valgrind-developers — Discussion list for developers of valgrind

[Valgrind-developers] [valgrind] s390x: Header file tweak

[Florian K. <fk...@so...>]

https://sourceware.org/git/gitweb.cgi?p=valgrind.git;h=c069c7a8cfbb90f5dbab915da78aa986743b1295

commit c069c7a8cfbb90f5dbab915da78aa986743b1295
Author: Florian Krohm <fl...@ei...>
Date:   Sat Mar 15 16:51:00 2025 +0000

    s390x: Header file tweak
    
    Thou shalt write header files in a way that they are compilable
    entities on their own.

Diff:
---
 VEX/priv/guest_s390_defs.h         | 1 +
 none/tests/s390x/disasm-test/vex.c | 1 -
 2 files changed, 1 insertion(+), 1 deletion(-)

diff --git a/VEX/priv/guest_s390_defs.h b/VEX/priv/guest_s390_defs.h
index a64d563ec9..607773ad14 100644
--- a/VEX/priv/guest_s390_defs.h
+++ b/VEX/priv/guest_s390_defs.h
@@ -34,6 +34,7 @@
 #include "libvex_basictypes.h"        // offsetof
 #include "guest_generic_bb_to_IR.h"   // DisResult
 #include "libvex_guest_s390x.h"       // VexGuestS390XState
+#include "main_util.h"                // STATIC_ASSERT
 
 
 /* Convert one s390 insn to IR.  See the type DisOneInstrFn in
diff --git a/none/tests/s390x/disasm-test/vex.c b/none/tests/s390x/disasm-test/vex.c
index fa7b48b7c3..a1847b42b5 100644
--- a/none/tests/s390x/disasm-test/vex.c
+++ b/none/tests/s390x/disasm-test/vex.c
@@ -24,7 +24,6 @@
 
 #include "libvex_ir.h"        // emptyIRSB
 #include "libvex.h"           // LibVEX_Init
-#include "main_util.h"        // GRRR for guest_s390_defs.h which needs STATIC_ASSERT
 #include "guest_s390_defs.h"  // disInstr_S390
 #include "host_s390_defs.h"   // s390_host_hwcaps
 #include "main_globals.h"     // vex_traceflags

[Valgrind-developers] [valgrind] Bug 501479 - Illumos DRD pthread_mutex_init wrapper errors

[Paul F. <pa...@so...>]

https://sourceware.org/git/gitweb.cgi?p=valgrind.git;h=4d578ef2153133c8480489f2f88a3dc1ffddef85

commit 4d578ef2153133c8480489f2f88a3dc1ffddef85
Author: Paul Floyd <pj...@wa...>
Date:   Fri Mar 14 21:56:35 2025 +0100

    Bug 501479 - Illumos DRD pthread_mutex_init wrapper errors

Diff:
---
 NEWS                         |  3 ++-
 drd/drd_pthread_intercepts.c | 32 ++++++++++++++++++++++++++++++++
 2 files changed, 34 insertions(+), 1 deletion(-)

diff --git a/NEWS b/NEWS
index e6a2855852..4d4c7eec46 100644
--- a/NEWS
+++ b/NEWS
@@ -59,7 +59,8 @@ are not entered into bugzilla tend to get forgotten about or ignored.
 499212  mmap() with MAP_ALIGNED() returns unaligned pointer
 501119  memcheck/tests/pointer-trace fails when run on NFS filesystem
 501194  Fix ML_(check_macho_and_get_rw_loads) so that it is correct for any number of segment commands
-501348 glibc built with -march=x86-64-v3 does not work due to ld.so memcmp
+501348  glibc built with -march=x86-64-v3 does not work due to ld.so memcmp
+501479  Illumos DRD pthread_mutex_init wrapper errors
 
 
 To see details of a given bug, visit
diff --git a/drd/drd_pthread_intercepts.c b/drd/drd_pthread_intercepts.c
index 7f20b0b4c2..bb92101931 100644
--- a/drd/drd_pthread_intercepts.c
+++ b/drd/drd_pthread_intercepts.c
@@ -876,6 +876,33 @@ PTH_FUNCS(int, pthreadZuonce, pthread_once_intercept,
           (pthread_once_t *once_control, void (*init_routine)(void)),
           (once_control, init_routine));
 
+#if defined(VGO_solaris)
+// see https://bugs.kde.org/show_bug.cgi?id=501479
+// temporary (?) workaround
+// Helgrind doesn't have this problem because it only redirects mutex_init
+// and not thread_mutex_init which also uses pthread_mutexattr_gettype
+// maybe DRD should do the same.
+typedef	struct{
+	int	pshared;
+	int	protocol;
+	int	prioceiling;
+	int	type;
+	int	robustness;
+} workaround_mattr_t;
+
+static int
+pthread_mutexattr_gettype_workaround(const pthread_mutexattr_t *attr, int *typep)
+{
+	workaround_mattr_t	*ap;
+
+	if (attr == NULL || (ap = attr->__pthread_mutexattrp) == NULL ||
+	    typep == NULL)
+		return (EINVAL);
+	*typep = ap->type;
+	return (0);
+}
+#endif
+
 static __always_inline
 int pthread_mutex_init_intercept(pthread_mutex_t *mutex,
                                  const pthread_mutexattr_t* attr)
@@ -885,8 +912,13 @@ int pthread_mutex_init_intercept(pthread_mutex_t *mutex,
    int mt;
    VALGRIND_GET_ORIG_FN(fn);
    mt = PTHREAD_MUTEX_DEFAULT;
+#if !defined(VGO_solaris)
    if (attr)
       pthread_mutexattr_gettype(attr, &mt);
+#else
+   if (attr)
+      pthread_mutexattr_gettype_workaround(attr, &mt);
+#endif
    VALGRIND_DO_CLIENT_REQUEST_STMT(VG_USERREQ_DRD_PRE_MUTEX_INIT,
                                    mutex, DRD_(pthread_to_drd_mutex_type)(mt),
                                    0, 0, 0);

[Valgrind-developers] Sourceware Survey 2025

[Mark W. <ma...@kl...>]

The Sourceware Project Leadership Committee would like to know who our
users are, which hosted projects they feel part of, what services they
rely on and what the priorities should be for new initiatives.

https://nextcloud.sfconservancy.org/apps/forms/s/xmGgmJFzSb2FZNd58cXMtAZp

The survey will run till the end of March and contains 20 quick
questions (none of the questions are mandatory, please feel free to
skip any you find not relevant, but any answer you give is helpful.)

The overview page (where we will also publish the summary/results) is
here https://sourceware.org/survey-2025 the direct link to the form is
https://nextcloud.sfconservancy.org/apps/forms/s/xmGgmJFzSb2FZNd58cXMtAZp

Also if you like to discuss this survey or give feedback another way
there is the Sourceware Open Office this Friday 14 March at 16:00 UTC
in #overseers on irc.libera.chat. To get the right time in your local
timezone: $ date -d "Fri Mar 14 16:00 UTC 2025"

Thanks,

 Frank Ch. Eigler, Christopher Faylor, Ian Kelling, Ian Lance Taylor,
 Tom Tromey, Jon Turney, Mark J. Wielaard, Elena Zannoni

[Valgrind-developers] [valgrind] Darwin: close fd after call to ML_(check_macho_and_get_rw_loads)

[Paul F. <pa...@so...>]

https://sourceware.org/git/gitweb.cgi?p=valgrind.git;h=7ba2fe4f7d16f5880035e3d4154f72e60df19eea

commit 7ba2fe4f7d16f5880035e3d4154f72e60df19eea
Author: Paul Floyd <pj...@wa...>
Date:   Thu Mar 13 20:47:28 2025 +0100

    Darwin: close fd after call to ML_(check_macho_and_get_rw_loads)
    
    I missed changing the code so that the VG_(close) happens for all
    platforms, not just non-Darwin.

Diff:
---
 coregrind/m_debuginfo/debuginfo.c | 16 ++++++----------
 1 file changed, 6 insertions(+), 10 deletions(-)

diff --git a/coregrind/m_debuginfo/debuginfo.c b/coregrind/m_debuginfo/debuginfo.c
index 18fd33535c..cd04720902 100644
--- a/coregrind/m_debuginfo/debuginfo.c
+++ b/coregrind/m_debuginfo/debuginfo.c
@@ -1349,26 +1349,22 @@ ULong VG_(di_notify_mmap)( Addr a, Bool allow_SkFileV, Int use_fd )
 
    expected_rw_load_count = 0;
 
+   Bool check_ok = False;
 #if defined(VGO_darwin)
-   if (!ML_(check_macho_and_get_rw_loads)( actual_fd, &expected_rw_load_count ))
-      return 0;
-#endif
-
+   check_ok = ML_(check_macho_and_get_rw_loads)( actual_fd, &expected_rw_load_count );
+#else
    /* We're only interested in mappings of object files. */
-#  if defined(VGO_linux) || defined(VGO_solaris) || defined(VGO_freebsd)
-
-   Bool elf_ok = ML_(check_elf_and_get_rw_loads) ( actual_fd, filename, &expected_rw_load_count, use_fd == -1 );
+   check_ok = ML_(check_elf_and_get_rw_loads) ( actual_fd, filename, &expected_rw_load_count, use_fd == -1 );
+#endif
 
    if (use_fd == -1) {
       VG_(close)( actual_fd );
    }
 
-   if (!elf_ok) {
+   if (!check_ok) {
       return 0;
    }
 
-#  endif
-
    /* See if we have a DebugInfo for this filename.  If not,
       create one. */
    di = find_or_create_DebugInfo_for( filename );

[Valgrind-developers] [valgrind] Darwin: change message for failure to open a binary

[Paul F. <pa...@so...>]

https://sourceware.org/git/gitweb.cgi?p=valgrind.git;h=b65a739f2a63047565ac2d654931c95ae9d44b62

commit b65a739f2a63047565ac2d654931c95ae9d44b62
Author: Paul Floyd <pj...@wa...>
Date:   Thu Mar 13 08:35:17 2025 +0100

    Darwin: change message for failure to open a binary
    
    Previously was always ELF, now either ELF or mach-o.

