dl_find_object_from_map (map, &dlfo); + + /* PT_LOAD segments for a non-contiguous link map are added to the + non-closeable mappings. */ + const ElfW(Phdr) *ph = map->l_phdr; + const ElfW(Phdr) *ph_end = map->l_phdr + map->l_phnum; + for (; ph < ph_end; ++ph) + if (ph->p_type == PT_LOAD) + { + if (_dlfo_nodelete_mappings != NULL) + { + /* Second pass only. */ + _dlfo_nodelete_mappings[nodelete] = dlfo; + ElfW(Addr) start = ph->p_vaddr + map->l_addr; + _dlfo_nodelete_mappings[nodelete].map_start = start; + _dlfo_nodelete_mappings[nodelete].map_end = start + ph->p_memsz; + } + ++nodelete; + } + return nodelete; +} + /* _dlfo_process_initial is called twice. First to compute the array sizes from the initial loaded mappings. Second to fill in the bases and infos arrays with the (still unsorted) data. Returns the @@ -476,29 +507,8 @@ _dlfo_process_initial (void) size_t nodelete = 0; if (!main_map->l_contiguous) - { - struct dl_find_object_internal dlfo; - _dl_find_object_from_map (main_map, &dlfo); - - /* PT_LOAD segments for a non-contiguous are added to the - non-closeable mappings. */ - for (const ElfW(Phdr) *ph = main_map->l_phdr, - *ph_end = main_map->l_phdr + main_map->l_phnum; - ph < ph_end; ++ph) - if (ph->p_type == PT_LOAD) - { - if (_dlfo_nodelete_mappings != NULL) - { - /* Second pass only. */ - _dlfo_nodelete_mappings[nodelete] = dlfo; - _dlfo_nodelete_mappings[nodelete].map_start - = ph->p_vaddr + main_map->l_addr; - _dlfo_nodelete_mappings[nodelete].map_end - = _dlfo_nodelete_mappings[nodelete].map_start + ph->p_memsz; - } - ++nodelete; - } - } + /* Contiguous case already handled in _dl_find_object_init. */ + nodelete = _dlfo_process_initial_noncontiguous_map (main_map, nodelete); size_t loaded = 0; for (Lmid_t ns = 0; ns < GL(dl_nns); ++ns) @@ -510,11 +520,18 @@ _dlfo_process_initial (void) /* lt_library link maps are implicitly NODELETE. */ if (l->l_type == lt_library || l->l_nodelete_active) { - if (_dlfo_nodelete_mappings != NULL) - /* Second pass only. */ - _dl_find_object_from_map - (l, _dlfo_nodelete_mappings + nodelete); - ++nodelete; + /* The kernel may have loaded ld.so with gaps. */ + if (!l->l_contiguous && is_rtld_link_map (l)) + nodelete + = _dlfo_process_initial_noncontiguous_map (l, nodelete); + else + { + if (_dlfo_nodelete_mappings != NULL) + /* Second pass only. */ + _dl_find_object_from_map + (l, _dlfo_nodelete_mappings + nodelete); + ++nodelete; + } } else if (l->l_type == lt_loaded) { @@ -764,7 +781,6 @@ _dl_find_object_update_1 (struct link_map **loaded, size_t count) /* Prefer newly loaded link map. */ assert (loaded_index1 > 0); _dl_find_object_from_map (loaded[loaded_index1 - 1], dlfo); - loaded[loaded_index1 - 1]->l_find_object_processed = 1; --loaded_index1; } diff --git a/elf/dl-find_object.h b/elf/dl-find_object.h index 9aa2439eaa..d9d75c4ad9 100644 --- a/elf/dl-find_object.h +++ b/elf/dl-find_object.h @@ -94,7 +94,7 @@ _dl_find_object_to_external (struct dl_find_object_internal *internal, } /* Extract the object location data from a link map and writes it to - *RESULT using relaxed MO stores. */ + *RESULT using relaxed MO stores. Set L->l_find_object_processed. */ static void __attribute__ ((unused)) _dl_find_object_from_map (struct link_map *l, struct dl_find_object_internal *result) @@ -141,8 +141,11 @@ _dl_find_object_from_map (struct link_map *l, break; } if (read_seg == 3) - return; + goto done; } + + done: + l->l_find_object_processed = 1; } /* Called by the dynamic linker to set up the data structures for the diff --git a/elf/rtld.c b/elf/rtld.c index 493f9696ea..ef4d96c053 100644 --- a/elf/rtld.c +++ b/elf/rtld.c @@ -1239,6 +1239,60 @@ rtld_setup_main_map (struct link_map *main_map) return has_interp; } +/* Set up the program header information for the dynamic linker + itself. It can be accessed via _r_debug and dl_iterate_phdr + callbacks, and it is used by _dl_find_object. */ +static void +rtld_setup_phdr (void) +{ + /* Starting from binutils-2.23, the linker will define the magic + symbol __ehdr_start to point to our own ELF header if it is + visible in a segment that also includes the phdrs. */ + + const ElfW(Ehdr) *rtld_ehdr = &__ehdr_start; + assert (rtld_ehdr->e_ehsize == sizeof *rtld_ehdr); + assert (rtld_ehdr->e_phentsize == sizeof (ElfW(Phdr))); + + const ElfW(Phdr) *rtld_phdr = (const void *) rtld_ehdr + rtld_ehdr->e_phoff; + + _dl_rtld_map.l_phdr = rtld_phdr; + _dl_rtld_map.l_phnum = rtld_ehdr->e_phnum; + + + _dl_rtld_map.l_contiguous = 1; + /* The linker may not have produced a contiguous object. The kernel + will load the object with actual gaps (unlike the glibc loader + for shared objects, which always produces a contiguous mapping). + See similar logic in rtld_setup_main_map above. */ + { + ElfW(Addr) expected_load_address = 0; + for (const ElfW(Phdr) *ph = rtld_phdr; ph < &rtld_phdr[rtld_ehdr->e_phnum]; + ++ph) + if (ph->p_type == PT_LOAD) + { + ElfW(Addr) mapstart = ph->p_vaddr & ~(GLRO(dl_pagesize) - 1); + if (_dl_rtld_map.l_contiguous && expected_load_address != 0 + && expected_load_address != mapstart) + _dl_rtld_map.l_contiguous = 0; + ElfW(Addr) allocend = ph->p_vaddr + ph->p_memsz; + /* The next expected address is the page following this load + segment. */ + expected_load_address = ((allocend + GLRO(dl_pagesize) - 1) + & ~(GLRO(dl_pagesize) - 1)); + } + } + + /* PT_GNU_RELRO is usually the last phdr. */ + size_t cnt = rtld_ehdr->e_phnum; + while (cnt-- > 0) + if (rtld_phdr[cnt].p_type == PT_GNU_RELRO) + { + _dl_rtld_map.l_relro_addr = rtld_phdr[cnt].p_vaddr; + _dl_rtld_map.l_relro_size = rtld_phdr[cnt].p_memsz; + break; + } +} + /* Adjusts the contents of the stack and related globals for the user entry point. The ld.so processed skip_args arguments and bumped _dl_argv and _dl_argc accordingly. Those arguments are removed from @@ -1705,33 +1759,7 @@ dl_main (const ElfW(Phdr) *phdr, ++GL(dl_ns)[LM_ID_BASE]._ns_nloaded; ++GL(dl_load_adds); - /* Starting from binutils-2.23, the linker will define the magic symbol - __ehdr_start to point to our own ELF header if it is visible in a - segment that also includes the phdrs. If that's not available, we use - the old method that assumes the beginning of the file is part of the - lowest-addressed PT_LOAD segment. */ - - /* Set up the program header information for the dynamic linker - itself. It is needed in the dl_iterate_phdr callbacks. */ - const ElfW(Ehdr) *rtld_ehdr = &__ehdr_start; - assert (rtld_ehdr->e_ehsize == sizeof *rtld_ehdr); - assert (rtld_ehdr->e_phentsize == sizeof (ElfW(Phdr))); - - const ElfW(Phdr) *rtld_phdr = (const void *) rtld_ehdr + rtld_ehdr->e_phoff; - - _dl_rtld_map.l_phdr = rtld_phdr; - _dl_rtld_map.l_phnum = rtld_ehdr->e_phnum; - - - /* PT_GNU_RELRO is usually the last phdr. */ - size_t cnt = rtld_ehdr->e_phnum; - while (cnt-- > 0) - if (rtld_phdr[cnt].p_type == PT_GNU_RELRO) - { - _dl_rtld_map.l_relro_addr = rtld_phdr[cnt].p_vaddr; - _dl_rtld_map.l_relro_size = rtld_phdr[cnt].p_memsz; - break; - } + rtld_setup_phdr (); /* Add the dynamic linker to the TLS list if it also uses TLS. */ if (_dl_rtld_map.l_tls_blocksize != 0) diff --git a/elf/tst-env-setuid.c b/elf/tst-env-setuid.c index 7209acd616..ff3eda7f91 100644 --- a/elf/tst-env-setuid.c +++ b/elf/tst-env-setuid.c @@ -40,6 +40,8 @@ static char SETGID_CHILD[] = "setgid-child"; # define PROFILE_LIB "tst-sonamemove-runmod2.so" #endif +#define LD_DEBUG_OUTPUT "/tmp/some-file" + struct envvar_t { const char *env; @@ -61,7 +63,7 @@ static const struct envvar_t filtered_envvars[] = { "MALLOC_TRIM_THRESHOLD_", FILTERED_VALUE }, { "RES_OPTIONS", FILTERED_VALUE }, { "LD_DEBUG", "all" }, - { "LD_DEBUG_OUTPUT", "/tmp/some-file" }, + { "LD_DEBUG_OUTPUT", LD_DEBUG_OUTPUT }, { "LD_WARN", FILTERED_VALUE }, { "LD_VERBOSE", FILTERED_VALUE }, { "LD_BIND_NOW", "0" }, @@ -74,6 +76,14 @@ static const struct envvar_t unfiltered_envvars[] = { "LD_ASSUME_KERNEL", UNFILTERED_VALUE }, }; +static void +unlink_ld_debug_output (pid_t pid) +{ + char *output = xasprintf ("%s.%d", LD_DEBUG_OUTPUT, pid); + unlink (output); + free (output); +} + static int test_child (void) { @@ -138,13 +148,21 @@ do_test (int argc, char **argv) /* Setgid child process. */ if (argc == 2 && strcmp (argv[1], SETGID_CHILD) == 0) { + pid_t ppid = getppid (); + if (getgid () == getegid ()) - /* This can happen if the file system is mounted nosuid. */ - FAIL_UNSUPPORTED ("SGID failed: GID and EGID match (%jd)\n", - (intmax_t) getgid ()); + { + /* This can happen if the file system is mounted nosuid. */ + unlink_ld_debug_output (ppid); + + FAIL_UNSUPPORTED ("SGID failed: GID and EGID match (%jd)\n", + (intmax_t) getgid ()); + } int ret = test_child (); + unlink_ld_debug_output (ppid); + if (ret != 0) exit (1); return 0; diff --git a/elf/tst-link-map-contiguous-ldso.c b/elf/tst-link-map-contiguous-ldso.c new file mode 100644 index 0000000000..04de808bb2 --- /dev/null +++ b/elf/tst-link-map-contiguous-ldso.c @@ -0,0 +1,98 @@ +/* Check that _dl_find_object behavior matches up with gaps. + Copyright (C) 2025 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + . */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +static int +do_test (void) +{ + struct link_map *l = xdlopen (LD_SO, RTLD_NOW); + if (!l->l_contiguous) + { + puts ("info: ld.so link map is not contiguous"); + + /* Try to find holes by probing with mmap. */ + int pagesize = getpagesize (); + bool gap_found = false; + ElfW(Addr) addr = l->l_map_start; + TEST_COMPARE (addr % pagesize, 0); + while (addr < l->l_map_end) + { + void *expected = (void *) addr; + void *ptr = xmmap (expected, 1, PROT_READ | PROT_WRITE, + MAP_PRIVATE | MAP_ANONYMOUS, -1); + struct dl_find_object dlfo; + int dlfo_ret = _dl_find_object (expected, &dlfo); + if (ptr == expected) + { + if (dlfo_ret < 0) + { + TEST_COMPARE (dlfo_ret, -1); + printf ("info: hole without mapping data found at %p\n", ptr); + } + else + FAIL ("object \"%s\" found in gap at %p", + dlfo.dlfo_link_map->l_name, ptr); + gap_found = true; + } + else if (dlfo_ret == 0) + { + if ((void *) dlfo.dlfo_link_map != (void *) l) + { + printf ("info: object \"%s\" found at %p\n", + dlfo.dlfo_link_map->l_name, ptr); + gap_found = true; + } + } + else + TEST_COMPARE (dlfo_ret, -1); + xmunmap (ptr, 1); + addr += pagesize; + } + if (!gap_found) + FAIL ("no ld.so gap found"); + } + else + { + puts ("info: ld.so link map is contiguous"); + + /* Assert that ld.so is truly contiguous in memory. */ + volatile long int *p = (volatile long int *) l->l_map_start; + volatile long int *end = (volatile long int *) l->l_map_end; + while (p < end) + { + *p; + ++p; + } + } + + xdlclose (l); + + return 0; +} + +#include diff --git a/elf/tst-link-map-contiguous-libc.c b/elf/tst-link-map-contiguous-libc.c new file mode 100644 index 0000000000..eb5728c765 --- /dev/null +++ b/elf/tst-link-map-contiguous-libc.c @@ -0,0 +1,57 @@ +/* Check that the entire libc.so program image is readable if contiguous. + Copyright (C) 2025 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + . */ + +#include +#include +#include +#include +#include +#include +#include + +static int +do_test (void) +{ + struct link_map *l = xdlopen (LIBC_SO, RTLD_NOW); + + /* The dynamic loader fills holes with PROT_NONE mappings. */ + if (!l->l_contiguous) + FAIL_EXIT1 ("libc.so link map is not contiguous"); + + /* Direct probing does not work because not everything is readable + due to PROT_NONE mappings. */ + int pagesize = getpagesize (); + ElfW(Addr) addr = l->l_map_start; + TEST_COMPARE (addr % pagesize, 0); + while (addr < l->l_map_end) + { + void *expected = (void *) addr; + void *ptr = xmmap (expected, 1, PROT_READ | PROT_WRITE, + MAP_PRIVATE | MAP_ANONYMOUS, -1); + if (ptr == expected) + FAIL ("hole in libc.so memory image after %lu bytes", + (unsigned long int) (addr - l->l_map_start)); + xmunmap (ptr, 1); + addr += pagesize; + } + + xdlclose (l); + + return 0; +} +#include diff --git a/elf/tst-link-map-contiguous-main.c b/elf/tst-link-map-contiguous-main.c new file mode 100644 index 0000000000..2d1a054f0f --- /dev/null +++ b/elf/tst-link-map-contiguous-main.c @@ -0,0 +1,45 @@ +/* Check that the entire main program image is readable if contiguous. + Copyright (C) 2025 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + . */ + +#include +#include +#include + +static int +do_test (void) +{ + struct link_map *l = xdlopen ("", RTLD_NOW); + if (!l->l_contiguous) + FAIL_UNSUPPORTED ("main link map is not contiguous"); + + /* This check only works if the kernel loaded the main program. The + dynamic loader replaces gaps with PROT_NONE mappings, resulting + in faults. */ + volatile long int *p = (volatile long int *) l->l_map_start; + volatile long int *end = (volatile long int *) l->l_map_end; + while (p < end) + { + *p; + ++p; + } + + xdlclose (l); + + return 0; +} +#include diff --git a/inet/bits/inet-fortified.h b/inet/bits/inet-fortified.h index 6738221a54..cc476ebcfd 100644 --- a/inet/bits/inet-fortified.h +++ b/inet/bits/inet-fortified.h @@ -45,15 +45,15 @@ __NTH (inet_pton (int __af, __fortify_clang_warning_only_if_bos0_lt (4, __dst, "inet_pton called with destination buffer size less than 4") { - size_t sz = 0; + size_t __sz = 0; if (__af == AF_INET) - sz = sizeof (struct in_addr); + __sz = sizeof (struct in_addr); else if (__af == AF_INET6) - sz = sizeof (struct in6_addr); + __sz = sizeof (struct in6_addr); else return __inet_pton_alias (__af, __src, __dst); - return __glibc_fortify (inet_pton, sz, sizeof (char), + return __glibc_fortify (inet_pton, __sz, sizeof (char), __glibc_objsize (__dst), __af, __src, __dst); }; diff --git a/libio/stdio.h b/libio/stdio.h index d042b36618..e0e70945fa 100644 --- a/libio/stdio.h +++ b/libio/stdio.h @@ -168,11 +168,11 @@ extern int renameat (int __oldfd, const char *__old, int __newfd, #ifdef __USE_GNU /* Flags for renameat2. */ # define RENAME_NOREPLACE (1 << 0) -# define AT_RENAME_NOREPLACE RENAME_NOREPLACE +# define AT_RENAME_NOREPLACE 0x0001 # define RENAME_EXCHANGE (1 << 1) -# define AT_RENAME_EXCHANGE RENAME_EXCHANGE +# define AT_RENAME_EXCHANGE 0x0002 # define RENAME_WHITEOUT (1 << 2) -# define AT_RENAME_WHITEOUT RENAME_WHITEOUT +# define AT_RENAME_WHITEOUT 0x0004 /* Rename file OLD relative to OLDFD to NEW relative to NEWFD, with additional flags. */ diff --git a/malloc/malloc.c b/malloc/malloc.c index 5ca390cc22..5f3e701fd1 100644 --- a/malloc/malloc.c +++ b/malloc/malloc.c @@ -230,6 +230,9 @@ /* For uintptr_t. */ #include +/* For stdc_count_ones. */ +#include + /* For va_arg, va_start, va_end. */ #include @@ -294,9 +297,9 @@ # define TCACHE_SMALL_BINS 64 # define TCACHE_LARGE_BINS 12 /* Up to 4M chunks */ # define TCACHE_MAX_BINS (TCACHE_SMALL_BINS + TCACHE_LARGE_BINS) -# define MAX_TCACHE_SMALL_SIZE tidx2usize (TCACHE_SMALL_BINS-1) +# define MAX_TCACHE_SMALL_SIZE tidx2csize (TCACHE_SMALL_BINS-1) -/* Only used to pre-fill the tunables. */ +# define tidx2csize(idx) (((size_t) idx) * MALLOC_ALIGNMENT + MINSIZE) # define tidx2usize(idx) (((size_t) idx) * MALLOC_ALIGNMENT + MINSIZE - SIZE_SZ) /* When "x" is from chunksize(). */ @@ -1932,7 +1935,7 @@ static struct malloc_par mp_ = , .tcache_count = TCACHE_FILL_COUNT, .tcache_small_bins = TCACHE_SMALL_BINS, - .tcache_max_bytes = MAX_TCACHE_SMALL_SIZE, + .tcache_max_bytes = MAX_TCACHE_SMALL_SIZE + 1, .tcache_unsorted_limit = 0 /* No limit. */ #endif }; @@ -3152,6 +3155,19 @@ tcache_key_initialize (void) if (__getrandom_nocancel_nostatus_direct (&tcache_key, sizeof(tcache_key), GRND_NONBLOCK) != sizeof (tcache_key)) + tcache_key = 0; + + /* We need tcache_key to be non-zero (otherwise tcache_double_free_verify's + clearing of e->key would go unnoticed and it would loop getting called + through __libc_free), and we want tcache_key not to be a + commonly-occurring value in memory, so ensure a minimum amount of one and + zero bits. */ + int minimum_bits = __WORDSIZE / 4; + int maximum_bits = __WORDSIZE - minimum_bits; + + while (labs ((intptr_t) tcache_key) <= 0x1000000 + || stdc_count_ones (tcache_key) < minimum_bits + || stdc_count_ones (tcache_key) > maximum_bits) { tcache_key = random_bits (); #if __WORDSIZE == 64 @@ -3208,11 +3224,10 @@ tcache_get_n (size_t tc_idx, tcache_entry **ep, bool mangled) if (__glibc_unlikely (misaligned_mem (e))) malloc_printerr ("malloc(): unaligned tcache chunk detected"); - void *ne = e == NULL ? NULL : REVEAL_PTR (e->next); if (!mangled) - *ep = ne; + *ep = REVEAL_PTR (e->next); else - *ep = PROTECT_PTR (ep, ne); + *ep = PROTECT_PTR (ep, REVEAL_PTR (e->next)); ++(tcache->num_slots[tc_idx]); e->key = 0; @@ -3229,7 +3244,7 @@ tcache_put (mchunkptr chunk, size_t tc_idx) static __always_inline void * tcache_get (size_t tc_idx) { - return tcache_get_n (tc_idx, & tcache->entries[tc_idx], false); + return tcache_get_n (tc_idx, &tcache->entries[tc_idx], false); } static __always_inline tcache_entry ** @@ -5587,15 +5602,13 @@ do_set_arena_max (size_t value) static __always_inline int do_set_tcache_max (size_t value) { + if (value > PTRDIFF_MAX) + return 0; + size_t nb = request2size (value); size_t tc_idx = csize2tidx (nb); - /* To check that value is not too big and request2size does not return an - overflown value. */ - if (value > nb) - return 0; - - if (nb > MAX_TCACHE_SMALL_SIZE) + if (tc_idx >= TCACHE_SMALL_BINS) tc_idx = large_csize2tidx (nb); LIBC_PROBE (memory_tunable_tcache_max_bytes, 2, value, mp_.tcache_max_bytes); @@ -5604,7 +5617,7 @@ do_set_tcache_max (size_t value) { if (tc_idx < TCACHE_SMALL_BINS) mp_.tcache_small_bins = tc_idx + 1; - mp_.tcache_max_bytes = nb; + mp_.tcache_max_bytes = nb + 1; return 1; } diff --git a/nptl/allocatestack.c b/nptl/allocatestack.c index 800ca89720..fb8a60a21d 100644 --- a/nptl/allocatestack.c +++ b/nptl/allocatestack.c @@ -240,7 +240,7 @@ setup_stack_prot (char *mem, size_t size, struct pthread *pd, /* Update the guard area of the thread stack MEM of size SIZE with the new GUARDISZE. It uses the method defined by PD stack_mode. */ static inline bool -adjust_stack_prot (char *mem, size_t size, const struct pthread *pd, +adjust_stack_prot (char *mem, size_t size, struct pthread *pd, size_t guardsize, size_t pagesize_m1) { /* The required guard area is larger than the current one. For @@ -258,11 +258,23 @@ adjust_stack_prot (char *mem, size_t size, const struct pthread *pd, so use the new guard placement with the new size. */ if (guardsize > pd->guardsize) { + /* There was no need to previously setup a guard page, so we need + to check whether the kernel supports guard advise. */ char *guard = guard_position (mem, size, guardsize, pd, pagesize_m1); - if (pd->stack_mode == ALLOCATE_GUARD_MADV_GUARD) - return __madvise (guard, guardsize, MADV_GUARD_INSTALL) == 0; - else if (pd->stack_mode == ALLOCATE_GUARD_PROT_NONE) - return __mprotect (guard, guardsize, PROT_NONE) == 0; + if (atomic_load_relaxed (&allocate_stack_mode) + == ALLOCATE_GUARD_MADV_GUARD) + { + if (__madvise (guard, guardsize, MADV_GUARD_INSTALL) == 0) + { + pd->stack_mode = ALLOCATE_GUARD_MADV_GUARD; + return true; + } + atomic_store_relaxed (&allocate_stack_mode, + ALLOCATE_GUARD_PROT_NONE); + } + + pd->stack_mode = ALLOCATE_GUARD_PROT_NONE; + return __mprotect (guard, guardsize, PROT_NONE) == 0; } /* The current guard area is larger than the required one. For _STACK_GROWS_DOWN is means change the guard as: diff --git a/nptl/cancellation.c b/nptl/cancellation.c index 156e63dcf0..bed0383a23 100644 --- a/nptl/cancellation.c +++ b/nptl/cancellation.c @@ -72,8 +72,8 @@ __syscall_cancel (__syscall_arg_t a1, __syscall_arg_t a2, __syscall_arg_t a5, __syscall_arg_t a6, __SYSCALL_CANCEL7_ARG_DEF __syscall_arg_t nr) { - int r = __internal_syscall_cancel (a1, a2, a3, a4, a5, a6, - __SYSCALL_CANCEL7_ARG nr); + long int r = __internal_syscall_cancel (a1, a2, a3, a4, a5, a6, + __SYSCALL_CANCEL7_ARG nr); return __glibc_unlikely (INTERNAL_SYSCALL_ERROR_P (r)) ? SYSCALL_ERROR_LABEL (INTERNAL_SYSCALL_ERRNO (r)) : r; diff --git a/nptl/tst-guard1.c b/nptl/tst-guard1.c index e3e06df0fc..1c73d3fc93 100644 --- a/nptl/tst-guard1.c +++ b/nptl/tst-guard1.c @@ -21,6 +21,7 @@ #include #include #include +#include #include #include #include @@ -202,7 +203,7 @@ tf (void *closure) /* Test 1: caller provided stack without guard. */ static void -do_test1 (void) +do_test1 (void *closure) { pthread_attr_t attr; xpthread_attr_init (&attr); @@ -227,7 +228,7 @@ do_test1 (void) /* Test 2: same as 1., but with a guard area. */ static void -do_test2 (void) +do_test2 (void *closure) { pthread_attr_t attr; xpthread_attr_init (&attr); @@ -250,18 +251,9 @@ do_test2 (void) xmunmap (stack, stacksize); } -/* Test 3: pthread_create with default values. */ +/* Test 3: pthread_create without a guard area. */ static void -do_test3 (void) -{ - pthread_t t = xpthread_create (NULL, tf, NULL); - void *status = xpthread_join (t); - TEST_VERIFY (status == 0); -} - -/* Test 4: pthread_create without a guard area. */ -static void -do_test4 (void) +do_test3 (void *closure) { pthread_attr_t attr; xpthread_attr_init (&attr); @@ -277,9 +269,18 @@ do_test4 (void) xpthread_attr_destroy (&attr); } +/* Test 4: pthread_create with default values. */ +static void +do_test4 (void *closure) +{ + pthread_t t = xpthread_create (NULL, tf, NULL); + void *status = xpthread_join (t); + TEST_VERIFY (status == 0); +} + /* Test 5: pthread_create with non default stack and guard size value. */ static void -do_test5 (void) +do_test5 (void *closure) { pthread_attr_t attr; xpthread_attr_init (&attr); @@ -299,7 +300,7 @@ do_test5 (void) test 3, but with a larger guard area. The pthread_create will need to increase the guard area. */ static void -do_test6 (void) +do_test6 (void *closure) { pthread_attr_t attr; xpthread_attr_init (&attr); @@ -320,7 +321,7 @@ do_test6 (void) pthread_create should use the cached stack from previous tests, but it would require to reduce the guard area. */ static void -do_test7 (void) +do_test7 (void *closure) { pthread_t t = xpthread_create (NULL, tf, NULL); void *status = xpthread_join (t); @@ -346,21 +347,40 @@ do_test (void) static