Diff:
---
 coregrind/m_debuginfo/debuginfo.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/coregrind/m_debuginfo/debuginfo.c b/coregrind/m_debuginfo/debuginfo.c
index 6c4a692656..18fd33535c 100644
--- a/coregrind/m_debuginfo/debuginfo.c
+++ b/coregrind/m_debuginfo/debuginfo.c
@@ -1329,12 +1329,16 @@ ULong VG_(di_notify_mmap)( Addr a, Bool allow_SkFileV, Int use_fd )
       SysRes fd = VG_(open)( filename, oflags, 0 );
       if (sr_isError(fd)) {
          if (sr_Err(fd) != VKI_EACCES) {
+#if defined(VGO_darwin)
+            const HChar* message = "can't open file to inspect mach-o header";
+#else
+            const HChar* message = "can't open file to inspect ELF header";
+#endif
             DebugInfo fake_di;
             VG_(memset)(&fake_di, 0, sizeof(fake_di));
             fake_di.fsm.filename = ML_(dinfo_strdup)("di.debuginfo.nmm",
                                                      filename);
-            ML_(symerr)(&fake_di, True,
-                        "can't open file to inspect ELF header");
+            ML_(symerr)(&fake_di, True, message);
          }
          return 0;
       }

[Valgrind-developers] [valgrind] s390x: Fix BZ 496950

[Florian K. <fk...@so...>]

https://sourceware.org/git/gitweb.cgi?p=valgrind.git;h=744e47c8ff5bd9a7908b63b9d053b32bac18d686

commit 744e47c8ff5bd9a7908b63b9d053b32bac18d686
Author: Florian Krohm <fl...@ei...>
Date:   Wed Mar 12 23:10:06 2025 +0000

    s390x: Fix BZ 496950
    
    Adding a new hardware capability requires changes in several places.
    Generally for a new FOO hardware capability:
    - add VEX_HWCAPS_S390X_FOO and update VEX_HWCAPS_S390X_ALL in libvex.h
    - test the corresponding facility bit in m_machine.c
    - add s390_host_has_foo in host_s390_defs.h
    - if an insn requiring FOO cannot be implemented on the host by other means
      - add EmFail_S390X_foo in libvex_emnote.h
    - handle EmFail_S390X_foo in main_main.c
    - update function show_hwcaps_s390x
    
    Occasionally something gets forgotten and this patch adds the forgotten
    pieces.
    Also add new hardware capabilities vxe2, vxd, msa, msa4, msa8, msa9
    
    Fixes: https://bugs.kde.org/show_bug.cgi?id=496950

Diff:
---
 NEWS                                    |  1 +
 VEX/priv/host_s390_defs.h               | 14 ++++++++++
 VEX/priv/main_main.c                    | 47 ++++++++++++++++++++++++++++++++-
 VEX/pub/libvex.h                        | 20 +++++++++++---
 VEX/pub/libvex_emnote.h                 | 36 +++++++++++++++++++++++--
 coregrind/m_extension/extension-s390x.c |  3 +--
 coregrind/m_machine.c                   |  8 ++++++
 7 files changed, 120 insertions(+), 9 deletions(-)

diff --git a/NEWS b/NEWS
index 7025d61af7..e6a2855852 100644
--- a/NEWS
+++ b/NEWS
@@ -44,6 +44,7 @@ are not entered into bugzilla tend to get forgotten about or ignored.
 495816  s390x: Fix disassembler segfault for C[G]RT and CL[G]RT
 496370  Illumos: signal handling is broken
 496571  False positive for null key passed to bpf_map_get_next_key syscall.
+496950  s390x: Fix hardware capabilities and EmFail codes
 497130  Recognize new DWARF5 DW_LANG constants
 497455  Update drd/scripts/download-and-build-gcc
 497723  Enabling Ada demangling breaks callgrind differentiation between
diff --git a/VEX/priv/host_s390_defs.h b/VEX/priv/host_s390_defs.h
index 957585c396..dcdb6713e9 100644
--- a/VEX/priv/host_s390_defs.h
+++ b/VEX/priv/host_s390_defs.h
@@ -939,6 +939,8 @@ extern UInt s390_host_hwcaps;
                       (s390_host_hwcaps & (VEX_HWCAPS_S390X_VX))
 #define s390_host_has_msa5 \
                       (s390_host_hwcaps & (VEX_HWCAPS_S390X_MSA5))
+#define s390_host_has_mi2 \
+                      (s390_host_hwcaps & (VEX_HWCAPS_S390X_MI2))
 #define s390_host_has_lsc2 \
                       (s390_host_hwcaps & (VEX_HWCAPS_S390X_LSC2))
 #define s390_host_has_vxe \
@@ -947,6 +949,18 @@ extern UInt s390_host_hwcaps;
                       (s390_host_hwcaps & (VEX_HWCAPS_S390X_NNPA))
 #define s390_host_has_dflt \
                       (s390_host_hwcaps & (VEX_HWCAPS_S390X_DFLT))
+#define s390_host_has_vxe2 \
+                      (s390_host_hwcaps & (VEX_HWCAPS_S390X_VXE2))
+#define s390_host_has_vxd \
+                      (s390_host_hwcaps & (VEX_HWCAPS_S390X_VXD))
+#define s390_host_has_msa \
+                      (s390_host_hwcaps & (VEX_HWCAPS_S390X_MSA))
+#define s390_host_has_msa4 \
+                      (s390_host_hwcaps & (VEX_HWCAPS_S390X_MSA4))
+#define s390_host_has_msa8 \
+                      (s390_host_hwcaps & (VEX_HWCAPS_S390X_MSA8))
+#define s390_host_has_msa9 \
+                      (s390_host_hwcaps & (VEX_HWCAPS_S390X_MSA9))
 #endif /* ndef __VEX_HOST_S390_DEFS_H */
 
 /*---------------------------------------------------------------*/
diff --git a/VEX/priv/main_main.c b/VEX/priv/main_main.c
index 91bb3bbbff..80f6d6a6d6 100644
--- a/VEX/priv/main_main.c
+++ b/VEX/priv/main_main.c
@@ -1561,7 +1561,45 @@ const HChar* LibVEX_EmNote_string ( VexEmNote ew )
      case EmFail_S390X_vx:
         return "Encountered an instruction that requires the vector facility.\n"
                "  That facility is not available on this host";
-     default: 
+     case EmFail_S390X_ppno:
+        return "Instruction ppno is not supported on this host.";
+     case EmFail_S390X_vxe:
+        return "Encountered an instruction that requires the vector-extensions"
+               " facility 1.\n"
+               "  That facility is not available on this host";
+     case EmFail_S390X_dflt:
+        return "Encountered an instruction that requires the deflate-conversion"
+               " facility.\n"
+               "  That facility is not available on this host";
+     case EmFail_S390X_nnpa:
+        return "Encountered an instruction that requires the"
+               " neural-network-processing-assist facility.\n"
+               "  That facility is not available on this host";
+     case EmFail_S390X_vxe2:
+        return "Encountered an instruction that requires the vector-extensions\n"
+               " facility 2.\n"
+               "  That facility is not available on this host";
+     case EmFail_S390X_vxd:
+        return "Encountered an instruction that requires the vector-packed-decimal\n"
+               " facility.\n"
+               "  That facility is not available on this host";
+     case EmFail_S390X_msa:
+        return "Encountered an instruction that requires the message-security"
+               " assist.\n"
+               "  That assist is not available on this host";
+     case EmFail_S390X_msa4:
+        return "Encountered an instruction that requires the"
+               " message-security-assist extension 4.\n"
+               "  That extension is not available on this host";
+     case EmFail_S390X_msa8:
+        return "Encountered an instruction that requires the"
+               " message-security-assist extension 8.\n"
+               "  That extension is not available on this host";
+     case EmFail_S390X_msa9:
+        return "Encountered an instruction that requires the"
+               " message-security-assist extension 9.\n"
+               "  That extension is not available on this host";
+     default:
         vpanic("LibVEX_EmNote_string: unknown warning");
    }
 }
@@ -1894,7 +1932,14 @@ static const HChar* show_hwcaps_s390x ( UInt hwcaps )
       { VEX_HWCAPS_S390X_MI2,   "mi2" },
       { VEX_HWCAPS_S390X_LSC2,  "lsc2" },
       { VEX_HWCAPS_S390X_VXE,   "vxe" },
+      { VEX_HWCAPS_S390X_NNPA,  "nnpa" },
       { VEX_HWCAPS_S390X_DFLT,  "dflt" },
+      { VEX_HWCAPS_S390X_VXE2,  "vxe2" },
+      { VEX_HWCAPS_S390X_VXD,   "vxd" },
+      { VEX_HWCAPS_S390X_MSA,   "msa" },
+      { VEX_HWCAPS_S390X_MSA4,  "msa4" },
+      { VEX_HWCAPS_S390X_MSA8,  "msa8" },
+      { VEX_HWCAPS_S390X_MSA9,  "msa9" },
    };
    /* Allocate a large enough buffer */
    static HChar buf[sizeof prefix + 
diff --git a/VEX/pub/libvex.h b/VEX/pub/libvex.h
index d99be23a02..2dbfbe770e 100644
--- a/VEX/pub/libvex.h
+++ b/VEX/pub/libvex.h
@@ -134,7 +134,7 @@ typedef
 /* s390x: Hardware capability encoding
 
    Bits [26:31] encode the machine model (see VEX_S390X_MODEL... below)
-   Bits [0:20]  encode specific hardware capabilities
+   Bits [0:25]  encode specific hardware capabilities
                 (see VEX_HWAPS_S390X_... below)
 */
 
@@ -174,12 +174,18 @@ typedef
 #define VEX_HWCAPS_S390X_LSC   (1<<16)  /* Conditional load/store facility */
 #define VEX_HWCAPS_S390X_PFPO  (1<<17)  /* Perform floating point ops facility */
 #define VEX_HWCAPS_S390X_VX    (1<<18)  /* Vector facility */
-#define VEX_HWCAPS_S390X_MSA5  (1<<19)  /* message security assistance facility */
-#define VEX_HWCAPS_S390X_MI2   (1<<20)  /* miscellaneous-instruction-extensions facility 2 */
+#define VEX_HWCAPS_S390X_MSA5  (1<<19)  /* Message-security-assistance facility 5 */
+#define VEX_HWCAPS_S390X_MI2   (1<<20)  /* Miscellaneous-instruction-extensions facility 2 */
 #define VEX_HWCAPS_S390X_LSC2  (1<<21)  /* Conditional load/store facility2 */
 #define VEX_HWCAPS_S390X_VXE   (1<<22)  /* Vector-enhancements facility */
 #define VEX_HWCAPS_S390X_NNPA  (1<<23)  /* NNPA facility */
 #define VEX_HWCAPS_S390X_DFLT  (1<<24)  /* Deflate-conversion facility */
+#define VEX_HWCAPS_S390X_VXE2  (1<<25)  /* Vector-enhancements facility 2 */
+#define VEX_HWCAPS_S390X_VXD   (1<<26)  /* Vector packed-decimal facility */
+#define VEX_HWCAPS_S390X_MSA   (1<<27)  /* Message-security assist */
+#define VEX_HWCAPS_S390X_MSA4  (1<<28)  /* Message-security-assist extension 4 */
+#define VEX_HWCAPS_S390X_MSA8  (1<<29)  /* Message-security-assist extension 8 */
+#define VEX_HWCAPS_S390X_MSA9  (1<<30)  /* Message-security-assist extension 9 */
 
 /* Special value representing all available s390x hwcaps */
 #define VEX_HWCAPS_S390X_ALL   (VEX_HWCAPS_S390X_LDISP | \
@@ -200,7 +206,13 @@ typedef
                                 VEX_HWCAPS_S390X_LSC2  | \
                                 VEX_HWCAPS_S390X_VXE   | \
                                 VEX_HWCAPS_S390X_NNPA  | \
-                                VEX_HWCAPS_S390X_DFLT)
+                                VEX_HWCAPS_S390X_DFLT  | \
+                                VEX_HWCAPS_S390X_VXE2  | \
+                                VEX_HWCAPS_S390X_VXD   | \
+                                VEX_HWCAPS_S390X_MSA   | \
+                                VEX_HWCAPS_S390X_MSA4  | \
+                                VEX_HWCAPS_S390X_MSA8  | \
+                                VEX_HWCAPS_S390X_MSA9)
 
 #define VEX_HWCAPS_S390X(x)  ((x) & ~VEX_S390X_MODEL_MASK)
 #define VEX_S390X_MODEL(x)   ((x) &  VEX_S390X_MODEL_MASK)
diff --git a/VEX/pub/libvex_emnote.h b/VEX/pub/libvex_emnote.h
index 77880a2702..c3ebcf4203 100644
--- a/VEX/pub/libvex_emnote.h
+++ b/VEX/pub/libvex_emnote.h
@@ -124,10 +124,42 @@ typedef
       /* ppno insn is not supported on this host */
       EmFail_S390X_ppno,
 
-      /* insn needs vector-enhancements facility which is not available on this
-         host */
+      /* insn needs vector-enhancements facility 1 which is not available on
+         this host */
       EmFail_S390X_vxe,
 
+      /* insn needs deflate-conversion facility which is not available on
+         this host */
+      EmFail_S390X_dflt,
+
+      /* insn needs neural-network-processing-assist facility which is not
+         available on this host */
+      EmFail_S390X_nnpa,
+
+      /* insn needs vector-enhancements facility 2 which is not available on
+         this host */
+      EmFail_S390X_vxe2,
+
+      /* insn needs vector-packed-decimal facility which is not available on
+         this host */
+      EmFail_S390X_vxd,
+
+      /* insn needs message-security assist which is not available on
+         this host */
+      EmFail_S390X_msa,
+
+      /* insn needs message-security-assist extension 4 which is not available
+         on this host */
+      EmFail_S390X_msa4,
+
+      /* insn needs message-security-assist extension 8 which is not available
+         on this host */
+      EmFail_S390X_msa8,
+
+      /* insn needs message-security-assist extension 9 which is not available
+         on this host */
+      EmFail_S390X_msa9,
+
       EmNote_NUMBER
    }
    VexEmNote;
diff --git a/coregrind/m_extension/extension-s390x.c b/coregrind/m_extension/extension-s390x.c
index 34e2e0409d..8a17c6fbdd 100644
--- a/coregrind/m_extension/extension-s390x.c
+++ b/coregrind/m_extension/extension-s390x.c
@@ -862,8 +862,7 @@ static enum ExtensionError do_extension_STFLE(ThreadState* tst, ULong variant)
       (S390_SETBITS(128, 131)
        /* 132: unassigned */
        /* 133: guarded-storage, not supported */
-       /* 134: vector packed decimal, not supported */
-       | S390_SETBITS(135, 135)
+       | S390_SETBITS(134, 135)
        /* 136: unassigned */
        /* 137: unassigned */
        | S390_SETBITS(138, 142)
diff --git a/coregrind/m_machine.c b/coregrind/m_machine.c
index 3b3671455f..eaa18b0b61 100644
--- a/coregrind/m_machine.c
+++ b/coregrind/m_machine.c
@@ -1620,11 +1620,13 @@ Bool VG_(machine_get_hwcaps)( void )
         UInt hwcaps_bit;
         const HChar name[6];   // may need adjustment for new facility names
      } fac_hwcaps[] = {
+        { False,  18,  VEX_HWCAPS_S390X_LDISP, "LDISP" },
         { False,  21,  VEX_HWCAPS_S390X_EIMM,  "EIMM"  },
         { False,  34,  VEX_HWCAPS_S390X_GIE,   "GIE"   },
         { False,  42,  VEX_HWCAPS_S390X_DFP,   "DFP"   },
         { False,  41,  VEX_HWCAPS_S390X_FGX,   "FGX"   },
         { False,  24,  VEX_HWCAPS_S390X_ETF2,  "ETF2"  },
+        { False,   7,  VEX_HWCAPS_S390X_STFLE, "STFLE" },
         { False,  30,  VEX_HWCAPS_S390X_ETF3,  "ETF3"  },
         { False,  25,  VEX_HWCAPS_S390X_STCKF, "STCKF" },
         { False,  37,  VEX_HWCAPS_S390X_FPEXT, "FPEXT" },
@@ -1637,6 +1639,12 @@ Bool VG_(machine_get_hwcaps)( void )
         { False, 135,  VEX_HWCAPS_S390X_VXE,   "VXE"   },
         { False, 151,  VEX_HWCAPS_S390X_DFLT,  "DFLT"  },
         { False, 165,  VEX_HWCAPS_S390X_NNPA,  "NNPA"  },
+        { False, 148,  VEX_HWCAPS_S390X_VXE2,  "VXE2"  },
+        { False, 134,  VEX_HWCAPS_S390X_VXD,   "VXD"   },
+        { False,  17,  VEX_HWCAPS_S390X_MSA,   "MSA"   },
+        { False,  77,  VEX_HWCAPS_S390X_MSA4,  "MSA4"  },
+        { False, 146,  VEX_HWCAPS_S390X_MSA8,  "MSA8"  },
+        { False, 155,  VEX_HWCAPS_S390X_MSA9,  "MSA9"  },
      };
 
      /* Set hwcaps according to the detected facilities */

[Valgrind-developers] [valgrind] s390x: Remove macro definitions for hardware facilities