const struct { const char *descr; - void (*test)(void); + void (*test) (void *); } tests[] = { { "user provided stack without guard", do_test1 }, { "user provided stack with guard", do_test2 }, - { "default attribute", do_test3 }, - { "default attribute without guard", do_test4 }, + /* N.B: do_test3 should be before do_test4 to check if a new thread + that uses the thread stack previously allocated without a guard + page correctly sets up the guard pages even on a kernel without + MADV_GUARD_INSTALL support (BZ 33356). */ + { "default attribute without guard", do_test3 }, + { "default attribute", do_test4 }, + /* Also checks if the guard is correctly removed from the cache thread + stack. */ + { "default attribute without guard", do_test3 }, { "non default stack and guard sizes", do_test5 }, { "reused stack with larger guard", do_test6 }, { "reused stack with smaller guard", do_test7 }, }; + /* Run each test with a clean state. */ + for (int i = 0; i < array_length (tests); i++) + { + printf ("debug: fork: test%01d: %s\n", i, tests[i].descr); + struct support_capture_subprocess result = + support_capture_subprocess (tests[i].test, NULL); + support_capture_subprocess_check (&result, tests[i].descr, 0, + sc_allow_none); + support_capture_subprocess_free (&result); + } + + /* And now run the same tests along with the thread stack cache. */ for (int i = 0; i < array_length (tests); i++) { printf ("debug: test%01d: %s\n", i, tests[i].descr); - tests[i].test(); + tests[i].test ( NULL); } return 0; diff --git a/nss/getXXbyYY_r.c b/nss/getXXbyYY_r.c index eae6c3480e..2b0735fb6a 100644 --- a/nss/getXXbyYY_r.c +++ b/nss/getXXbyYY_r.c @@ -157,19 +157,15 @@ __merge_einval (LOOKUP_TYPE *a, #define CHECK_MERGE(err, status) \ ({ \ - do \ + if (err) \ { \ - if (err) \ - { \ - __set_errno (err); \ - if (err == ERANGE) \ - status = NSS_STATUS_TRYAGAIN; \ - else \ - status = NSS_STATUS_UNAVAIL; \ - break; \ - } \ + __set_errno (err); \ + if (err == ERANGE) \ + status = NSS_STATUS_TRYAGAIN; \ + else \ + status = NSS_STATUS_UNAVAIL; \ + break; \ } \ - while (0); \ }) /* Type of the lookup function we need here. */ diff --git a/stdio-common/tst-renameat2.c b/stdio-common/tst-renameat2.c index 12aa0f8b0f..6213e1376d 100644 --- a/stdio-common/tst-renameat2.c +++ b/stdio-common/tst-renameat2.c @@ -28,6 +28,12 @@ #include #include +/* These constants are defined with different token sequences, + matching the Linux definitions, to avoid preprocessor warnings. */ +_Static_assert (RENAME_NOREPLACE == AT_RENAME_NOREPLACE, "RENAME_NOREPLACE"); +_Static_assert (RENAME_EXCHANGE == AT_RENAME_EXCHANGE, "RENAME_EXCHANGE"); +_Static_assert (RENAME_WHITEOUT == AT_RENAME_WHITEOUT, "RENAME_WHITEOUT"); + /* Directory with the temporary files. */ static char *directory; static int directory_fd; diff --git a/stdlib/abort.c b/stdlib/abort.c index caa9e6dc04..904244a2fb 100644 --- a/stdlib/abort.c +++ b/stdlib/abort.c @@ -19,6 +19,7 @@ #include #include #include +#include #include /* Try to get a machine dependent instruction which will make the @@ -42,7 +43,10 @@ __libc_rwlock_define_initialized (static, lock); void __abort_fork_reset_child (void) { - __libc_rwlock_init (lock); + /* Reinitialize lock without calling pthread_rwlock_init, to + avoid a valgrind DRD false positive. */ + __libc_rwlock_define_initialized (, reset_lock); + memcpy (&lock, &reset_lock, sizeof (lock)); } void diff --git a/support/support_capture_subprocess.c b/support/support_capture_subprocess.c index b4e4bf9502..c89e65b534 100644 --- a/support/support_capture_subprocess.c +++ b/support/support_capture_subprocess.c @@ -133,6 +133,27 @@ copy_and_spawn_sgid (const char *child_id, gid_t gid) if (chmod (execname, 02750) != 0) FAIL_UNSUPPORTED ("cannot make \"%s\" SGID: %m ", execname); + /* Now we can drop the privilege of that group. */ + const int count = 64; + gid_t groups[count]; + int ngroups = getgroups(count, groups); + + if (ngroups < 0) + FAIL_UNSUPPORTED ("Could not get group list again for user %jd\n", + (intmax_t) getuid ()); + + int n = 0; + for (int i = 0; i < ngroups; i++) + { + if (groups[i] != gid) + { + if (n != i) + groups[n] = groups[i]; + n++; + } + } + setgroups (n, groups); + /* We have the binary, now spawn the subprocess. Avoid using support_subprogram because we only want the program exit status, not the contents. */ diff --git a/support/support_subprocess.c b/support/support_subprocess.c index be00dde3a7..8bf9a33ea2 100644 --- a/support/support_subprocess.c +++ b/support/support_subprocess.c @@ -25,6 +25,7 @@ #include #include #include +#include static struct support_subprocess support_subprocess_init (void) @@ -60,6 +61,8 @@ support_subprocess (void (*callback) (void *), void *closure) xclose (result.stdout_pipe[1]); xclose (result.stderr_pipe[1]); callback (closure); + /* Make sure that temporary files are deleted. */ + support_delete_temp_files (); _exit (0); } xclose (result.stdout_pipe[1]); diff --git a/sysdeps/aarch64/Makefile b/sysdeps/aarch64/Makefile index bb97d31355..9479fb9679 100644 --- a/sysdeps/aarch64/Makefile +++ b/sysdeps/aarch64/Makefile @@ -79,8 +79,18 @@ sysdep_routines += \ tests += \ tst-sme-jmp \ + tst-sme-signal \ tst-sme-za-state \ # tests +tests-internal += \ + tst-sme-clone \ + tst-sme-clone3 \ + tst-sme-fork \ + tst-sme-vfork \ + # tests-internal + +$(objpfx)tst-sme-clone3: $(objpfx)clone3.o $(objpfx)__arm_za_disable.o + endif ifeq ($(subdir),malloc) diff --git a/sysdeps/aarch64/__longjmp.S b/sysdeps/aarch64/__longjmp.S index 70ac02c44b..53b42e1bdc 100644 --- a/sysdeps/aarch64/__longjmp.S +++ b/sysdeps/aarch64/__longjmp.S @@ -26,16 +26,8 @@ ENTRY (__longjmp) #if IS_IN(libc) - /* Disable ZA state of SME in libc.a and libc.so, but not in ld.so. - The calling convention of __libc_arm_za_disable allows to do - this thus allowing to avoid saving to and reading from stack. - As a result we also don't need to sign the return address and - check it after returning because it is not stored to stack. */ - mov x13, x30 - cfi_register (x30, x13) - bl __libc_arm_za_disable - mov x30, x13 - cfi_register (x13, x30) + /* Disable ZA state of SME in libc.a and libc.so, but not in ld.so. */ + CALL_LIBC_ARM_ZA_DISABLE #endif cfi_def_cfa (x0, 0) diff --git a/sysdeps/aarch64/fpu/pow_sve.c b/sysdeps/aarch64/fpu/pow_sve.c index b8c1b39dca..becf1a8410 100644 --- a/sysdeps/aarch64/fpu/pow_sve.c +++ b/sysdeps/aarch64/fpu/pow_sve.c @@ -31,8 +31,8 @@ The SVE algorithm drops the tail in the exp computation at the price of a lower accuracy, slightly above 1ULP. The SVE algorithm also drops the special treatement of small (< 2^-65) and - large (> 2^63) finite values of |y|, as they only affect non-round to nearest - modes. + large (> 2^63) finite values of |y|, as they only affect non-round to + nearest modes. Maximum measured error is 1.04 ULPs: SV_NAME_D2 (pow) (0x1.3d2d45bc848acp+63, -0x1.a48a38b40cd43p-12) @@ -156,42 +156,22 @@ sv_zeroinfnan (svbool_t pg, svuint64_t i) a double. (int32_t)KI is the k used in the argument reduction and exponent adjustment of scale, positive k here means the result may overflow and negative k means the result may underflow. */ -static inline double -specialcase (double tmp, uint64_t sbits, uint64_t ki) -{ - double scale; - if ((ki & 0x80000000) == 0) - { - /* k > 0, the exponent of scale might have overflowed by <= 460. */ - sbits -= 1009ull << 52; - scale = asdouble (sbits); - return 0x1p1009 * (scale + scale * tmp); - } - /* k < 0, need special care in the subnormal range. */ - sbits += 1022ull << 52; - /* Note: sbits is signed scale. */ - scale = asdouble (sbits); - double y = scale + scale * tmp; - return 0x1p-1022 * y; -} - -/* Scalar fallback for special cases of SVE pow's exp. */ static inline svfloat64_t -sv_call_specialcase (svfloat64_t x1, svuint64_t u1, svuint64_t u2, - svfloat64_t y, svbool_t cmp) +specialcase (svfloat64_t tmp, svuint64_t sbits, svuint64_t ki, svbool_t cmp) { - svbool_t p = svpfirst (cmp, svpfalse ()); - while (svptest_any (cmp, p)) - { - double sx1 = svclastb (p, 0, x1); - uint64_t su1 = svclastb (p, 0, u1); - uint64_t su2 = svclastb (p, 0, u2); - double elem = specialcase (sx1, su1, su2); - svfloat64_t y2 = sv_f64 (elem); - y = svsel (p, y2, y); - p = svpnext_b64 (cmp, p); - } - return y; + svbool_t p_pos = svcmpge_n_f64 (cmp, svreinterpret_f64_u64 (ki), 0.0); + + /* Scale up or down depending on sign of k. */ + svint64_t offset + = svsel_s64 (p_pos, sv_s64 (1009ull << 52), sv_s64 (-1022ull << 52)); + svfloat64_t factor + = svsel_f64 (p_pos, sv_f64 (0x1p1009), sv_f64 (0x1p-1022)); + + svuint64_t offset_sbits + = svsub_u64_x (cmp, sbits, svreinterpret_u64_s64 (offset)); + svfloat64_t scale = svreinterpret_f64_u64 (offset_sbits); + svfloat64_t res = svmad_f64_x (cmp, scale, tmp, scale); + return svmul_f64_x (cmp, res, factor); } /* Compute y+TAIL = log(x) where the rounded result is y and TAIL has about @@ -214,8 +194,8 @@ sv_log_inline (svbool_t pg, svuint64_t ix, svfloat64_t *tail, /* log(x) = k*Ln2 + log(c) + log1p(z/c-1). */ /* SVE lookup requires 3 separate lookup tables, as opposed to scalar version - that uses array of structures. We also do the lookup earlier in the code to - make sure it finishes as early as possible. */ + that uses array of structures. We also do the lookup earlier in the code + to make sure it finishes as early as possible. */ svfloat64_t invc = svld1_gather_index (pg, __v_pow_log_data.invc, i); svfloat64_t logc = svld1_gather_index (pg, __v_pow_log_data.logc, i); svfloat64_t logctail = svld1_gather_index (pg, __v_pow_log_data.logctail, i); @@ -325,14 +305,14 @@ sv_exp_inline (svbool_t pg, svfloat64_t x, svfloat64_t xtail, svbool_t oflow = svcmpge (pg, abstop, HugeExp); oflow = svand_z (pg, uoflow, svbic_z (pg, oflow, uflow)); - /* For large |x| values (512 < |x| < 1024) scale * (1 + TMP) can overflow - or underflow. */ + /* Handle underflow and overlow in scale. + For large |x| values (512 < |x| < 1024), scale * (1 + TMP) can + overflow or underflow. */ svbool_t special = svbic_z (pg, uoflow, svorr_z (pg, uflow, oflow)); + if (__glibc_unlikely (svptest_any (pg, special))) + z = svsel (special, specialcase (tmp, sbits, ki, special), z); - /* Update result with special and large cases. */ - z = sv_call_specialcase (tmp, sbits, ki, z, special); - - /* Handle underflow and overflow. */ + /* Handle underflow and overflow in exp. */ svbool_t x_is_neg = svcmplt (pg, x, 0); svuint64_t sign_mask = svlsl_x (pg, sign_bias, 52 - V_POW_EXP_TABLE_BITS); @@ -353,7 +333,7 @@ sv_exp_inline (svbool_t pg, svfloat64_t x, svfloat64_t xtail, } static inline double -pow_sc (double x, double y) +pow_specialcase (double x, double y) { uint64_t ix = asuint64 (x); uint64_t iy = asuint64 (y); @@ -382,6 +362,14 @@ pow_sc (double x, double y) return x; } +/* Scalar fallback for special case routines with custom signature. */ +static svfloat64_t NOINLINE +sv_pow_specialcase (svfloat64_t x1, svfloat64_t x2, svfloat64_t y, + svbool_t cmp) +{ + return sv_call2_f64 (pow_specialcase, x1, x2, y, cmp); +} + svfloat64_t SV_NAME_D2 (pow) (svfloat64_t x, svfloat64_t y, const svbool_t pg) { const struct data *d = ptr_barrier (&data); @@ -444,7 +432,7 @@ svfloat64_t SV_NAME_D2 (pow) (svfloat64_t x, svfloat64_t y, const svbool_t pg) /* Cases of zero/inf/nan x or y. */ if (__glibc_unlikely (svptest_any (svptrue_b64 (), special))) - vz = sv_call2_f64 (pow_sc, x, y, vz, special); + vz = sv_pow_specialcase (x, y, vz, special); return vz; } diff --git a/sysdeps/aarch64/fpu/powf_sve.c b/sysdeps/aarch64/fpu/powf_sve.c index 7046990aa1..76f54b3522 100644 --- a/sysdeps/aarch64/fpu/powf_sve.c +++ b/sysdeps/aarch64/fpu/powf_sve.c @@ -116,11 +116,10 @@ zeroinfnan (uint32_t ix) preamble of scalar powf except that we do not update ix and sign_bias. This is done in the preamble of the SVE powf. */ static inline float -powf_specialcase (float x, float y, float z) +powf_specialcase (float x, float y) { uint32_t ix = asuint (x); uint32_t iy = asuint (y); - /* Either (x < 0x1p-126 or inf or nan) or (y is 0 or inf or nan). */ if (__glibc_unlikely (zeroinfnan (iy))) { if (2 * iy == 0) @@ -142,32 +141,15 @@ powf_specialcase (float x, float y, float z) x2 = -x2; return iy & 0x80000000 ? 