[Florian K. <fk...@so...>]

https://sourceware.org/git/gitweb.cgi?p=valgrind.git;h=aee4a0a72e9a0dc49af18806643a49749db7eb5c

commit aee4a0a72e9a0dc49af18806643a49749db7eb5c
Author: Florian Krohm <fl...@ei...>
Date:   Wed Mar 12 22:36:33 2025 +0000

    s390x: Remove macro definitions for hardware facilities
    
    Remove the facility-bit macros S390_FAC_...
    They are guaranteed to never change and are used only in m_machine.c
    
    Part of fixing https://bugs.kde.org/show_bug.cgi?id=496950

Diff:
---
 VEX/pub/libvex_s390x_common.h | 37 -------------------------------------
 coregrind/m_machine.c         | 34 +++++++++++++++++-----------------
 2 files changed, 17 insertions(+), 54 deletions(-)

diff --git a/VEX/pub/libvex_s390x_common.h b/VEX/pub/libvex_s390x_common.h
index dda5ffd312..ea6a59d274 100644
--- a/VEX/pub/libvex_s390x_common.h
+++ b/VEX/pub/libvex_s390x_common.h
@@ -68,43 +68,6 @@
 #define S390_INNERLOOP_FRAME_SIZE ((8+1+1)*8 + 160)
 
 
-/*--------------------------------------------------------------*/
-/*--- Facility bits                                          ---*/
-/*--------------------------------------------------------------*/
-
-/* The value of the macro is the number of the facility bit as per POP. */
-#define S390_FAC_MSA     17  // message-security-assist
-#define S390_FAC_LDISP   18  // long displacement
-#define S390_FAC_HFPMAS  20  // HFP multiply-and-add-subtract
-#define S390_FAC_EIMM    21  // extended immediate
-#define S390_FAC_HFPUNX  23  // HFP unnormalized extension
-#define S390_FAC_ETF2    24  // ETF2-enhancement
-#define S390_FAC_STCKF   25  // store clock fast insn
-#define S390_FAC_PENH    26  // parsing-enhancement
-#define S390_FAC_ETF3    30  // ETF3-enhancement
-#define S390_FAC_XCPUT   31  // extract-CPU-time
-#define S390_FAC_GIE     34  // general insn extension
-#define S390_FAC_EXEXT   35  // execute extension
-#define S390_FAC_FPEXT   37  // floating-point extension
-#define S390_FAC_FPSE    41  // floating-point support enhancement
-#define S390_FAC_DFP     42  // decimal floating point
-#define S390_FAC_PFPO    44  // perform floating point operation insn
-#define S390_FAC_HIGHW   45  // high-word extension
-#define S390_FAC_LSC     45  // load/store on condition
-#define S390_FAC_DFPZC   48  // DFP zoned-conversion
-#define S390_FAC_MISC    49  // miscellaneous insn
-#define S390_FAC_CTREXE  50  // constrained transactional execution
-#define S390_FAC_LSC2    53  // load/store on condition 2 and load and zero rightmost byte
-#define S390_FAC_MSA5    57  // message-security-assist 5
-#define S390_FAC_MI2     58  // miscellaneous-instruction-extensions 2
-#define S390_FAC_TREXE   73  // transactional execution
-#define S390_FAC_MSA4    77  // message-security-assist 4
-#define S390_FAC_VX      129 // vector facility
-#define S390_FAC_VXE     135 // vector enhancements facility 1
-#define S390_FAC_VXE2    148 // vector enhancements facility 2
-#define S390_FAC_DFLT    151 // deflate-conversion facility
-#define S390_FAC_NNPA    165 // NNPA facility
-
 /*--------------------------------------------------------------*/
 /*--- Extensions                                             ---*/
 /*--------------------------------------------------------------*/
diff --git a/coregrind/m_machine.c b/coregrind/m_machine.c
index dec4f2373e..3b3671455f 100644
--- a/coregrind/m_machine.c
+++ b/coregrind/m_machine.c
@@ -1620,23 +1620,23 @@ Bool VG_(machine_get_hwcaps)( void )
         UInt hwcaps_bit;
         const HChar name[6];   // may need adjustment for new facility names
      } fac_hwcaps[] = {
-        { False, S390_FAC_EIMM,  VEX_HWCAPS_S390X_EIMM,  "EIMM"  },
-        { False, S390_FAC_GIE,   VEX_HWCAPS_S390X_GIE,   "GIE"   },
-        { False, S390_FAC_DFP,   VEX_HWCAPS_S390X_DFP,   "DFP"   },
-        { False, S390_FAC_FPSE,  VEX_HWCAPS_S390X_FGX,   "FGX"   },
-        { False, S390_FAC_ETF2,  VEX_HWCAPS_S390X_ETF2,  "ETF2"  },
-        { False, S390_FAC_ETF3,  VEX_HWCAPS_S390X_ETF3,  "ETF3"  },
-        { False, S390_FAC_STCKF, VEX_HWCAPS_S390X_STCKF, "STCKF" },
-        { False, S390_FAC_FPEXT, VEX_HWCAPS_S390X_FPEXT, "FPEXT" },
-        { False, S390_FAC_LSC,   VEX_HWCAPS_S390X_LSC,   "LSC"   },
-        { False, S390_FAC_PFPO,  VEX_HWCAPS_S390X_PFPO,  "PFPO"  },
-        { False, S390_FAC_VX,    VEX_HWCAPS_S390X_VX,    "VX"    },
-        { False, S390_FAC_MSA5,  VEX_HWCAPS_S390X_MSA5,  "MSA5"  },
-        { False, S390_FAC_MI2,   VEX_HWCAPS_S390X_MI2,   "MI2"   },
-        { False, S390_FAC_LSC2,  VEX_HWCAPS_S390X_LSC2,  "LSC2"  },
-        { False, S390_FAC_VXE,   VEX_HWCAPS_S390X_VXE,   "VXE"   },
-        { False, S390_FAC_DFLT,  VEX_HWCAPS_S390X_DFLT,  "DFLT"  },
-        { False, S390_FAC_NNPA,  VEX_HWCAPS_S390X_NNPA,  "NNPA"  },
+        { False,  21,  VEX_HWCAPS_S390X_EIMM,  "EIMM"  },
+        { False,  34,  VEX_HWCAPS_S390X_GIE,   "GIE"   },
+        { False,  42,  VEX_HWCAPS_S390X_DFP,   "DFP"   },
+        { False,  41,  VEX_HWCAPS_S390X_FGX,   "FGX"   },
+        { False,  24,  VEX_HWCAPS_S390X_ETF2,  "ETF2"  },
+        { False,  30,  VEX_HWCAPS_S390X_ETF3,  "ETF3"  },
+        { False,  25,  VEX_HWCAPS_S390X_STCKF, "STCKF" },
+        { False,  37,  VEX_HWCAPS_S390X_FPEXT, "FPEXT" },
+        { False,  45,  VEX_HWCAPS_S390X_LSC,   "LSC"   },
+        { False,  44,  VEX_HWCAPS_S390X_PFPO,  "PFPO"  },
+        { False, 129,  VEX_HWCAPS_S390X_VX,    "VX"    },
+        { False,  57,  VEX_HWCAPS_S390X_MSA5,  "MSA5"  },
+        { False,  58,  VEX_HWCAPS_S390X_MI2,   "MI2"   },
+        { False,  53,  VEX_HWCAPS_S390X_LSC2,  "LSC2"  },
+        { False, 135,  VEX_HWCAPS_S390X_VXE,   "VXE"   },
+        { False, 151,  VEX_HWCAPS_S390X_DFLT,  "DFLT"  },
+        { False, 165,  VEX_HWCAPS_S390X_NNPA,  "NNPA"  },
      };
 
      /* Set hwcaps according to the detected facilities */

[Valgrind-developers] [valgrind] m_debuginfo/debuginfo.c (di_notify_mmap): Ignore memfd_create mappings

[Mark W. <ma...@so...>]

https://sourceware.org/git/gitweb.cgi?p=valgrind.git;h=9616164273be6bd91552683d0543170de27af852

commit 9616164273be6bd91552683d0543170de27af852
Author: Mark Wielaard <ma...@kl...>
Date:   Wed Mar 12 13:42:12 2025 +0100

    m_debuginfo/debuginfo.c (di_notify_mmap): Ignore memfd_create mappings
    
    When valgrind observes a mapping of a memfd_create anonymous file it
    would generate a warning:
    
     WARNING: Serious error when reading debug info
     When reading debug info from /memfd:mozilla-ipc (deleted):
     failed to stat64/stat this file
    
    All such memfd_create anonymous files are always prefixed with
    "memfd:" and obviously cannot be stat64/stat. So ignore such warnings.
    
    Tested-by: ish...@yk...