1 / x2 : x2; } - /* We need a return here in case x<0 and y is integer, but all other tests - need to be run. */ - return z; + /* Return x for convenience, but make sure result is never used. */ + return x; } /* Scalar fallback for special case routines with custom signature. */ static svfloat32_t NOINLINE -sv_call_powf_sc (svfloat32_t x1, svfloat32_t x2, svfloat32_t y) +sv_call_powf_sc (svfloat32_t x1, svfloat32_t x2, svfloat32_t y, svbool_t cmp) { - /* Special cases of x or y: zero, inf and nan. */ - svbool_t xspecial = sv_zeroinfnan (svptrue_b32 (), svreinterpret_u32 (x1)); - svbool_t yspecial = sv_zeroinfnan (svptrue_b32 (), svreinterpret_u32 (x2)); - svbool_t cmp = svorr_z (svptrue_b32 (), xspecial, yspecial); - - svbool_t p = svpfirst (cmp, svpfalse ()); - while (svptest_any (cmp, p)) - { - float sx1 = svclastb (p, 0, x1); - float sx2 = svclastb (p, 0, x2); - float elem = svclastb (p, 0, y); - elem = powf_specialcase (sx1, sx2, elem); - svfloat32_t y2 = sv_f32 (elem); - y = svsel (p, y2, y); - p = svpnext_b32 (cmp, p); - } - return y; + return sv_call2_f32 (powf_specialcase, x1, x2, y, cmp); } /* Compute core for half of the lanes in double precision. */ @@ -223,15 +205,15 @@ sv_powf_core (const svbool_t pg, svuint32_t i, svuint32_t iz, svint32_t k, const svbool_t ptrue = svptrue_b64 (); /* Unpack and promote input vectors (pg, y, z, i, k and sign_bias) into two - * in order to perform core computation in double precision. */ + in order to perform core computation in double precision. */ const svbool_t pg_lo = svunpklo (pg); const svbool_t pg_hi = svunpkhi (pg); - svfloat64_t y_lo - = svcvt_f64_x (pg, svreinterpret_f32 (svunpklo (svreinterpret_u32 (y)))); - svfloat64_t y_hi - = svcvt_f64_x (pg, svreinterpret_f32 (svunpkhi (svreinterpret_u32 (y)))); - svfloat64_t z_lo = svcvt_f64_x (pg, svreinterpret_f32 (svunpklo (iz))); - svfloat64_t z_hi = svcvt_f64_x (pg, svreinterpret_f32 (svunpkhi (iz))); + svfloat64_t y_lo = svcvt_f64_x ( + ptrue, svreinterpret_f32 (svunpklo (svreinterpret_u32 (y)))); + svfloat64_t y_hi = svcvt_f64_x ( + ptrue, svreinterpret_f32 (svunpkhi (svreinterpret_u32 (y)))); + svfloat64_t z_lo = svcvt_f64_x (ptrue, svreinterpret_f32 (svunpklo (iz))); + svfloat64_t z_hi = svcvt_f64_x (ptrue, svreinterpret_f32 (svunpkhi (iz))); svuint64_t i_lo = svunpklo (i); svuint64_t i_hi = svunpkhi (i); svint64_t k_lo = svunpklo (k); @@ -312,7 +294,7 @@ svfloat32_t SV_NAME_F2 (pow) (svfloat32_t x, svfloat32_t y, const svbool_t pg) (23 - V_POWF_EXP2_TABLE_BITS)); /* Compute core in extended precision and return intermediate ylogx results - * to handle cases of underflow and underflow in exp. */ + to handle cases of underflow and overflow in exp. */ svfloat32_t ylogx; svfloat32_t ret = sv_powf_core (yint_or_xpos, i, iz, k, y, sign_bias, &ylogx, d); @@ -330,7 +312,7 @@ svfloat32_t SV_NAME_F2 (pow) (svfloat32_t x, svfloat32_t y, const svbool_t pg) ret = svsel (yint_or_xpos, ret, sv_f32 (__builtin_nanf (""))); if (__glibc_unlikely (svptest_any (cmp, cmp))) - return sv_call_powf_sc (x, y, ret); + return sv_call_powf_sc (x, y, ret, cmp); return ret; } diff --git a/sysdeps/aarch64/fpu/sinh_advsimd.c b/sysdeps/aarch64/fpu/sinh_advsimd.c index 0d6a4856f8..b6b60262c6 100644 --- a/sysdeps/aarch64/fpu/sinh_advsimd.c +++ b/sysdeps/aarch64/fpu/sinh_advsimd.c @@ -24,36 +24,26 @@ static const struct data { struct v_expm1_data d; uint64x2_t halff; -#if WANT_SIMD_EXCEPT - uint64x2_t tiny_bound, thresh; -#else float64x2_t large_bound; -#endif } data = { .d = V_EXPM1_DATA, .halff = V2 (0x3fe0000000000000), -#if WANT_SIMD_EXCEPT - /* 2^-26, below which sinh(x) rounds to x. */ - .tiny_bound = V2 (0x3e50000000000000), - /* asuint(large_bound) - asuint(tiny_bound). */ - .thresh = V2 (0x0230000000000000), -#else /* 2^9. expm1 helper overflows for large input. */ .large_bound = V2 (0x1p+9), -#endif }; static float64x2_t NOINLINE VPCS_ATTR -special_case (float64x2_t x) +special_case (float64x2_t x, float64x2_t t, float64x2_t halfsign, + uint64x2_t special) { - return v_call_f64 (sinh, x, x, v_u64 (-1)); + return v_call_f64 (sinh, x, vmulq_f64 (t, halfsign), special); } /* Approximation for vector double-precision sinh(x) using expm1. sinh(x) = (exp(x) - exp(-x)) / 2. The greatest observed error is 2.52 ULP: - _ZGVnN2v_sinh(-0x1.a098a2177a2b9p-2) got -0x1.ac2f05bb66fccp-2 - want -0x1.ac2f05bb66fc9p-2. */ + _ZGVnN2v_sinh(0x1.9f6ff2ab6fb19p-2) got 0x1.aaed83a3153ccp-2 + want 0x1.aaed83a3153c9p-2. */ float64x2_t VPCS_ATTR V_NAME_D1 (sinh) (float64x2_t x) { const struct data *d = ptr_barrier (&data); @@ -63,21 +53,16 @@ float64x2_t VPCS_ATTR V_NAME_D1 (sinh) (float64x2_t x) float64x2_t halfsign = vreinterpretq_f64_u64 ( vbslq_u64 (v_u64 (0x8000000000000000), ix, d->halff)); -#if WANT_SIMD_EXCEPT - uint64x2_t special = vcgeq_u64 ( - vsubq_u64 (vreinterpretq_u64_f64 (ax), d->tiny_bound), d->thresh); -#else uint64x2_t special = vcageq_f64 (x, d->large_bound); -#endif - - /* Fall back to scalar variant for all lanes if any of them are special. */ - if (__glibc_unlikely (v_any_u64 (special))) - return special_case (x); /* Up to the point that expm1 overflows, we can use it to calculate sinh using a slight rearrangement of the definition of sinh. This allows us to retain acceptable accuracy for very small inputs. */ float64x2_t t = expm1_inline (ax, &d->d); t = vaddq_f64 (t, vdivq_f64 (t, vaddq_f64 (t, v_f64 (1.0)))); + + if (__glibc_unlikely (v_any_u64 (special))) + return special_case (x, t, halfsign, special); + return vmulq_f64 (t, halfsign); } diff --git a/sysdeps/aarch64/fpu/sv_math.h b/sysdeps/aarch64/fpu/sv_math.h index 3d576df4cc..65d7f0ff20 100644 --- a/sysdeps/aarch64/fpu/sv_math.h +++ b/sysdeps/aarch64/fpu/sv_math.h @@ -24,11 +24,29 @@ #include "vecmath_config.h" +#if !defined(__ARM_FEATURE_SVE_BITS) || __ARM_FEATURE_SVE_BITS == 0 +/* If not specified by -msve-vector-bits, assume maximum vector length. */ +# define SVE_VECTOR_BYTES 256 +#else +# define SVE_VECTOR_BYTES (__ARM_FEATURE_SVE_BITS / 8) +#endif +#define SVE_NUM_FLTS (SVE_VECTOR_BYTES / sizeof (float)) +#define SVE_NUM_DBLS (SVE_VECTOR_BYTES / sizeof (double)) +/* Predicate is stored as one bit per byte of VL so requires VL / 64 bytes. */ +#define SVE_NUM_PG_BYTES (SVE_VECTOR_BYTES / sizeof (uint64_t)) + #define SV_NAME_F1(fun) _ZGVsMxv_##fun##f #define SV_NAME_D1(fun) _ZGVsMxv_##fun #define SV_NAME_F2(fun) _ZGVsMxvv_##fun##f #define SV_NAME_D2(fun) _ZGVsMxvv_##fun +static inline void +svstr_p (uint8_t *dst, svbool_t p) +{ + /* Predicate STR does not currently have an intrinsic. */ + __asm__("str %0, [%x1]\n" : : "Upa"(p), "r"(dst) : "memory"); +} + /* Double precision. */ static inline svint64_t sv_s64 (int64_t x) @@ -51,33 +69,35 @@ sv_f64 (double x) static inline svfloat64_t sv_call_f64 (double (*f) (double), svfloat64_t x, svfloat64_t y, svbool_t cmp) { - svbool_t p = svpfirst (cmp, svpfalse ()); - while (svptest_any (cmp, p)) + double tmp[SVE_NUM_DBLS]; + uint8_t pg_bits[SVE_NUM_PG_BYTES]; + svstr_p (pg_bits, cmp); + svst1 (svptrue_b64 (), tmp, svsel (cmp, x, y)); + + for (int i = 0; i < svcntd (); i++) { - double elem = svclastb_n_f64 (p, 0, x); - elem = (*f) (elem); - svfloat64_t y2 = svdup_n_f64 (elem); - y = svsel_f64 (p, y2, y); - p = svpnext_b64 (cmp, p); + if (pg_bits[i] & 1) + tmp[i] = f (tmp[i]); } - return y; + return svld1 (svptrue_b64 (), tmp); } static inline svfloat64_t sv_call2_f64 (double (*f) (double, double), svfloat64_t x1, svfloat64_t x2, svfloat64_t y, svbool_t cmp) { - svbool_t p = svpfirst (cmp, svpfalse ()); - while (svptest_any (cmp, p)) + double tmp1[SVE_NUM_DBLS], tmp2[SVE_NUM_DBLS]; + uint8_t pg_bits[SVE_NUM_PG_BYTES]; + svstr_p (pg_bits, cmp); + svst1 (svptrue_b64 (), tmp1, svsel (cmp, x1, y)); + svst1 (cmp, tmp2, x2); + + for (int i = 0; i < svcntd (); i++) { - double elem1 = svclastb_n_f64 (p, 0, x1); - double elem2 = svclastb_n_f64 (p, 0, x2); - double ret = (*f) (elem1, elem2); - svfloat64_t y2 = svdup_n_f64 (ret); - y = svsel_f64 (p, y2, y); - p = svpnext_b64 (cmp, p); + if (pg_bits[i] & 1) + tmp1[i] = f (tmp1[i], tmp2[i]); } - return y; + return svld1 (svptrue_b64 (), tmp1); } static inline svuint64_t @@ -109,33 +129,40 @@ sv_f32 (float x) static inline svfloat32_t sv_call_f32 (float (*f) (float), svfloat32_t x, svfloat32_t y, svbool_t cmp) { - svbool_t p = svpfirst (cmp, svpfalse ()); - while (svptest_any (cmp, p)) + float tmp[SVE_NUM_FLTS]; + uint8_t pg_bits[SVE_NUM_PG_BYTES]; + svstr_p (pg_bits, cmp); + svst1 (svptrue_b32 (), tmp, svsel (cmp, x, y)); + + for (int i = 0; i < svcntd (); i++) { - float elem = svclastb_n_f32 (p, 0, x); - elem = f (elem); - svfloat32_t y2 = svdup_n_f32 (elem); - y = svsel_f32 (p, y2, y); - p = svpnext_b32 (cmp, p); + uint8_t p = pg_bits[i]; + if (p & 1) + tmp[i * 2] = f (tmp[i * 2]); + if (p & (1 << 4)) + tmp[i * 2 + 1] = f (tmp[i * 2 + 1]); } - return y; + return svld1 (svptrue_b32 (), tmp); } static inline svfloat32_t sv_call2_f32 (float (*f) (float, float), svfloat32_t x1, svfloat32_t x2, svfloat32_t y, svbool_t cmp) { - svbool_t p = svpfirst (cmp, svpfalse ()); - while (svptest_any (cmp, p)) + float tmp1[SVE_NUM_FLTS], tmp2[SVE_NUM_FLTS]; + uint8_t pg_bits[SVE_NUM_PG_BYTES]; + svstr_p (pg_bits, cmp); + svst1 (svptrue_b32 (), tmp1, svsel (cmp, x1, y)); + svst1 (cmp, tmp2, x2); + + for (int i = 0; i < svcntd (); i++) { - float elem1 = svclastb_n_f32 (p, 0, x1); - float elem2 = svclastb_n_f32 (p, 0, x2); - float ret = f (elem1, elem2); - svfloat32_t y2 = svdup_n_f32 (ret); - y = svsel_f32 (p, y2, y); - p = svpnext_b32 (cmp, p); + uint8_t p = pg_bits[i]; + if (p & 1) + tmp1[i * 2] = f (tmp1[i * 2], tmp2[i * 2]); + if (p & (1 << 4)) + tmp1[i * 2 + 1] = f (tmp1[i * 2 + 1], tmp2[i * 2 + 1]); } - return y; + return svld1 (svptrue_b32 (), tmp1); } - #endif diff --git a/sysdeps/aarch64/fpu/tan_advsimd.c b/sysdeps/aarch64/fpu/tan_advsimd.c index 825c9754b3..d391a003d8 100644 --- a/sysdeps/aarch64/fpu/tan_advsimd.c +++ b/sysdeps/aarch64/fpu/tan_advsimd.c @@ -25,9 +25,7 @@ static const struct data float64x2_t poly[9]; double half_pi[2]; float64x2_t two_over_pi, shift; -#if !WANT_SIMD_EXCEPT float64x2_t range_val; -#endif } data = { /* Coefficients generated using FPMinimax. */ .poly = { V2 (0x1.5555555555556p-2), V2 (0x1.1111111110a63p-3), @@ -38,20 +36,17 @@ static const struct data .half_pi = { 0x1.921fb54442d18p0, 0x1.1a62633145c07p-54 }, .two_over_pi = V2 (0x1.45f306dc9c883p-1), .shift = V2 (0x1.8p52), -#if !WANT_SIMD_EXCEPT .range_val = V2 (0x1p23), -#endif }; #define RangeVal 0x4160000000000000 /* asuint64(0x1p23). */ #define TinyBound 0x3e50000000000000 /* asuint64(2^-26). */ -#define Thresh 0x310000000000000 /* RangeVal - TinyBound. */ /* Special cases (fall back to scalar calls). */ static float64x2_t VPCS_ATTR NOINLINE -special_case (float64x2_t x) +special_case (float64x2_t x, float64x2_t n, float64x2_t d, uint64x2_t special) { - return v_call_f64 (tan, x, x, v_u64 (-1)); + return v_call_f64 (tan, x, vdivq_f64 (n, d), special); } /* Vector approximation for double-precision tan. @@ -65,14 +60,6 @@ float64x2_t VPCS_ATTR V_NAME_D1 (tan) (float64x2_t x) very large inputs. Fall back to scalar routine for all lanes if any are too large, or Inf/NaN. If fenv exceptions are expected, also fall back for tiny input to avoid underflow. */ -#if WANT_SIMD_EXCEPT - uint64x2_t iax = vreinterpretq_u64_f64 (vabsq_f64 (x)); - /* iax - tiny_bound > range_val - tiny_bound. */ - uint64x2_t special - = vcgtq_u64 (vsubq_u64 (iax, v_u64 (TinyBound)), v_u64 (Thresh)); - if (__glibc_unlikely (v_any_u64 (special))) - return special_case (x); -#endif /* q = nearest integer to 2 * x / pi. */ float64x2_t q @@ -81,9 +68,8 @@ float64x2_t VPCS_ATTR V_NAME_D1 (tan) (float64x2_t x) /* Use q to reduce x to r in [-pi/4, pi/4], by: r = x - q * pi/2, in extended precision. */ - float64x2_t r = x; float64x2_t half_pi = vld1q_f64 (dat->half_pi); - r = vfmsq_laneq_f64 (r, q, half_pi, 0); + float64x2_t r = vfmsq_laneq_f64 (x, q, half_pi, 0); r = vfmsq_laneq_f64 (r, q, half_pi, 1); /* Further reduce r to [-pi/8, pi/8], to be reconstructed using double angle formula. */ @@ -114,12 +100,13 @@ float64x2_t VPCS_ATTR V_NAME_D1 (tan) (float64x2_t x) uint64x2_t no_recip = vtstq_u64 (vreinterpretq_u64_s64 (qi), v_u64 (1)); -#if !WANT_SIMD_EXCEPT uint64x2_t special = vcageq_f64 (x, dat->range_val); + float64x2_t swap = vbslq_f64 (no_recip, n, vnegq_f64 (d)); + d = vbslq_f64 (no_recip, d, n); + n = swap; + if (__glibc_unlikely (v_any_u64 (special))) - return special_case (x); -#endif + return special_case (x, n, d, special); - return vdivq_f64 (vbslq_f64 (no_recip, n, vnegq_f64 (d)), - vbslq_f64 (no_recip, d, n)); + return vdivq_f64 (n, d); } diff --git a/sysdeps/aarch64/fpu/tanpi_sve.c b/sysdeps/aarch64/fpu/tanpi_sve.c index 57c643ae29..bfe6828e1f 100644 --- a/sysdeps/aarch64/fpu/tanpi_sve.c +++ b/sysdeps/aarch64/fpu/tanpi_sve.c @@ -1,6 +1,6 @@ /* Double-precision (SVE) tanpi function - Copyright (C) 2024 Free Software Foundation, Inc. + Copyright (C) 2024-2025 Free Software Foundation, Inc. This file is part of the GNU C Library. The GNU C Library is free software; you can redistribute it and/or @@ -58,10 +58,10 @@ svfloat64_t SV_NAME_D1 (tanpi) (svfloat64_t x, const svbool_t pg) svfloat64_t r2 = svmul_x (pg, r, r); svfloat64_t r4 = svmul_x (pg, r2, r2); - svfloat64_t c_1_3 = svld1rq (pg, &d->c1); - svfloat64_t c_5_7 = svld1rq (pg, &d->c5); - svfloat64_t c_9_11 = svld1rq (pg, &d->c9); - svfloat64_t c_13_14 = svld1rq (pg, &d->c13); + svfloat64_t c_1_3 = svld1rq (svptrue_b64 (), &d->c1); + svfloat64_t c_5_7 = svld1rq (svptrue_b64 (), &d->c5); + svfloat64_t c_9_11 = svld1rq (svptrue_b64 (), &d->c9); + svfloat64_t c_13_14 = svld1rq (svptrue_b64 (), &d->c13); svfloat64_t p01 = svmla_lane (sv_f64 (d->c0), r2, c_1_3, 0); svfloat64_t p23 = svmla_lane (sv_f64 (d->c2), r2, c_1_3, 1); svfloat64_t p45 = svmla_lane (sv_f64 (d->c4), r2, c_5_7, 0); diff --git a/sysdeps/aarch64/fpu/tanpif_sve.c b/sysdeps/aarch64/fpu/tanpif_sve.c index 0285f56f34..6894379564 100644 --- a/sysdeps/aarch64/fpu/tanpif_sve.c +++ b/sysdeps/aarch64/fpu/tanpif_sve.c @@ -1,6 +1,6 @@ /* Single-precision (SVE) tanpi function - Copyright (C) 2024 Free Software Foundation, Inc. + Copyright (C) 2024-2025 Free Software Foundation, Inc. This file is part of the GNU C Library. The GNU C Library is free software; you can redistribute it and/or @@ -37,7 +37,7 @@ const static struct v_tanpif_data svfloat32_t SV_NAME_F1 (tanpi) (svfloat32_t x, const svbool_t pg) { const struct v_tanpif_data *d = ptr_barrier (&tanpif_data); - svfloat32_t odd_coeffs = svld1rq (pg, &d->c1); + svfloat32_t odd_coeffs = svld1rq (svptrue_b32 (), &d->c1); svfloat32_t n = svrintn_x (pg, x); /* inf produces nan that propagates. */ diff --git a/sysdeps/aarch64/setjmp.S b/sysdeps/aarch64/setjmp.S index 53c5e7d8cc..92cedfad83 100644 --- a/sysdeps/aarch64/setjmp.S +++ b/sysdeps/aarch64/setjmp.S @@ -37,16 +37,8 @@ ENTRY_ALIGN (__sigsetjmp, 2) 1: #if IS_IN(libc) - /* Disable ZA state of SME in libc.a and libc.so, but not in ld.so. - The calling convention of __libc_arm_za_disable allows to do - this thus allowing to avoid saving to and reading from stack. - As a result we also don't need to sign the return address and - check it after returning because it is not stored to stack. */ - mov x13, x30 - cfi_register (x30, x13) - bl __libc_arm_za_disable - mov x30, x13 - cfi_register (x13, x30) + /* Disable ZA state of SME in libc.a and libc.so, but not in ld.so. */ + CALL_LIBC_ARM_ZA_DISABLE #endif stp x19, x20, [x0, #JB_X19<<3] diff --git a/sysdeps/aarch64/tst-sme-clone.c b/sysdeps/aarch64/tst-sme-clone.c new file mode 100644 index 0000000000..b6ad54fa37 --- /dev/null +++ b/sysdeps/aarch64/tst-sme-clone.c @@ -0,0 +1,54 @@ +/* Test that ZA state of SME is cleared in both parent and child + when clone() syscall is used. + Copyright (C) 2025 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + . */ + +#include "tst-sme-skeleton.c" + +#include +#include + +static int +fun (void * const arg) +{ + printf ("in child: %s\n", (const char *)arg); + /* Check that ZA state of SME was disabled in child. */ + check_sme_za_state ("after clone in child", /* Clear. */ true); + return 0; +} + +static char __attribute__((aligned(16))) +stack[1024 * 1024]; + +static void +run (struct blk *ptr) +{ + char *syscall_name = (char *)"clone"; + printf ("in parent: before %s\n", syscall_name); + + /* Enabled ZA state so that effect of disabling be observable. */ + enable_sme_za_state (ptr); + check_sme_za_state ("before clone", /* Clear. */ false); + + pid_t pid = xclone (fun, syscall_name, stack, sizeof (stack), + CLONE_NEWUSER | CLONE_NEWNS | SIGCHLD); + + /* Check that ZA state of SME was disabled in parent. */ + check_sme_za_state ("after clone in parent", /* Clear. */ true); + + TEST_VERIFY (xwaitpid (pid, NULL, 0) == pid); +} diff --git a/sysdeps/aarch64/tst-sme-clone3.c b/sysdeps/aarch64/tst-sme-clone3.c new file mode 100644 index 0000000000..f420d5984d --- /dev/null +++ b/sysdeps/aarch64/tst-sme-clone3.c @@ -0,0 +1,84 @@ +/* Test that ZA state of SME is cleared in both parent and child + when clone3() syscall is used. + Copyright (C) 2025 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + . */ + +#include "tst-sme-skeleton.c" + +#include + +#include +#include +#include + +/* Since clone3 is not a public symbol, we link this test explicitly + with clone3.o and have to provide this declaration. */ +int __clone3 (struct clone_args *cl_args, size_t size, + int (*func)(void *arg), void *arg); + +static int +fun (void * const arg) +{ + printf ("in child: %s\n", (const char *)arg); + /* Check that ZA state of SME was disabled in child. */ + check_sme_za_state ("after clone3 in child", /* Clear. */ true); + return 0; +} + +static char __attribute__((aligned(16))) +stack[1024 * 1024]; + +/* Required by __arm_za_disable.o and provided by the startup code + as a hidden symbol. */ +uint64_t _dl_hwcap2; + +static void +run (struct blk *ptr) +{ + _dl_hwcap2 = getauxval (AT_HWCAP2); + + char *syscall_name = (char *)"clone3"; + struct clone_args args = { + .flags = CLONE_VM | CLONE_VFORK, + .exit_signal = SIGCHLD, + .stack = (uintptr_t) stack, + .stack_size = sizeof (stack), + }; + printf ("in parent: before %s\n", syscall_name); + + /* Enabled ZA state so that effect of disabling be observable. */ + enable_sme_za_state (ptr); + check_sme_za_state ("before clone3", /* Clear. */ false); + + pid_t pid = __clone3 (&args, sizeof (args), fun, syscall_name); + + /* Check that ZA state of SME was disabled in parent. */ + check_sme_za_state ("after clone3 in parent", /* Clear. */ true); + + printf ("%s child pid: %d\n", syscall_name, pid); + + xwaitpid (pid, NULL, 0); + printf ("in parent: after %s\n", syscall_name); +} + +/* Workaround to simplify linking with clone3.o. */ +void __syscall_error(int code) +{ + int err = -code; + fprintf (stderr, "syscall error %d (%s)\n", err, strerror (err)); + exit (err); +} diff --git a/sysdeps/aarch64/tst-sme-fork.c b/sysdeps/aarch64/tst-sme-fork.c new file mode 100644 index 0000000000..b003b08884 --- /dev/null +++ b/sysdeps/aarch64/tst-sme-fork.c @@ -0,0 +1,43 @@ +/* Test that ZA state of SME is cleared in both parent and child + when fork() function is used. + Copyright (C) 2025 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + . */ + +#include "tst-sme-skeleton.c" + +static void +run (struct blk *blk) +{ + /* Enabled ZA state so that effect of disabling be observable. */ + enable_sme_za_state (blk); + check_sme_za_state ("before fork", /* Clear. */ false); + fflush (stdout); + + pid_t pid = xfork (); + + if (pid == 0) + { + /* Check that ZA state of SME was disabled in child. */ + check_sme_za_state ("after fork in child", /* Clear. */ true); + exit (0); + } + + /* Check that ZA state of SME was disabled in parent. */ + check_sme_za_state ("after fork in parent", /* Clear. */ true); + + TEST_VERIFY (xwaitpid (pid, NULL, 0) == pid); +} diff --git a/sysdeps/aarch64/tst-sme-helper.h b/sysdeps/aarch64/tst-sme-helper.h index f049416c2b..ab9c503e45 100644 --- a/sysdeps/aarch64/tst-sme-helper.h +++ b/sysdeps/aarch64/tst-sme-helper.h @@ -16,9 +16,6 @@ License along with the GNU C Library; if not, see . */ -/* Streaming SVE vector register size. */ -static unsigned long svl; - struct blk { void *za_save_buffer; uint16_t num_za_save_slices; @@ -68,10 +65,10 @@ start_za (void) /* Load data into ZA byte by byte from p. */ static void __attribute__ ((noinline)) -load_za (const void *p) +load_za (const void *buf, unsigned long svl) { register unsigned long x15 asm ("x15") = 0; - register unsigned long x16 asm ("x16") = (unsigned long)p; + register unsigned long x16 asm ("x16") = (unsigned long)buf; register unsigned long x17 asm ("x17") = svl; asm volatile ( diff --git a/sysdeps/aarch64/tst-sme-jmp.c b/sysdeps/aarch64/tst-sme-jmp.c index 