Diff:
---
 coregrind/m_debuginfo/debuginfo.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/coregrind/m_debuginfo/debuginfo.c b/coregrind/m_debuginfo/debuginfo.c
index 15686fc5c4..6c4a692656 100644
--- a/coregrind/m_debuginfo/debuginfo.c
+++ b/coregrind/m_debuginfo/debuginfo.c
@@ -1196,7 +1196,9 @@ ULong VG_(di_notify_mmap)( Addr a, Bool allow_SkFileV, Int use_fd )
    if (sr_isError(statres)) {
       DebugInfo fake_di;
       Bool quiet = VG_(strstr)(filename, "/var/run/nscd/") != NULL
-                   || VG_(strstr)(filename, "/dev/shm/") != NULL;
+                   || VG_(strstr)(filename, "/dev/shm/") != NULL
+                   || VG_(strncmp)("/memfd:", filename,
+                                   VG_(strlen)("/memfd:")) == 0;
       if (!quiet && VG_(clo_verbosity) > 1) {
          VG_(memset)(&fake_di, 0, sizeof(fake_di));
          fake_di.fsm.filename = ML_(dinfo_strdup)("di.debuginfo.nmm", filename);

[Valgrind-developers] [valgrind/VALGRIND_3_24_BRANCH] gdbserver_tests: filter out new Missing rpms message

[Mark W. <ma...@so...>]

https://sourceware.org/git/gitweb.cgi?p=valgrind.git;h=7b2806871f15977db3a64a59f70d48f55ba3228e

commit 7b2806871f15977db3a64a59f70d48f55ba3228e
Author: Mark Wielaard <ma...@kl...>
Date:   Thu Jan 16 17:30:37 2025 +0100

    gdbserver_tests: filter out new Missing rpms message
    
    As seen on the fedora 40 s390x tester. GDB now might output something
    like:
    
    Missing rpms, try: dnf --enablerepo='*debug*' install glibc-debuginfo-2.39-33.fc40.s390x
    
    Filter those messages out to get zero fail test results again.
    
    (cherry picked from commit 090f8ce59b5f3d3ec39e032ee5e9524ce4f51a44)

Diff:
---
 gdbserver_tests/filter_gdb.in              | 1 +
 gdbserver_tests/filter_memcheck_monitor.in | 1 +
 2 files changed, 2 insertions(+)

diff --git a/gdbserver_tests/filter_gdb.in b/gdbserver_tests/filter_gdb.in
index 094ea933ec..2bef9f3ee5 100755
--- a/gdbserver_tests/filter_gdb.in
+++ b/gdbserver_tests/filter_gdb.in
@@ -265,6 +265,7 @@ s/^0x........ in \(\w\+ (\)/\1/
 
 #	delete any missing debuginfo messages
 /^Missing debuginfo.*/d
+/^Missing rpms.*/d
 
 EOF
 
diff --git a/gdbserver_tests/filter_memcheck_monitor.in b/gdbserver_tests/filter_memcheck_monitor.in
index 6e8a49a303..e407e7b921 100755
--- a/gdbserver_tests/filter_memcheck_monitor.in
+++ b/gdbserver_tests/filter_memcheck_monitor.in
@@ -31,6 +31,7 @@ $dir/filter_vgdb                                       |
 $SED -e '/Cannot access memory at address 0x......../d'                   \
     -e '/^[1-9][0-9]*	\.\.\/sysdeps\/powerpc\/powerpc32\/dl-start\.S: No such file or directory\./d' \
     -e '/^Missing separate debuginfo/d'                                  \
+    -e '/^Missing rpms/d'                                                \
     -e '/^Try: zypper install -C/d'                                      \
     -e 's/Test 3:   FAIL: expected si_code==2, not 128/Test 3:   PASS/'  \
     -e 's/in use at exit: [0-9][0-9,]* bytes in [0-9][0-9]* blocks/in use at exit: ... bytes in ... blocks/' \

[Valgrind-developers] [valgrind/VALGRIND_3_24_BRANCH] add_hardwired_spec for ld-linux-x86-64.so.2 memcmp

[Mark W. <ma...@so...>]

https://sourceware.org/git/gitweb.cgi?p=valgrind.git;h=2c17a8a3e865816ee48e7f78148bbc7240c2a38c

commit 2c17a8a3e865816ee48e7f78148bbc7240c2a38c
Author: Mark Wielaard <ma...@kl...>
Date:   Wed Mar 12 13:57:19 2025 +0100

    add_hardwired_spec for ld-linux-x86-64.so.2 memcmp
    
    With RPATH processing ld.so compiled for x86-64-v3 uses an optimized
    avx2 memcmp (bcmp) which causes (false postive) memcheck invalid reads
    of size 32 warnings.
    
    Fix this my adding a hardwire spec with a simpler memcmp for
    ld-linux-x86-64.so.2.
    
    https://bugs.kde.org/show_bug.cgi?id=501348
    
    Authored-by: Florian Weimer <fw...@re...>
    
    (cherry picked from commit 9ffdeb4927b8505cc5a9ea54f608f4972166bb0e)

Diff:
---
 NEWS                            |  1 +
 coregrind/m_redir.c             |  9 +++++++++
 coregrind/m_trampoline.S        | 19 +++++++++++++++++++
 coregrind/pub_core_trampoline.h |  1 +
 4 files changed, 30 insertions(+)

diff --git a/NEWS b/NEWS
index 7f9e005c59..fe225c666c 100644
--- a/NEWS
+++ b/NEWS
@@ -11,6 +11,7 @@ The following bugs have been fixed or resolved on this branch.
 498317  FdBadUse is not a valid CoreError type in a suppression
         even though it's generated by --gen-suppressions=yes
 498143  False positive on EVIOCGRAB ioctl
+501348  glibc built with -march=x86-64-v3 does not work due to ld.so memcmp
 
 To see details of a given bug, visit
   https://bugs.kde.org/show_bug.cgi?id=XXXXXX
diff --git a/coregrind/m_redir.c b/coregrind/m_redir.c
index f9e8d8801a..64ebea3b63 100644
--- a/coregrind/m_redir.c
+++ b/coregrind/m_redir.c
@@ -1425,6 +1425,15 @@ void VG_(redir_initialise) ( void )
          NULL
 #        else
          complain_about_stripped_glibc_ldso
+#        endif
+      );
+      add_hardwired_spec(
+         "ld-linux-x86-64.so.2", "memcmp",
+         (Addr)&VG_(amd64_linux_REDIR_FOR_memcmp),
+#        ifndef GLIBC_MANDATORY_STRLEN_REDIRECT
+         NULL
+#        else
+         complain_about_stripped_glibc_ldso
 #        endif
       );
    }
diff --git a/coregrind/m_trampoline.S b/coregrind/m_trampoline.S
index e897963be2..27f1b35d2b 100644
--- a/coregrind/m_trampoline.S
+++ b/coregrind/m_trampoline.S
@@ -241,6 +241,25 @@ VG_(amd64_linux_REDIR_FOR_strcmp):
 	ret
 .size VG_(amd64_linux_REDIR_FOR_strcmp), .-VG_(amd64_linux_REDIR_FOR_strcmp)
 
+.global VG_(amd64_linux_REDIR_FOR_memcmp)
+.type   VG_(amd64_linux_REDIR_FOR_memcmp), @function
+VG_(amd64_linux_REDIR_FOR_memcmp):
+	xorl	%ecx, %ecx
+	jmp	2f
+1:
+	movzbl	(%rdi, %rcx), %eax
+	movzbl	(%rsi, %rcx), %r8d
+	addq	$1, %rcx
+	subl	%r8d, %eax
+	jne	3f
+2:
+	cmpq	%rcx, %rdx
+	jne	1b
+	xorl	%eax, %eax
+3:
+	ret
+.size VG_(amd64_linux_REDIR_FOR_memcmp), .-VG_(amd64_linux_REDIR_FOR_memcmp)
+
 .global VG_(amd64_linux_REDIR_FOR_index)
 .type   VG_(amd64_linux_REDIR_FOR_index), @function
 VG_(amd64_linux_REDIR_FOR_index):
diff --git a/coregrind/pub_core_trampoline.h b/coregrind/pub_core_trampoline.h
index d0bd6b8597..3d2f7e9f32 100644
--- a/coregrind/pub_core_trampoline.h
+++ b/coregrind/pub_core_trampoline.h
@@ -82,6 +82,7 @@ extern Addr VG_(amd64_linux_REDIR_FOR_vtime);
 extern Addr VG_(amd64_linux_REDIR_FOR_vgetcpu);
 extern UInt VG_(amd64_linux_REDIR_FOR_strlen)( void* );
 extern Int VG_(amd64_linux_REDIR_FOR_strcmp)( void*, void* );
+extern Int VG_(amd64_linux_REDIR_FOR_memcmp)( void*, void*, SizeT );
 extern Char* VG_(amd64_linux_REDIR_FOR_index) ( const Char*, Int );
 #endif

[Valgrind-developers] [valgrind] add_hardwired_spec for ld-linux-x86-64.so.2 memcmp

[Mark W. <ma...@so...>]

https://sourceware.org/git/gitweb.cgi?p=valgrind.git;h=9ffdeb4927b8505cc5a9ea54f608f4972166bb0e

commit 9ffdeb4927b8505cc5a9ea54f608f4972166bb0e
Author: Mark Wielaard <ma...@kl...>
Date:   Wed Mar 12 13:57:19 2025 +0100

    add_hardwired_spec for ld-linux-x86-64.so.2 memcmp
    
    With RPATH processing ld.so compiled for x86-64-v3 uses an optimized
    avx2 memcmp (bcmp) which causes (false postive) memcheck invalid reads
    of size 32 warnings.
    