103897ad36..b2d21c6e1a 100644 --- a/sysdeps/aarch64/tst-sme-jmp.c +++ b/sysdeps/aarch64/tst-sme-jmp.c @@ -29,6 +29,9 @@ #include "tst-sme-helper.h" +/* Streaming SVE vector register size. */ +static unsigned long svl; + static uint8_t *za_orig; static uint8_t *za_dump; static uint8_t *za_save; @@ -82,7 +85,7 @@ longjmp_test (void) FAIL_EXIT1 ("svcr != 0: %lu", svcr); set_tpidr2 (&blk); start_za (); - load_za (za_orig); + load_za (za_orig, svl); print_data ("za save space", za_save); p = get_tpidr2 (); @@ -131,7 +134,7 @@ setcontext_test (void) FAIL_EXIT1 ("svcr != 0: %lu", svcr); set_tpidr2 (&blk); start_za (); - load_za (za_orig); + load_za (za_orig, svl); print_data ("za save space", za_save); p = get_tpidr2 (); diff --git a/sysdeps/aarch64/tst-sme-signal.c b/sysdeps/aarch64/tst-sme-signal.c new file mode 100644 index 0000000000..b4b07bcc44 --- /dev/null +++ b/sysdeps/aarch64/tst-sme-signal.c @@ -0,0 +1,115 @@ +/* Test handling of SME state in a signal handler. + Copyright (C) 2025 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + . */ + +#include "tst-sme-skeleton.c" + +#include + +static struct _aarch64_ctx * +extension (void *p) +{ + return p; +} + +#ifndef TPIDR2_MAGIC +#define TPIDR2_MAGIC 0x54504902 +#endif + +#ifndef ZA_MAGIC +#define ZA_MAGIC 0x54366345 +#endif + +#ifndef ZT_MAGIC +#define ZT_MAGIC 0x5a544e01 +#endif + +#ifndef EXTRA_MAGIC +#define EXTRA_MAGIC 0x45585401 +#endif + +/* We use a pipe to make sure that the final check of the SME state + happens after signal handler finished. */ +static int pipefd[2]; + +#define WRITE(msg) xwrite (1, msg, sizeof (msg)); + +static void +handler (int signo, siginfo_t *si, void *ctx) +{ + TEST_VERIFY (signo == SIGUSR1); + WRITE ("in the handler\n"); + check_sme_za_state ("during signal", true /* State is clear. */); + ucontext_t *uc = ctx; + void *p = uc->uc_mcontext.__reserved; + unsigned int found = 0; + uint32_t m; + while ((m = extension (p)->magic)) + { + if (m == TPIDR2_MAGIC) + { + WRITE ("found TPIDR2_MAGIC\n"); + found += 1; + } + if (m == ZA_MAGIC) + { + WRITE ("found ZA_MAGIC\n"); + found += 1; + } + if (m == ZT_MAGIC) + { + WRITE ("found ZT_MAGIC\n"); + found += 1; + } + if (m == EXTRA_MAGIC) + { + WRITE ("found EXTRA_MAGIC\n"); + struct { struct _aarch64_ctx h; uint64_t data; } *e = p; + p = (char *)e->data; + continue; + } + p = (char *)p + extension (p)->size; + } + TEST_COMPARE (found, 3); + + /* Signal that the wait is over (see below). */ + char message = '\0'; + xwrite (pipefd[1], &message, 1); +} + +static void +run (struct blk *blk) +{ + xpipe (pipefd); + + struct sigaction sigact; + sigemptyset (&sigact.sa_mask); + sigact.sa_flags = 0; + sigact.sa_flags |= SA_SIGINFO; + sigact.sa_sigaction = handler; + xsigaction (SIGUSR1, &sigact, NULL); + + enable_sme_za_state (blk); + check_sme_za_state ("before signal", false /* State is not clear. */); + xraise (SIGUSR1); + + /* Wait for signal handler to complete. */ + char response; + xread (pipefd[0], &response, 1); + + check_sme_za_state ("after signal", false /* State is not clear. */); +} diff --git a/sysdeps/aarch64/tst-sme-skeleton.c b/sysdeps/aarch64/tst-sme-skeleton.c new file mode 100644 index 0000000000..ba84dda1cb --- /dev/null +++ b/sysdeps/aarch64/tst-sme-skeleton.c @@ -0,0 +1,101 @@ +/* Template for SME tests. + Copyright (C) 2025 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + . */ + +#include +#include +#include +#include +#include + +#include +#include +#include +#include +#include + +#include "tst-sme-helper.h" + +/* Streaming SVE vector register size. */ +static unsigned long svl; + +static uint8_t *state; + +static void +enable_sme_za_state (struct blk *blk) +{ + start_za (); + set_tpidr2 (blk); + load_za (blk, svl); +} + +/* Check if SME state is disabled (when CLEAR is true) or + enabled (when CLEAR is false). */ +static void +check_sme_za_state (const char msg[], bool clear) +{ + unsigned long svcr = get_svcr (); + void *tpidr2 = get_tpidr2 (); + printf ("[%s]\n", msg); + printf ("svcr = %016lx\n", svcr); + printf ("tpidr2 = %016lx\n", (unsigned long)tpidr2); + if (clear) + { + TEST_VERIFY (svcr == 0); + TEST_VERIFY (tpidr2 == NULL); + } + else + { + TEST_VERIFY (svcr != 0); + TEST_VERIFY (tpidr2 != NULL); + } +} + +/* Should be defined in actual test that includes this + skeleton file. */ +static void +run (struct blk *ptr); + +static int +do_test (void) +{ + unsigned long hwcap2 = getauxval (AT_HWCAP2); + if ((hwcap2 & HWCAP2_SME) == 0) + return EXIT_UNSUPPORTED; + + /* Get current streaming SVE vector length in bytes. */ + svl = get_svl (); + printf ("svl: %lu\n", svl); + + TEST_VERIFY_EXIT (!(svl < 16 || svl % 16 != 0 || svl >= (1 << 16))); + + /* Initialise buffer for ZA state of SME. */ + state = xmalloc (svl * svl); + memset (state, 1, svl * svl); + struct blk blk = { + .za_save_buffer = state, + .num_za_save_slices = svl, + .__reserved = {0}, + }; + + run (&blk); + + free (state); + return 0; +} + +#include diff --git a/sysdeps/aarch64/tst-sme-vfork.c b/sysdeps/aarch64/tst-sme-vfork.c new file mode 100644 index 0000000000..3feea065e5 --- /dev/null +++ b/sysdeps/aarch64/tst-sme-vfork.c @@ -0,0 +1,43 @@ +/* Test that ZA state of SME is cleared in both parent and child + when vfork() function is used. + Copyright (C) 2025 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + . */ + +#include "tst-sme-skeleton.c" + +static void +run (struct blk *blk) +{ + /* Enabled ZA state so that effect of disabling be observable. */ + enable_sme_za_state (blk); + check_sme_za_state ("before vfork", /* Clear. */ false); + fflush (stdout); + + pid_t pid = vfork (); + + if (pid == 0) + { + /* Check that ZA state of SME was disabled in child. */ + check_sme_za_state ("after vfork in child", /* Clear. */ true); + _exit (0); + } + + /* Check that ZA state of SME was disabled in parent. */ + check_sme_za_state ("after vfork in parent", /* Clear. */ true); + + TEST_VERIFY (xwaitpid (pid, NULL, 0) == pid); +} diff --git a/sysdeps/aarch64/tst-sme-za-state.c b/sysdeps/aarch64/tst-sme-za-state.c index 63f6eebeb4..00118ef506 100644 --- a/sysdeps/aarch64/tst-sme-za-state.c +++ b/sysdeps/aarch64/tst-sme-za-state.c @@ -16,47 +16,9 @@ License along with the GNU C Library; if not, see . */ -#include -#include -#include -#include -#include - -#include -#include -#include - -#include "tst-sme-helper.h" - -static uint8_t *state; - -static void -enable_sme_za_state (struct blk *ptr) -{ - set_tpidr2 (ptr); - start_za (); - load_za (state); -} +#include "tst-sme-skeleton.c" -static void -check_sme_za_state (const char msg[], bool clear) -{ - unsigned long svcr = get_svcr (); - void *tpidr2 = get_tpidr2 (); - printf ("[%s]\n", msg); - printf ("svcr = %016lx\n", svcr); - printf ("tpidr2 = %016lx\n", (unsigned long)tpidr2); - if (clear) - { - TEST_VERIFY (svcr == 0); - TEST_VERIFY (tpidr2 == NULL); - } - else - { - TEST_VERIFY (svcr != 0); - TEST_VERIFY (tpidr2 != NULL); - } -} +#include static void run (struct blk *ptr) @@ -88,32 +50,3 @@ run (struct blk *ptr) TEST_COMPARE (ret, 42); check_sme_za_state ("after longjmp", /* Clear. */ true); } - -static int -do_test (void) -{ - unsigned long hwcap2 = getauxval (AT_HWCAP2); - if ((hwcap2 & HWCAP2_SME) == 0) - return EXIT_UNSUPPORTED; - - /* Get current streaming SVE vector register size. */ - svl = get_svl (); - printf ("svl: %lu\n", svl); - TEST_VERIFY_EXIT (!(svl < 16 || svl % 16 != 0 || svl >= (1 << 16))); - - /* Initialise buffer for ZA state of SME. */ - state = xmalloc (svl * svl); - memset (state, 1, svl * svl); - struct blk blk = { - .za_save_buffer = state, - .num_za_save_slices = svl, - .__reserved = {0}, - }; - - run (&blk); - - free (state); - return 0; -} - -#include diff --git a/sysdeps/i386/Makefile b/sysdeps/i386/Makefile index ee6470d78e..c0c017b899 100644 --- a/sysdeps/i386/Makefile +++ b/sysdeps/i386/Makefile @@ -60,6 +60,15 @@ $(objpfx)tst-ld-sse-use.out: ../sysdeps/i386/tst-ld-sse-use.sh $(objpfx)ld.so @echo "Checking ld.so for SSE register use. This will take a few seconds..." $(BASH) $< $(objpfx) '$(NM)' '$(OBJDUMP)' '$(READELF)' > $@; \ $(evaluate-test) + +tests-special += $(objpfx)check-gnu-tls.out + +$(objpfx)check-gnu-tls.out: $(common-objpfx)libc.so + LC_ALL=C $(READELF) -V -W $< \ + | sed -ne '/.gnu.version_d/, /.gnu.version_r/ p' \ + | grep GLIBC_ABI_GNU_TLS > $@; \ + $(evaluate-test) +generated += check-gnu-tls.out else CFLAGS-.os += $(if $(filter rtld-%.os,$(@F)), $(rtld-CFLAGS)) endif diff --git a/sysdeps/i386/Versions b/sysdeps/i386/Versions index 36e23b466a..9c84c8ef04 100644 --- a/sysdeps/i386/Versions +++ b/sysdeps/i386/Versions @@ -28,6 +28,11 @@ libc { GLIBC_2.13 { __fentry__; } + GLIBC_ABI_GNU_TLS { + # This symbol is used only for empty version map and will be removed + # by scripts/versions.awk. + __placeholder_only_for_empty_version_map; + } } libm { GLIBC_2.1 { diff --git a/sysdeps/powerpc/powerpc64/le/power10/strcmp.S b/sysdeps/powerpc/powerpc64/le/power10/strcmp.S new file mode 100644 index 0000000000..0d4a53317c --- /dev/null +++ b/sysdeps/powerpc/powerpc64/le/power10/strcmp.S @@ -0,0 +1,185 @@ +/* Optimized strcmp implementation for PowerPC64/POWER10. + Copyright (C) 2025 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + . */ +#include + +#ifndef STRCMP +# define STRCMP strcmp +#endif + +/* Implements the function + int [r3] strcmp (const char *s1 [r3], const char *s2 [r4]). */ + + +#define COMPARE_16(vreg1,vreg2,offset) \ + lxv vreg1+32,offset(r3); \ + lxv vreg2+32,offset(r4); \ + vcmpnezb. v7,vreg1,vreg2; \ + bne cr6,L(different); \ + +#define COMPARE_32(vreg1,vreg2,offset,label1,label2) \ + lxvp vreg1+32,offset(r3); \ + lxvp vreg2+32,offset(r4); \ + vcmpnezb. v7,vreg1+1,vreg2+1; \ + bne cr6,L(label1); \ + vcmpnezb. v7,vreg1,vreg2; \ + bne cr6,L(label2); \ + +#define TAIL(vreg1,vreg2) \ + vctzlsbb r6,v7; \ + vextubrx r5,r6,vreg1; \ + vextubrx r4,r6,vreg2; \ + subf r3,r4,r5; \ + blr; \ + +#define CHECK_N_BYTES(reg1,reg2,len_reg) \ + sldi r0,len_reg,56; \ + lxvl 32+v4,reg1,r0; \ + lxvl 32+v5,reg2,r0; \ + add reg1,reg1,len_reg; \ + add reg2,reg2,len_reg; \ + vcmpnezb. v7,v4,v5; \ + vctzlsbb r6,v7; \ + cmpld cr7,r6,len_reg; \ + blt cr7,L(different); \ + + + .machine power10 +ENTRY_TOCLESS (STRCMP, 4) + li r11,16 + /* eq bit of cr1 used as swap status flag to indicate if + source pointers were swapped. */ + crclr 4*cr1+eq + andi. r7,r3,15 + sub r7,r11,r7 /* r7(nalign1) = 16 - (str1 & 15). */ + andi. r9,r4,15 + sub r5,r11,r9 /* r5(nalign2) = 16 - (str2 & 15). */ + cmpld cr7,r7,r5 + beq cr7,L(same_aligned) + blt cr7,L(nalign1_min) + /* Swap r3 and r4, and r7 and r5 such that r3 and r7 hold the + pointer which is closer to the next 16B boundary so that only + one CHECK_N_BYTES is needed before entering the loop below. */ + mr r8,r4 + mr r4,r3 + mr r3,r8 + mr r12,r7 + mr r7,r5 + mr r5,r12 + crset 4*cr1+eq /* Set bit on swapping source pointers. */ + + .p2align 5 +L(nalign1_min): + CHECK_N_BYTES(r3,r4,r7) + + .p2align 5 +L(s1_aligned): + /* r9 and r5 is number of bytes to be read after and