    Fix this my adding a hardwire spec with a simpler memcmp for
    ld-linux-x86-64.so.2.
    
    https://bugs.kde.org/show_bug.cgi?id=501348
    
    Authored-by: Florian Weimer <fw...@re...>

Diff:
---
 NEWS                            |  1 +
 coregrind/m_redir.c             |  9 +++++++++
 coregrind/m_trampoline.S        | 19 +++++++++++++++++++
 coregrind/pub_core_trampoline.h |  1 +
 4 files changed, 30 insertions(+)

diff --git a/NEWS b/NEWS
index 16cfeef2a4..7025d61af7 100644
--- a/NEWS
+++ b/NEWS
@@ -58,6 +58,7 @@ are not entered into bugzilla tend to get forgotten about or ignored.
 499212  mmap() with MAP_ALIGNED() returns unaligned pointer
 501119  memcheck/tests/pointer-trace fails when run on NFS filesystem
 501194  Fix ML_(check_macho_and_get_rw_loads) so that it is correct for any number of segment commands
+501348 glibc built with -march=x86-64-v3 does not work due to ld.so memcmp
 
 
 To see details of a given bug, visit
diff --git a/coregrind/m_redir.c b/coregrind/m_redir.c
index 5e7bc42f97..63172b9717 100644
--- a/coregrind/m_redir.c
+++ b/coregrind/m_redir.c
@@ -1431,6 +1431,15 @@ void VG_(redir_initialise) ( void )
          NULL
 #        else
          complain_about_stripped_glibc_ldso
+#        endif
+      );
+      add_hardwired_spec(
+         "ld-linux-x86-64.so.2", "memcmp",
+         (Addr)&VG_(amd64_linux_REDIR_FOR_memcmp),
+#        ifndef GLIBC_MANDATORY_STRLEN_REDIRECT
+         NULL
+#        else
+         complain_about_stripped_glibc_ldso
 #        endif
       );
    }
diff --git a/coregrind/m_trampoline.S b/coregrind/m_trampoline.S
index 4860a2259f..2c2cc0dc2a 100644
--- a/coregrind/m_trampoline.S
+++ b/coregrind/m_trampoline.S
@@ -241,6 +241,25 @@ VG_(amd64_linux_REDIR_FOR_strcmp):
 	ret
 .size VG_(amd64_linux_REDIR_FOR_strcmp), .-VG_(amd64_linux_REDIR_FOR_strcmp)
 
+.global VG_(amd64_linux_REDIR_FOR_memcmp)
+.type   VG_(amd64_linux_REDIR_FOR_memcmp), @function
+VG_(amd64_linux_REDIR_FOR_memcmp):
+	xorl	%ecx, %ecx
+	jmp	2f
+1:
+	movzbl	(%rdi, %rcx), %eax
+	movzbl	(%rsi, %rcx), %r8d
+	addq	$1, %rcx
+	subl	%r8d, %eax
+	jne	3f
+2:
+	cmpq	%rcx, %rdx
+	jne	1b
+	xorl	%eax, %eax
+3:
+	ret
+.size VG_(amd64_linux_REDIR_FOR_memcmp), .-VG_(amd64_linux_REDIR_FOR_memcmp)
+
 .global VG_(amd64_linux_REDIR_FOR_index)
 .type   VG_(amd64_linux_REDIR_FOR_index), @function
 VG_(amd64_linux_REDIR_FOR_index):
diff --git a/coregrind/pub_core_trampoline.h b/coregrind/pub_core_trampoline.h
index db497c51a5..11d791df7c 100644
--- a/coregrind/pub_core_trampoline.h
+++ b/coregrind/pub_core_trampoline.h
@@ -82,6 +82,7 @@ extern Addr VG_(amd64_linux_REDIR_FOR_vtime);
 extern Addr VG_(amd64_linux_REDIR_FOR_vgetcpu);
 extern UInt VG_(amd64_linux_REDIR_FOR_strlen)( void* );
 extern Int VG_(amd64_linux_REDIR_FOR_strcmp)( void*, void* );
+extern Int VG_(amd64_linux_REDIR_FOR_memcmp)( void*, void*, SizeT );
 extern Char* VG_(amd64_linux_REDIR_FOR_index) ( const Char*, Int );
 #endif

Re: [Valgrind-developers] [valgrind] coregrind/m_debuginfo: don't try to examine zero sized mmapped files

[Mark W. <ma...@kl...>]

Hi Chiaki,

On Wed, 2025-03-12 at 04:40 +0900, ISHIKAWA,chiaki wrote:
> It works.
> I don't get the noisy warnings any more.

Thanks for testing. I pushed the attached commit.

Cheers,

Mark

[Valgrind-developers] [valgrind] Solaris suppressions: add suppressions for Solaris 11.4 getaddrinfo

[Paul F. <pa...@so...>]

https://sourceware.org/git/gitweb.cgi?p=valgrind.git;h=78b3b79fcda79dee94d3a181afc4c4a0ed95fbb4

commit 78b3b79fcda79dee94d3a181afc4c4a0ed95fbb4
Author: Paul Floyd <pj...@wa...>
Date:   Wed Mar 12 08:33:11 2025 +0100

    Solaris suppressions: add suppressions for Solaris 11.4 getaddrinfo
    
    getaddrinfo, the bane of all threading tests.

Diff:
---
 solaris12.supp | 64 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 64 insertions(+)

diff --git a/solaris12.supp b/solaris12.supp
index 2c68508f5c..29f8086619 100644
--- a/solaris12.supp
+++ b/solaris12.supp
@@ -166,3 +166,67 @@
    fun:pthread_barrier_wait
 }
 
+#----------------------------------------------------------------------------#
+# The getattrinfo testcase causes issues on all platforms, and we assume
+# that they are due to private libc locking mechanisms.
+
+{
+   Solaris:_nsc_getdoorbuf
+   drd:ConflictingAccess
+   fun:tsdalloc
+   fun:_nsc_getdoorbuf
+}
+
+{
+   Solaris:nss_dbop_search
+   drd:ConflictingAccess
+   fun:nss_dbop_search
+}
+
+{
+   Solaris:thr_create_once
+   drd:ConflictingAccess
+   fun:thr_keycreate_once
+}
+
+{
+   Solaris:_nsc_getdoorbsize
+   drd:ConflictingAccess
+   fun:_nsc_getdoorbsize
+}
+
+{
+   Solaris:_nsc_proc_is_cache
+   drd:ConflictingAccess
+   fun:_nsc_proc_is_cache
+}
+
+{
+   Solaris:__rpc_getconfip
+   drd:ConflictingAccess
+   fun:__rpc_getconfip
+}
+
+{
+   Solaris:_nsc_proc_is_cache
+   Helgrind:Race
+   fun:_nsc_proc_is_cache
+}
+
+{
+   Solaris:thr_create_once
+   Helgrind:Race
+   fun:thr_keycreate_once
+}
+
+{
+   Solaris:_nsc_proc_is_cache
+   Helgrind:Race
+   fun:_nsc_proc_is_cache
+}
+
+{
+   Solaris:nss_dbop_search
+   Helgrind:Race
+   fun:nss_dbop_search
+}

[Valgrind-developers] [valgrind] Solaris regtest: fdleak filter, change __xnet_socket to just socket

[Paul F. <pa...@so...>]

https://sourceware.org/git/gitweb.cgi?p=valgrind.git;h=92480c0fee983d2bd0ee7fdb8ed01783834b30ac

commit 92480c0fee983d2bd0ee7fdb8ed01783834b30ac
Author: Paul Floyd <pj...@wa...>
Date:   Wed Mar 12 08:12:10 2025 +0100

    Solaris regtest: fdleak filter, change __xnet_socket to just socket

Diff:
---
 none/tests/filter_fdleak | 1 +
 1 file changed, 1 insertion(+)

diff --git a/none/tests/filter_fdleak b/none/tests/filter_fdleak
index f84008b6db..18a65b4568 100755
--- a/none/tests/filter_fdleak
+++ b/none/tests/filter_fdleak
@@ -21,6 +21,7 @@ perl -p -e 's/socket\.c:[1-9][0-9]*/in \/...libc.../' |
 perl -p -e 's/ __close / close /' |
 perl -p -e 's/ __write / write /' |
 sed '/ _so_socket /d;' |
+sed 's/__xnet_socket/socket/' |
 awk '/ socket /{sub(/ by /, " at "); print; next}{print}' |
 awk '/ at .* _syscall6 /{getline; getline; sub(/ by /, " at "); print; next}{print}' |
 sed '/__open/d' |

[Valgrind-developers] [valgrind] Illumos regtest: add a filter for shell 'Abort' messages

[Paul F. <pa...@so...>]

https://sourceware.org/git/gitweb.cgi?p=valgrind.git;h=36ab09596af75e5c977ca43e20fc4dd598964454

commit 36ab09596af75e5c977ca43e20fc4dd598964454
Author: Paul Floyd <pj...@wa...>
Date:   Tue Mar 11 20:50:07 2025 +0100

    Illumos regtest: add a filter for shell 'Abort' messages

Diff:
---
 tests/filter_stderr_basic.in | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tests/filter_stderr_basic.in b/tests/filter_stderr_basic.in
index 7d38a0d4fd..d14ca727df 100755
--- a/tests/filter_stderr_basic.in
+++ b/tests/filter_stderr_basic.in
@@ -57,7 +57,7 @@ perl -n -e 'print if !/^(Illegal instruction|Segmentation fault|Alarm clock|Abor
 # Similar as above, but for ksh on Solaris/illumos.
 perl -n -e 'print if !/^(Memory fault|Killed) $/' |
 # bash on Illumos
-$SED "/sh: [1-9][0-9]*: \(Memory fault\|Alarm call\|Terminated\|Killed\)/d" |
+$SED "/sh: [1-9][0-9]*: \(Memory fault\|Alarm call\|Terminated\|Killed\|Abort\)/d" |
 
 # Translate intercepted glibc functions back to their canonical name
 perl -p -e "s/: memcpy\@\@?GLIBC_[.1-9]+ \(vg_replace_strmem.c:.*?\)/: memcpy \(vg_replace_strmem.c:...\)/" |

Re: [Valgrind-developers] [valgrind] coregrind/m_debuginfo: don't try to examine zero sized mmapped files

[ISHIKAWA,chiaki <ish...@yk...>]

Dear Mark,

It works.
I don't get the noisy warnings any more.