before + page boundary correspondingly. */ + sub r5,r5,r7 + subfic r9,r5,16 + /* Now let r7 hold the count of quadwords which can be + checked without crossing a page boundary. quadword offset is + (str2>>4)&0xFF. */ + rlwinm r7,r4,28,0xFF + /* Below check is required only for first iteration. For second + iteration and beyond, the new loop counter is always 255. */ + cmpldi r7,255 + beq L(L3) + /* Get the initial loop count by 255-((str2>>4)&0xFF). */ + subfic r11,r7,255 + + .p2align 5 +L(L1): + mtctr r11 + + .p2align 5 +L(L2): + COMPARE_16(v4,v5,0) /* Load 16B blocks using lxv. */ + addi r3,r3,16 + addi r4,r4,16 + bdnz L(L2) + /* Cross the page boundary of s2, carefully. */ + + .p2align 5 +L(L3): + CHECK_N_BYTES(r3,r4,r5) + CHECK_N_BYTES(r3,r4,r9) + li r11,255 /* Load the new loop counter. */ + b L(L1) + + .p2align 5 +L(same_aligned): + CHECK_N_BYTES(r3,r4,r7) + /* Align s1 to 32B and adjust s2 address. + Use lxvp only if both s1 and s2 are 32B aligned. */ + COMPARE_16(v4,v5,0) + COMPARE_16(v4,v5,16) + COMPARE_16(v4,v5,32) + COMPARE_16(v4,v5,48) + addi r3,r3,64 + addi r4,r4,64 + COMPARE_16(v4,v5,0) + COMPARE_16(v4,v5,16) + + clrldi r6,r3,59 + subfic r5,r6,32 + add r3,r3,r5 + add r4,r4,r5 + andi. r5,r4,0x1F + beq cr0,L(32B_aligned_loop) + + .p2align 5 +L(16B_aligned_loop): + COMPARE_16(v4,v5,0) + COMPARE_16(v4,v5,16) + COMPARE_16(v4,v5,32) + COMPARE_16(v4,v5,48) + addi r3,r3,64 + addi r4,r4,64 + b L(16B_aligned_loop) + + /* Calculate and return the difference. */ +L(different): + vctzlsbb r6,v7 + vextubrx r5,r6,v4 + vextubrx r4,r6,v5 + bt 4*cr1+eq,L(swapped) + subf r3,r4,r5 + blr + + /* If src pointers were swapped, then swap the + indices and calculate the return value. */ +L(swapped): + subf r3,r5,r4 + blr + + .p2align 5 +L(32B_aligned_loop): + COMPARE_32(v14,v16,0,tail1,tail2) + COMPARE_32(v14,v16,32,tail1,tail2) + COMPARE_32(v14,v16,64,tail1,tail2) + COMPARE_32(v14,v16,96,tail1,tail2) + addi r3,r3,128 + addi r4,r4,128 + b L(32B_aligned_loop) + +L(tail1): TAIL(v15,v17) +L(tail2): TAIL(v14,v16) + +END (STRCMP) +libc_hidden_builtin_def (strcmp) diff --git a/sysdeps/powerpc/powerpc64/le/power10/strlen.S b/sysdeps/powerpc/powerpc64/le/power10/strlen.S index ec644d5bff..29a5a7d960 100644 --- a/sysdeps/powerpc/powerpc64/le/power10/strlen.S +++ b/sysdeps/powerpc/powerpc64/le/power10/strlen.S @@ -31,7 +31,7 @@ # define FUNCNAME RAWMEMCHR # endif # define MCOUNT_NARGS 2 -# define VREG_ZERO v20 +# define VREG_ZERO v17 # define OFF_START_LOOP 256 # define RAWMEMCHR_SUBTRACT_VECTORS \ vsububm v4,v4,v18; \ diff --git a/sysdeps/powerpc/powerpc64/le/power10/strncmp.S b/sysdeps/powerpc/powerpc64/le/power10/strncmp.S new file mode 100644 index 0000000000..6e09fcb7f2 --- /dev/null +++ b/sysdeps/powerpc/powerpc64/le/power10/strncmp.S @@ -0,0 +1,252 @@ +/* Optimized strncmp implementation for PowerPC64/POWER10. + Copyright (C) 2025 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + . */ + +#include + +/* Implements the function + + int [r3] strncmp (const char *s1 [r3], const char *s2 [r4], size_t [r5] n) + + The implementation uses unaligned doubleword access to avoid specialized + code paths depending of data alignment for first 32 bytes and uses + vectorised loops after that. */ + +#ifndef STRNCMP +# define STRNCMP strncmp +#endif + +#define COMPARE_16(vreg1,vreg2,offset) \ + lxv vreg1+32,offset(r3); \ + lxv vreg2+32,offset(r4); \ + vcmpnezb. v7,vreg1,vreg2; \ + bne cr6,L(different); \ + cmpldi cr7,r5,16; \ + ble cr7,L(ret0); \ + addi r5,r5,-16; + +#define COMPARE_32(vreg1,vreg2,offset,label1,label2) \ + lxvp vreg1+32,offset(r3); \ + lxvp vreg2+32,offset(r4); \ + vcmpnezb. v7,vreg1+1,vreg2+1; \ + bne cr6,L(label1); \ + vcmpnezb. v7,vreg1,vreg2; \ + bne cr6,L(label2); \ + cmpldi cr7,r5,32; \ + ble cr7,L(ret0); \ + addi r5,r5,-32; + +#define TAIL_FIRST_16B(vreg1,vreg2) \ + vctzlsbb r6,v7; \ + cmpld cr7,r5,r6; \ + ble cr7,L(ret0); \ + vextubrx r5,r6,vreg1; \ + vextubrx r4,r6,vreg2; \ + subf r3,r4,r5; \ + blr; + +#define TAIL_SECOND_16B(vreg1,vreg2) \ + vctzlsbb r6,v7; \ + addi r0,r6,16; \ + cmpld cr7,r5,r0; \ + ble cr7,L(ret0); \ + vextubrx r5,r6,vreg1; \ + vextubrx r4,r6,vreg2; \ + subf r3,r4,r5; \ + blr; + +#define CHECK_N_BYTES(reg1,reg2,len_reg) \ + sldi r6,len_reg,56; \ + lxvl 32+v4,reg1,r6; \ + lxvl 32+v5,reg2,r6; \ + add reg1,reg1,len_reg; \ + add reg2,reg2,len_reg; \ + vcmpnezb v7,v4,v5; \ + vctzlsbb r6,v7; \ + cmpld cr7,r6,len_reg; \ + blt cr7,L(different); \ + cmpld cr7,r5,len_reg; \ + ble cr7,L(ret0); \ + sub r5,r5,len_reg; \ + + .machine power10 +ENTRY_TOCLESS (STRNCMP, 4) + /* Check if size is 0. */ + cmpdi cr0,r5,0 + beq cr0,L(ret0) + andi. r7,r3,4095 + andi. r8,r4,4095 + cmpldi cr0,r7,4096-16 + cmpldi cr1,r8,4096-16 + bgt cr0,L(crosses) + bgt cr1,L(crosses) + COMPARE_16(v4,v5,0) + addi r3,r3,16 + addi r4,r4,16 + +L(crosses): + andi. r7,r3,15 + subfic r7,r7,16 /* r7(nalign1) = 16 - (str1 & 15). */ + andi. r9,r4,15 + subfic r8,r9,16 /* r8(nalign2) = 16 - (str2 & 15). */ + cmpld cr7,r7,r8 + beq cr7,L(same_aligned) + blt cr7,L(nalign1_min) + + /* nalign2 is minimum and s2 pointer is aligned. */ + CHECK_N_BYTES(r3,r4,r8) + /* Are we on the 64B hunk which crosses a page? */ + andi. r10,r3,63 /* Determine offset into 64B hunk. */ + andi. r8,r3,15 /* The offset into the 16B hunk. */ + neg r7,r3 + andi. r9,r7,15 /* Number of bytes after a 16B cross. */ + rlwinm. r7,r7,26,0x3F /* ((r4-4096))>>6&63. */ + beq L(compare_64_pagecross) + mtctr r7 + b L(compare_64B_unaligned) + + /* nalign1 is minimum and s1 pointer is aligned. */ +L(nalign1_min): + CHECK_N_BYTES(r3,r4,r7) + /* Are we on the 64B hunk which crosses a page? */ + andi. r10,r4,63 /* Determine offset into 64B hunk. */ + andi. r8,r4,15 /* The offset into the 16B hunk. */ + neg r7,r4 + andi. r9,r7,15 /* Number of bytes after a 16B cross. */ + rlwinm. r7,r7,26,0x3F /* ((r4-4096))>>6&63. */ + beq L(compare_64_pagecross) + mtctr r7 + + .p2align 5 +L(compare_64B_unaligned): + COMPARE_16(v4,v5,0) + COMPARE_16(v4,v5,16) + COMPARE_16(v4,v5,32) + COMPARE_16(v4,v5,48) + addi r3,r3,64 + addi r4,r4,64 + bdnz L(compare_64B_unaligned) + + /* Cross the page boundary of s2, carefully. Only for first + iteration we have to get the count of 64B blocks to be checked. + From second iteration and beyond, loop counter is always 63. */ +L(compare_64_pagecross): + li r11, 63 + mtctr r11 + cmpldi r10,16 + ble L(cross_4) + cmpldi r10,32 + ble L(cross_3) + cmpldi r10,48 + ble L(cross_2) +L(cross_1): + CHECK_N_BYTES(r3,r4,r9) + CHECK_N_BYTES(r3,r4,r8) + COMPARE_16(v4,v5,0) + COMPARE_16(v4,v5,16) + COMPARE_16(v4,v5,32) + addi r3,r3,48 + addi r4,r4,48 + b L(compare_64B_unaligned) +L(cross_2): + COMPARE_16(v4,v5,0) + addi r3,r3,16 + addi r4,r4,16 + CHECK_N_BYTES(r3,r4,r9) + CHECK_N_BYTES(r3,r4,r8) + COMPARE_16(v4,v5,0) + COMPARE_16(v4,v5,16) + addi r3,r3,32 + addi r4,r4,32 + b L(compare_64B_unaligned) +L(cross_3): + COMPARE_16(v4,v5,0) + COMPARE_16(v4,v5,16) + addi r3,r3,32 + addi r4,r4,32 + CHECK_N_BYTES(r3,r4,r9) + CHECK_N_BYTES(r3,r4,r8) + COMPARE_16(v4,v5,0) + addi r3,r3,16 + addi r4,r4,16 + b L(compare_64B_unaligned) +L(cross_4): + COMPARE_16(v4,v5,0) + COMPARE_16(v4,v5,16) + COMPARE_16(v4,v5,32) + addi r3,r3,48 + addi r4,r4,48 + CHECK_N_BYTES(r3,r4,r9) + CHECK_N_BYTES(r3,r4,r8) + b L(compare_64B_unaligned) + +L(same_aligned): + CHECK_N_BYTES(r3,r4,r7) + /* Align s1 to 32B and adjust s2 address. + Use lxvp only if both s1 and s2 are 32B aligned. */ + COMPARE_16(v4,v5,0) + COMPARE_16(v4,v5,16) + COMPARE_16(v4,v5,32) + COMPARE_16(v4,v5,48) + addi r3,r3,64 + addi r4,r4,64 + COMPARE_16(v4,v5,0) + COMPARE_16(v4,v5,16) + addi r5,r5,32 + + clrldi r6,r3,59 + subfic r7,r6,32 + add r3,r3,r7 + add r4,r4,r7 + subf r5,r7,r5 + andi. r7,r4,0x1F + beq cr0,L(32B_aligned_loop) + + .p2align 5 +L(16B_aligned_loop): + COMPARE_16(v4,v5,0) + COMPARE_16(v4,v5,16) + COMPARE_16(v4,v5,32) + COMPARE_16(v4,v5,48) + addi r3,r3,64 + addi r4,r4,64 + b L(16B_aligned_loop) + + /* Calculate and return the difference. */ +L(different): + TAIL_FIRST_16B(v4,v5) + + .p2align 5 +L(32B_aligned_loop): + COMPARE_32(v14,v16,0,tail1,tail2) + COMPARE_32(v14,v16,32,tail1,tail2) + COMPARE_32(v14,v16,64,tail1,tail2) + COMPARE_32(v14,v16,96,tail1,tail2) + addi r3,r3,128 + addi r4,r4,128 + b L(32B_aligned_loop) + +L(tail1): TAIL_FIRST_16B(v15,v17) +L(tail2): TAIL_SECOND_16B(v14,v16) + + .p2align 5 +L(ret0): + li r3,0 + blr + +END(STRNCMP) +libc_hidden_builtin_def(strncmp) diff --git a/sysdeps/powerpc/powerpc64/multiarch/Makefile b/sysdeps/powerpc/powerpc64/multiarch/Makefile index e321ce54e0..c9178223a8 100644 --- a/sysdeps/powerpc/powerpc64/multiarch/Makefile +++ b/sysdeps/powerpc/powerpc64/multiarch/Makefile @@ -32,7 +32,7 @@ sysdep_routines += memcpy-power8-cached memcpy-power7 memcpy-a2 memcpy-power6 \ ifneq (,$(filter %le,$(config-machine))) sysdep_routines += memcmp-power10 memcpy-power10 memmove-power10 memset-power10 \ rawmemchr-power9 rawmemchr-power10 \ - strcmp-power9 strncmp-power9 \ + strcmp-power9 strcmp-power10 strncmp-power9 strncmp-power10 \ strcpy-power9 strcat-power10 stpcpy-power9 \ strlen-power9 strncpy-power9 stpncpy-power9 strlen-power10 endif diff --git a/sysdeps/powerpc/powerpc64/multiarch/ifunc-impl-list.c b/sysdeps/powerpc/powerpc64/multiarch/ifunc-impl-list.c index 016d05fd16..f2b9cccde3 100644 --- a/sysdeps/powerpc/powerpc64/multiarch/ifunc-impl-list.c +++ b/sysdeps/powerpc/powerpc64/multiarch/ifunc-impl-list.c @@ -164,6 +164,9 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/powerpc/powerpc64/multiarch/strncmp.c. */ IFUNC_IMPL (i, name, strncmp, #ifdef __LITTLE_ENDIAN__ + IFUNC_IMPL_ADD (array, i, strncmp, hwcap2 & PPC_FEATURE2_ARCH_3_1 + && hwcap & PPC_FEATURE_HAS_VSX, + __strncmp_power10) IFUNC_IMPL_ADD (array, i, strncmp, hwcap2 & PPC_FEATURE2_ARCH_3_00 && hwcap & PPC_FEATURE_HAS_ALTIVEC, __strncmp_power9) @@ -366,6 +369,10 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, /* Support sysdeps/powerpc/powerpc64/multiarch/strcmp.c. */ IFUNC_IMPL (i, name, strcmp, #ifdef __LITTLE_ENDIAN__ + IFUNC_IMPL_ADD (array, i, strcmp, + (hwcap2 & PPC_FEATURE2_ARCH_3_1) + && (hwcap & PPC_FEATURE_HAS_VSX), + __strcmp_power10) IFUNC_IMPL_ADD (array, i, strcmp, hwcap2 & PPC_FEATURE2_ARCH_3_00 && hwcap & PPC_FEATURE_HAS_ALTIVEC, diff --git a/sysdeps/powerpc/powerpc64/multiarch/strcmp-power10.S b/sysdeps/powerpc/powerpc64/multiarch/strcmp-power10.S new file mode 100644 index 0000000000..a4ee7fb53c --- /dev/null +++ b/sysdeps/powerpc/powerpc64/multiarch/strcmp-power10.S @@ -0,0 +1,26 @@ +/* Optimized strcmp implementation for POWER10/PPC64. + Copyright (C) 2025 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + . */ + +#if defined __LITTLE_ENDIAN__ && IS_IN (libc) +#define STRCMP __strcmp_power10 + +#undef libc_hidden_builtin_def +#define libc_hidden_builtin_def(name) + +#include +#endif /* __LITTLE_ENDIAN__ && IS_IN (libc) */ diff --git a/sysdeps/powerpc/powerpc64/multiarch/strcmp.c b/sysdeps/powerpc/powerpc64/multiarch/strcmp.c index 7c77c084a7..3c636e3bbc 100644 --- a/sysdeps/powerpc/powerpc64/multiarch/strcmp.c +++ b/sysdeps/powerpc/powerpc64/multiarch/strcmp.c @@ -29,12 +29,16 @@ extern __typeof (strcmp) __strcmp_power7 attribute_hidden; extern __typeof (strcmp) __strcmp_power8 attribute_hidden; # ifdef __LITTLE_ENDIAN__ extern __typeof (strcmp) __strcmp_power9 attribute_hidden; +extern __typeof (strcmp) __strcmp_power10 attribute_hidden; # endif # undef strcmp libc_ifunc_redirected (__redirect_strcmp, strcmp, # ifdef __LITTLE_ENDIAN__ + (hwcap2 & PPC_FEATURE2_ARCH_3_1 + && hwcap & PPC_FEATURE_HAS_VSX) + ? __strcmp_power10 : (hwcap2 & PPC_FEATURE2_ARCH_3_00 && hwcap & PPC_FEATURE_HAS_ALTIVEC) ? __strcmp_power9 : diff --git a/sysdeps/powerpc/powerpc64/multiarch/strncmp-power10.S b/sysdeps/powerpc/powerpc64/multiarch/strncmp-power10.S new file mode 100644 index 0000000000..bb25bc75b8 --- /dev/null +++ b/sysdeps/powerpc/powerpc64/multiarch/strncmp-power10.S @@ -0,0 +1,25 @@ +/* Copyright (C) 2025 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + . */ + +#if defined __LITTLE_ENDIAN__ && IS_IN (libc) +#define STRNCMP __strncmp_power10 + +#undef libc_hidden_builtin_def +#define libc_hidden_builtin_def(name) + +#include +#endif diff --git a/sysdeps/powerpc/powerpc64/multiarch/strncmp.c b/sysdeps/powerpc/powerpc64/multiarch/strncmp.c index 4cfe27fa45..0a664a620d 100644 --- a/sysdeps/powerpc/powerpc64/multiarch/strncmp.c +++ b/sysdeps/powerpc/powerpc64/multiarch/strncmp.c @@ -29,6 +29,7 @@ extern __typeof (strncmp) __strncmp_ppc attribute_hidden; extern __typeof (strncmp) __strncmp_power8 attribute_hidden; # ifdef __LITTLE_ENDIAN__ extern __typeof (strncmp) __strncmp_power9 attribute_hidden; +extern __typeof (strncmp) __strncmp_power10 attribute_hidden; # endif # undef strncmp @@ -36,6 +37,9 @@ extern __typeof (strncmp) __strncmp_power9 attribute_hidden; ifunc symbol properly. */ libc_ifunc_redirected (__redirect_strncmp, strncmp, # ifdef __LITTLE_ENDIAN__ + (hwcap2 & PPC_FEATURE2_ARCH_3_1 + && hwcap & PPC_FEATURE_HAS_VSX) + ? __strncmp_power10 : (hwcap2 & PPC_FEATURE2_ARCH_3_00 && hwcap & PPC_FEATURE_HAS_ALTIVEC) ? __strncmp_power9 : diff --git a/sysdeps/pthread/tst-fopen-threaded.c b/sysdeps/pthread/tst-fopen-threaded.c index ade58ad19e..c17f1eaa13 100644 --- a/sysdeps/pthread/tst-fopen-threaded.c +++ b/sysdeps/pthread/tst-fopen-threaded.c @@ -34,11 +34,13 @@ #include #include #include +#include #include #include #include #include +#include #define NUM_THREADS 100 #define ITERS 10 @@ -111,7 +113,8 @@ threadOpenCloseRoutine (void *argv) /* Wait for all threads to be ready to call fopen and fclose. */ xpthread_barrier_wait (&barrier); - FILE *fd = xfopen ("/tmp/openclosetest", "w+"); + char *file = (char *) argv; + FILE *fd = xfopen (file, "w+"); xfclose (fd); return NULL; } @@ -235,6 +238,10 @@ do_test (void) xfclose (fd_file); } + char *tempdir = support_create_temp_directory ("openclosetest-"); + char *file = xasprintf ("%s/file", tempdir); + add_temp_file (file); + /* Test 3: Concurrent open/close. */ for (int reps = 1; reps <= ITERS; reps++) { @@ -243,7 +250,7 @@ do_test (void) { threads[i] = xpthread_create (support_small_stack_thread_attribute (), - threadOpenCloseRoutine, NULL); + threadOpenCloseRoutine, file); } for (int i = 0; i < NUM_THREADS; i++) { @@ -252,6 +259,9 @@ do_test (void) xpthread_barrier_destroy (&barrier); } + free (file); + free (tempdir); + return 0; } diff --git a/sysdeps/unix/sysv/linux/aarch64/clone.S b/sysdeps/unix/sysv/linux/aarch64/clone.S index 40015c6933..53f1efd728 100644 --- a/sysdeps/unix/sysv/linux/aarch64/clone.S +++ b/sysdeps/unix/sysv/linux/aarch64/clone.S @@ -45,6 +45,9 @@ ENTRY(__clone) and x1, x1, -16 cbz x1, .Lsyscall_error + /* Clear ZA state of SME. */ + CALL_LIBC_ARM_ZA_DISABLE + /* Do the system call. */ /* X0:flags, x1:newsp, x2:parenttidptr, x3:newtls, x4:childtid. */ mov x0, x2 /* flags */ diff --git a/sysdeps/unix/sysv/linux/aarch64/clone3.S b/sysdeps/unix/sysv/linux/aarch64/clone3.S index c9ca845ef2..bc978b7e10 100644 --- a/sysdeps/unix/sysv/linux/aarch64/clone3.S +++ b/sysdeps/unix/sysv/linux/aarch64/clone3.S @@ -46,6 +46,9 @@ ENTRY(__clone3) cbz x10, .Lsyscall_error /* No NULL cl_args pointer. */ cbz x2, .Lsyscall_error /* No NULL function pointer. */ + /* Clear ZA state of SME. */ + CALL_LIBC_ARM_ZA_DISABLE + /* Do the system call, the kernel expects: x8: system call number x0: cl_args diff --git a/sysdeps/unix/sysv/linux/aarch64/setcontext.S b/sysdeps/unix/sysv/linux/aarch64/setcontext.S index d9716f012e..8e98594663 100644 --- a/sysdeps/unix/sysv/linux/aarch64/setcontext.S +++ b/sysdeps/unix/sysv/linux/aarch64/setcontext.S @@ -49,15 +49,7 @@ ENTRY (__setcontext) b C_SYMBOL_NAME (__syscall_error) 1: /* Clear ZA state of SME. */ - /* The calling convention of __libc_arm_za_disable allows to do - this thus allowing to avoid saving to and reading from stack. - As a result we also don't need to sign the return address and - check it after returning because it is not stored to stack. */ - mov x13, x30 - cfi_register (x30, x13) - bl __libc_arm_za_disable - mov x30, x13 - cfi_register (x13, x30) + CALL_LIBC_ARM_ZA_DISABLE /* Restore the general purpose registers. */ mov x0, x9 cfi_def_cfa (x0, 0) diff --git a/sysdeps/unix/sysv/linux/aarch64/sysdep.h b/sysdeps/unix/sysv/linux/aarch64/sysdep.h index f0e8d64eef..8a7690d4a8 100644 --- a/sysdeps/unix/sysv/linux/aarch64/sysdep.h +++ b/sysdeps/unix/sysv/linux/aarch64/sysdep.h @@ -150,6 +150,19 @@ mov x8, SYS_ify (syscall_name); \ svc 0 +/* Clear ZA state of SME (ASM version). */ +/* The __libc_arm_za_disable function has special calling convention + that allows to call it without stack manipulation and preserving + most of the registers. */ + .macro CALL_LIBC_ARM_ZA_DISABLE + cfi_remember_state + mov x13, x30 + cfi_register(x30, x13) + bl __libc_arm_za_disable + mov x30, x13 + cfi_restore_state + .endm + #else /* not __ASSEMBLER__ */ # define VDSO_NAME "LINUX_2.6.39" @@ -230,6 +243,32 @@ #undef HAVE_INTERNAL_BRK_ADDR_SYMBOL #define HAVE_INTERNAL_BRK_ADDR_SYMBOL 1 +/* Clear ZA state of SME (C version). */ +/* The __libc_arm_za_disable function has special calling convention + that allows to call it without stack manipulation and preserving + most of the registers. */ +#define CALL_LIBC_ARM_ZA_DISABLE() \ +({ \ + unsigned long int __tmp; \ + asm volatile ( \ + " .cfi_remember_state\n" \ + " mov %0, x30\n" \ + " .cfi_register x30, %0\n" \ + " bl __libc_arm_za_disable\n" \ + " mov x30, %0\n" \ + " .cfi_restore_state\n" \ + : "=r" (__tmp) \ + : \ + : "x14", "x15", "x16", "x17", "x18", "memory" ); \ +}) + +/* Do clear ZA state of SME before making normal clone syscall. */ +#define INLINE_CLONE_SYSCALL(a0, a1, a2, a3, a4) \ +({ \ + CALL_LIBC_ARM_ZA_DISABLE (); \ + INLINE_SYSCALL_CALL (clone, a0, a1, a2, a3, a4); \ +}) + #endif /* __ASSEMBLER__ */ #endif /* linux/aarch64/sysdep.h */ diff --git a/sysdeps/unix/sysv/linux/aarch64/vfork.S b/sysdeps/unix/sysv/linux/aarch64/vfork.S index d5943a7485..2600bc9be3 100644 --- a/sysdeps/unix/sysv/linux/aarch64/vfork.S +++ b/sysdeps/unix/sysv/linux/aarch64/vfork.S @@ -27,6 +27,9 @@ ENTRY (__vfork) + /* Clear ZA state of SME. */ + CALL_LIBC_ARM_ZA_DISABLE + mov x0, #0x4111 /* CLONE_VM | CLONE_VFORK | SIGCHLD */ mov x1, sp DO_CALL (clone, 2) diff --git a/sysdeps/x86/Makefile b/sysdeps/x86/Makefile index 4fbd48e1c8..9e1c8cce85 100644 --- a/sysdeps/x86/Makefile +++ b/sysdeps/x86/Makefile @@ -135,6 +135,15 @@ LDFLAGS-tst-tls23 += -rdynamic tst-tls23-mod.so-no-z-defs = yes $(objpfx)tst-tls23-mod.so: $(libsupport) + +tests-special += $(objpfx)check-gnu2-tls.out + +$(objpfx)check-gnu2-tls.out: $(common-objpfx)libc.so + LC_ALL=C $(READELF) -V -W $< \ + | sed -ne '/.gnu.version_d/, /.gnu.version_r/ p' \ + | grep GLIBC_ABI_GNU2_TLS > $@; \ + $(evaluate-test) +generated += check-gnu2-tls.out endif ifeq ($(subdir),gmon) diff --git a/sysdeps/x86/Versions b/sysdeps/x86/Versions index 4b10c4b5d7..e8dcfccbe4 100644 --- a/sysdeps/x86/Versions +++ b/sysdeps/x86/Versions @@ -7,4 +7,9 @@ libc { GLIBC_2.33 { __x86_get_cpuid_feature_leaf; } + GLIBC_ABI_GNU2_TLS { + # This symbol is used only for empty version map and will be removed + # by scripts/versions.awk. + __placeholder_only_for_empty_version_map; + } } diff --git a/sysdeps/x86/cpu-features.c b/sysdeps/x86/cpu-features.c index b7d1506135..b67ef541dd 100644 --- a/sysdeps/x86/cpu-features.c +++ b/sysdeps/x86/cpu-features.c @@ -543,6 +543,8 @@ enum intel_microarch INTEL_BIGCORE_PANTHERLAKE, INTEL_BIGCORE_GRANITERAPIDS, INTEL_BIGCORE_DIAMONDRAPIDS, + INTEL_BIGCORE_WILDCATLAKE, + INTEL_BIGCORE_NOVALAKE, /* Mixed (bigcore + atom SOC). */ INTEL_MIXED_LAKEFIELD, @@ -702,6 +704,8 @@ intel_get_fam6_microarch (unsigned int model, return INTEL_BIGCORE_ARROWLAKE; case 0xCC: return INTEL_BIGCORE_PANTHERLAKE; + case 0xD5: + return INTEL_BIGCORE_WILDCATLAKE; case 0xAD: case 0xAE: return INTEL_BIGCORE_GRANITERAPIDS; @@ -818,6 +822,17 @@ disable_tsx: break; } } + else if (family == 18) + switch (model) + { + case 0x01: + case 0x03: + microarch = INTEL_BIGCORE_NOVALAKE; + break; + + default: + break; + } else if (family == 19) switch (model) { @@ -934,6 +949,8 @@ disable_tsx: case INTEL_BIGCORE_LUNARLAKE: case INTEL_BIGCORE_ARROWLAKE: case INTEL_BIGCORE_PANTHERLAKE: + case INTEL_BIGCORE_WILDCATLAKE: + case INTEL_BIGCORE_NOVALAKE: case INTEL_BIGCORE_SAPPHIRERAPIDS: case INTEL_BIGCORE_EMERALDRAPIDS: case INTEL_BIGCORE_GRANITERAPIDS: diff --git a/sysdeps/x86_64/Makefile b/sysdeps/x86_64/Makefile index c3e1065c81..fe9f1cdddb 100644 --- a/sysdeps/x86_64/Makefile +++ b/sysdeps/x86_64/Makefile @@ -212,6 +212,15 @@ LDFLAGS-tst-plt-rewrite2 = -Wl,-z,now LDFLAGS-tst-plt-rewritemod2.so = -Wl,-z,now,-z,undefs tst-plt-rewrite2-ENV = GLIBC_TUNABLES=glibc.cpu.plt_rewrite=2 $(objpfx)tst-plt-rewrite2: $(objpfx)tst-plt-rewritemod2.so + +tests-special += $(objpfx)check-dt-x86-64-plt.out + +$(objpfx)check-dt-x86-64-plt.out: $(common-objpfx)libc.so + LC_ALL=C $(READELF) -V -W $< \ + | sed -ne '/.gnu.version_d/, /.gnu.version_r/ p' \ + | grep GLIBC_ABI_DT_X86_64_PLT > $@; \ + $(evaluate-test) +generated += check-dt-x86-64-plt.out endif test-internal-extras += tst-gnu2-tls2mod1 diff --git a/sysdeps/x86_64/Versions b/sysdeps/x86_64/Versions index e94758b236..6a989ad3b3 100644 --- a/sysdeps/x86_64/Versions +++ b/sysdeps/x86_64/Versions @@ -5,6 +5,11 @@ libc { GLIBC_2.13 { __fentry__; } + GLIBC_ABI_DT_X86_64_PLT { + # This symbol is used only for empty version map and will be removed + # by scripts/versions.awk. + __placeholder_only_for_empty_version_map; + } } libm { GLIBC_2.1 { diff --git a/sysdeps/x86_64/multiarch/ifunc-wmemset.h b/sysdeps/x86_64/multiarch/ifunc-wmemset.h index f95cca6ae5..50af138230 100644 --- a/sysdeps/x86_64/multiarch/ifunc-wmemset.h +++ b/sysdeps/x86_64/multiarch/ifunc-wmemset.h @@ -35,7 +35,7 @@ IFUNC_SELECTOR (void) if (X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, AVX2) && X86_ISA_CPU_FEATURES_ARCH_P (cpu_features, - AVX_Fast_Unaligned_Load, !)) + AVX_Fast_Unaligned_Load,)) { if (X86_ISA_CPU_FEATURE_USABLE_P (cpu_features, AVX512VL)) {