Thank you for the quick fix!

Chiaki


On 2025/03/10 14:43, ISHIKAWA,chiaki wrote:
> Dear Mark,
>
> Thank you for your analysis.
>
> I will try to create a localized version of valgrind after applying 
> the patch and see how it goes.
>
> It might take a couple of days due to my day job, etc.
>
> Chiaki
>
> On 2025/03/10 5:45, Mark Wielaard wrote:
>> Hi Chiaki,
>>
>> On Mon, Mar 10, 2025 at 03:32:16AM +0900, ISHIKAWA,chiaki wrote:
>>> Does this help reduce the occurences of presumably non-critical
>>> error messages like the following?
>> No. These are indeed related messages, but caused by using memfd
>> allocated memory.
>>
>>> I get this when I try to run mozilla's thunderbird mail client under
>>> valgrind.
>>> They appear many times.
>>>
>>>    23:44.69 GECKO(2640779) --2640783-- WARNING: Serious error when
>>> reading debug info
>>>    23:44.69 GECKO(2640779) --2640783-- When reading debug info from
>>> /memfd:mozilla-ipc (deleted):
>>>    23:44.69 GECKO(2640779) --2640783-- failed to stat64/stat this file
>>>    23:44.69 GECKO(2640779) --2640783-- WARNING: Serious error when
>>> reading debug info
>>>    23:44.69 GECKO(2640779) --2640783-- When reading debug info from
>>> /memfd:mozilla-ipc (deleted):
>>>    23:44.69 GECKO(2640779) --2640783-- failed to stat64/stat this file
>>>
>>> I have no idea why valgrind wants try to to read debug info from
>>> memory region.
>> I presume because valgrind thinks it might be a real memory mapped
>> file. The code triggering this warning is:
>>
>>     /* Don't let the stat call fail silently.  Filter out some known
>>        sources of noise before complaining, though. */
>>     if (sr_isError(statres)) {
>>        DebugInfo fake_di;
>>        Bool quiet = VG_(strstr)(filename, "/var/run/nscd/") != NULL
>>                     || VG_(strstr)(filename, "/dev/shm/") != NULL;
>>        if (!quiet && VG_(clo_verbosity) > 1) {
>>           VG_(memset)(&fake_di, 0, sizeof(fake_di));
>>           fake_di.fsm.filename = 
>> ML_(dinfo_strdup)("di.debuginfo.nmm", filename);
>>           ML_(symerr)(&fake_di, True, "failed to stat64/stat this 
>> file");
>>        }
>>        return 0;
>>     }
>>
>> So we probably should add "/memfd:" to the patterns for which not to
>> warn.
>>
>> https://www.man7.org/linux/man-pages/man2/memfd_create.2.html says:
>>
>>         int memfd_create(const char *name, unsigned int flags);
>>
>>      The name supplied in name is used as a filename and will be
>>      displayed as the target of the corresponding symbolic link in the
>>      directory /proc/self/fd/.  The displayed name is always prefixed
>>      with memfd: and serves only for debugging purposes.
>>
>> So I would propose something like the following (untested) patch:
>>
>> diff --git a/coregrind/m_debuginfo/debuginfo.c 
>> b/coregrind/m_debuginfo/debuginfo.c
>> index 15686fc5c46f..6c4a6926560b 100644
>> --- a/coregrind/m_debuginfo/debuginfo.c
>> +++ b/coregrind/m_debuginfo/debuginfo.c
>> @@ -1196,7 +1196,9 @@ ULong VG_(di_notify_mmap)( Addr a, Bool 
>> allow_SkFileV, Int use_fd )
>>      if (sr_isError(statres)) {
>>         DebugInfo fake_di;
>>         Bool quiet = VG_(strstr)(filename, "/var/run/nscd/") != NULL
>> -                   || VG_(strstr)(filename, "/dev/shm/") != NULL;
>> +                   || VG_(strstr)(filename, "/dev/shm/") != NULL
>> +                   || VG_(strncmp)("/memfd:", filename,
>> +                                   VG_(strlen)("/memfd:")) == 0;
>>         if (!quiet && VG_(clo_verbosity) > 1) {
>>            VG_(memset)(&fake_di, 0, sizeof(fake_di));
>>            fake_di.fsm.filename = 
>> ML_(dinfo_strdup)("di.debuginfo.nmm", filename);
>>
>> Does that work for you?
>>
>> Thanks,
>>
>> Mark
>
>
>
>
> _______________________________________________
> Valgrind-developers mailing list
> Val...@li...
> https://lists.sourceforge.net/lists/listinfo/valgrind-developers

[Valgrind-developers] [valgrind] Illumos fdleak: dup2 spurious errors

[Paul F. <pa...@so...>]

https://sourceware.org/git/gitweb.cgi?p=valgrind.git;h=dd169be0e65fecd1e4b7593dc3da7b87dc7ecf03

commit dd169be0e65fecd1e4b7593dc3da7b87dc7ecf03
Author: Paul Floyd <pj...@wa...>
Date:   Tue Mar 11 20:36:30 2025 +0100

    Illumos fdleak: dup2 spurious errors
    
    We should also put back checks that calls like dup2 aren't
    using reserved file descriptors.

Diff:
---
 coregrind/m_syswrap/syswrap-solaris.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/coregrind/m_syswrap/syswrap-solaris.c b/coregrind/m_syswrap/syswrap-solaris.c
index 3da0d216cd..dc0068d90c 100644
--- a/coregrind/m_syswrap/syswrap-solaris.c
+++ b/coregrind/m_syswrap/syswrap-solaris.c
@@ -3989,11 +3989,13 @@ PRE(sys_fcntl)
       PRE_REG_READ3(long, "fcntl", int, fildes, int, cmd, int, arg);
       /* Check if a client program isn't going to poison any of V's output
          fds. */
+      /*
       if (ARG2 == VKI_F_DUP2FD &&
           !ML_(fd_allowed)(ARG3, "fcntl(F_DUP2FD)", tid, False)) {
          SET_STATUS_Failure(VKI_EBADF);
          return;
       }
+      */
       break;
 
    /* These ones use ARG3 as "native lock" (input only). */

[Valgrind-developers] [valgrind] Regtest: filter out pathname from memcheck dw4 test

[Paul F. <pa...@so...>]

https://sourceware.org/git/gitweb.cgi?p=valgrind.git;h=e479ebc7fa189561bd1459aa60d1374bf67a6746

commit e479ebc7fa189561bd1459aa60d1374bf67a6746
Author: Paul Floyd <pj...@wa...>
Date:   Tue Mar 11 09:10:42 2025 +0100

    Regtest: filter out pathname from memcheck dw4 test

Diff:
---
 memcheck/tests/dw4.stderr.exp | 4 ++--
 memcheck/tests/filter_dw4.in  | 3 +++
 2 files changed, 5 insertions(+), 2 deletions(-)

diff --git a/memcheck/tests/dw4.stderr.exp b/memcheck/tests/dw4.stderr.exp
index e4b523f982..02c5955110 100644
--- a/memcheck/tests/dw4.stderr.exp
+++ b/memcheck/tests/dw4.stderr.exp
@@ -35,12 +35,12 @@ Unaddressable byte(s) found during client check request
 Uninitialised byte(s) found during client check request
    at 0x........: croak (dw4.c:32)
    by 0x........: main (dw4.c:89)
- Address 0x........ is in a rw- mapped file /export/home/paulf/valgrind/memcheck/tests/dw4 segment
+ Address 0x........ is in a rw- mapped file DIRECTORY/dw4 segment
 
 Uninitialised byte(s) found during client check request
    at 0x........: croak (dw4.c:32)
    by 0x........: main (dw4.c:91)
- Address 0x........ is in a rw- mapped file /export/home/paulf/valgrind/memcheck/tests/dw4 segment
+ Address 0x........ is in a rw- mapped file DIRECTORY/dw4 segment
 
 Unaddressable byte(s) found during client check request
    at 0x........: croak (dw4.c:25)
diff --git a/memcheck/tests/filter_dw4.in b/memcheck/tests/filter_dw4.in
index 96b6ab8f50..d729ad6e81 100755
--- a/memcheck/tests/filter_dw4.in
+++ b/memcheck/tests/filter_dw4.in
@@ -9,6 +9,9 @@ $SED "s/inside a block of size [0-9]* alloc'd/inside a block of size ... alloc'd
 # remove directory name and pid from mapped filename
 $SED "s/file .*valgrind-dw4-test.[1-9][0-9]*/file valgrind-dw4-test.PID/" |
 
+# remove name from mapped file segment
+$SED "s#is in a rw- mapped file .*/dw4 segment#is in a rw- mapped file DIRECTORY/dw4 segment#" |
+
 ./filter_stderr "$@"
 
 exit 0

[Valgrind-developers] [valgrind] Fix more int to pointer warnings.

[Paul F. <pa...@so...>]

https://sourceware.org/git/gitweb.cgi?p=valgrind.git;h=ca9cba8346b921a2beb7b61ca8a707204ee5aeff

commit ca9cba8346b921a2beb7b61ca8a707204ee5aeff
Author: Paul Floyd <pj...@wa...>
Date:   Tue Mar 11 07:40:48 2025 +0100

    Fix more int to pointer warnings.
    
    No more produced with GCC 14 on OI.

Diff:
---
 VEX/priv/host_riscv64_defs.c                 | 6 +++---
 VEX/priv/host_riscv64_isel.c                 | 2 +-
 coregrind/m_gdbserver/valgrind-low-riscv64.c | 2 +-
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/VEX/priv/host_riscv64_defs.c b/VEX/priv/host_riscv64_defs.c
index 6296b2e7fe..7e6b95380a 100644
--- a/VEX/priv/host_riscv64_defs.c
+++ b/VEX/priv/host_riscv64_defs.c
@@ -2647,7 +2647,7 @@ VexInvalRange unchainXDirect_RISCV64(VexEndness  endness_host,
    UChar* p = place_to_unchain;
    vassert(((HWord)p & 1) == 0);
    vassert(is_addr48_to_ireg_EXACTLY_18B(p, 5 /*x5/t0*/,
-                                         (ULong)place_to_jump_to_EXPECTED));
+                                         (ULong)(HWord)place_to_jump_to_EXPECTED));
    vassert(p[18] == 0x82 && p[19] == 0x82);
 
    /* And what we want to change it to is:
@@ -2664,7 +2664,7 @@ VexInvalRange unchainXDirect_RISCV64(VexEndness  endness_host,
 
       The replacement has the same length as the original.
    */
-   (void)addr48_to_ireg_EXACTLY_18B(p, 5 /*x5/t0*/, (ULong)disp_cp_chain_me);
+   (void)addr48_to_ireg_EXACTLY_18B(p, 5 /*x5/t0*/, (ULong)(HWord)disp_cp_chain_me);
    p[18] = 0x82;
    p[19] = 0x89;
 
@@ -2686,7 +2686,7 @@ VexInvalRange patchProfInc_RISCV64(VexEndness   endness_host,
    vassert(p[18] == 0x83 && p[19] == 0x32 && p[20] == 0x03 && p[21] == 0x00);
    vassert(p[22] == 0x85 && p[23] == 0x02);
    vassert(p[24] == 0x23 && p[25] == 0x30 && p[26] == 0x53 && p[27] == 0x00);
-   (void)addr48_to_ireg_EXACTLY_18B(p, 6 /*x6/t1*/, (ULong)location_of_counter);
+   (void)addr48_to_ireg_EXACTLY_18B(p, 6 /*x6/t1*/, (ULong)(HWord)location_of_counter);
    VexInvalRange vir = {(HWord)p, 28};
    return vir;
 }
diff --git a/VEX/priv/host_riscv64_isel.c b/VEX/priv/host_riscv64_isel.c
index 6921a47db4..c476d133c7 100644
--- a/VEX/priv/host_riscv64_isel.c
+++ b/VEX/priv/host_riscv64_isel.c
@@ -542,7 +542,7 @@ static Bool doHelperCall(/*OUT*/ UInt*   stackAdjustAfterCall,
       The same applies to nextFArgReg which records a number of used
       floating-point registers f10/fa0 .. f17/fa7.
     */
-   addInstr(env, RISCV64Instr_Call(*retloc, (Addr64)cee->addr, cond, nextArgReg,
+   addInstr(env, RISCV64Instr_Call(*retloc, (Addr64)(HWord)cee->addr, cond, nextArgReg,
                                    nextFArgReg));
 
    return True;
diff --git a/coregrind/m_gdbserver/valgrind-low-riscv64.c b/coregrind/m_gdbserver/valgrind-low-riscv64.c
index 75088fc1ec..5c200318c0 100644
--- a/coregrind/m_gdbserver/valgrind-low-riscv64.c
+++ b/coregrind/m_gdbserver/valgrind-low-riscv64.c
@@ -264,7 +264,7 @@ static CORE_ADDR** target_get_dtv (ThreadState *tst)
             void *private;
         } tcbhead_t;
    */
-   return (CORE_ADDR**)(void *)(riscv->guest_x4 - 2 * sizeof(void *));
+   return (CORE_ADDR**)(void *)((HWord)riscv->guest_x4 - 2 * sizeof(void *));
 }
 
 static struct valgrind_target_ops low_target = {

[Valgrind-developers] [valgrind] s390x: disasm-test: Fix thinko

[Florian K. <fk...@so...>]

https://sourceware.org/git/gitweb.cgi?p=valgrind.git;h=1d9cca850c5401784fe389b6792cb23c803d6342

commit 1d9cca850c5401784fe389b6792cb23c803d6342
Author: Florian Krohm <fl...@ei...>
Date:   Mon Mar 10 12:36:21 2025 +0000

    s390x: disasm-test: Fix thinko
    
    Remove verbiage about limited support for opcodes with optional operands.
    Optional operands are of no concern for disassembly. The reason is that
    disassembly sees as its input only a sequence of bytes which have no
    indication as to whether some of its bits were explicitly specified or not.
    We're testing the disassembler here - not the assembler.

Diff:
---
 none/tests/s390x/disasm-test/README | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/none/tests/s390x/disasm-test/README b/none/tests/s390x/disasm-test/README
index 54df05cf98..a771c41c98 100644
--- a/none/tests/s390x/disasm-test/README
+++ b/none/tests/s390x/disasm-test/README
@@ -169,10 +169,6 @@ Status
 Only opcodes with extended mnemonics as identified in Appendix J of the
 Principles of Operation are considered.
 
-There is only partial support for opcodes with optional operands.
-In the sense that the generated testcases will always include the
-optional operand.
-
 
 TODO
 ----

Re: [Valgrind-developers] [valgrind] coregrind/m_debuginfo: don't try to examine zero sized mmapped files

[ISHIKAWA,chiaki <ish...@yk...>]

Dear Mark,

Thank you for your analysis.

I will try to create a localized version of valgrind after applying the 
patch and see how it goes.

It might take a couple of days due to my day job, etc.

Chiaki

On 2025/03/10 5:45, Mark Wielaard wrote:
> Hi Chiaki,
>
> On Mon, Mar 10, 2025 at 03:32:16AM +0900, ISHIKAWA,chiaki wrote:
>> Does this help reduce the occurences of presumably non-critical
>> error messages like the following?
> No. These are indeed related messages, but caused by using memfd
> allocated memory.
>
>> I get this when I try to run mozilla's thunderbird mail client under
>> valgrind.
>> They appear many times.
>>
>>    23:44.69 GECKO(2640779) --2640783-- WARNING: Serious error when
>> reading debug info
>>    23:44.69 GECKO(2640779) --2640783-- When reading debug info from
>> /memfd:mozilla-ipc (deleted):
>>    23:44.69 GECKO(2640779) --2640783-- failed to stat64/stat this file
>>    23:44.69 GECKO(2640779) --2640783-- WARNING: Serious error when
>> reading debug info
>>    23:44.69 GECKO(2640779) --2640783-- When reading debug info from
>> /memfd:mozilla-ipc (deleted):
>>    23:44.69 GECKO(2640779) --2640783-- failed to stat64/stat this file
>>
>> I have no idea why valgrind wants try to to read debug info from
>> memory region.
> I presume because valgrind thinks it might be a real memory mapped
> file. The code triggering this warning is:
>
>     /* Don't let the stat call fail silently.  Filter out some known
>        sources of noise before complaining, though. */
>     if (sr_isError(statres)) {
>        DebugInfo fake_di;
>        Bool quiet = VG_(strstr)(filename, "/var/run/nscd/") != NULL
>                     || VG_(strstr)(filename, "/dev/shm/") != NULL;
>        if (!quiet && VG_(clo_verbosity) > 1) {
>           VG_(memset)(&fake_di, 0, sizeof(fake_di));
>           fake_di.fsm.filename = ML_(dinfo_strdup)("di.debuginfo.nmm", filename);
>           ML_(symerr)(&fake_di, True, "failed to stat64/stat this file");
>        }
>        return 0;
>     }
>
> So we probably should add "/memfd:" to the patterns for which not to
> warn.
>
> https://www.man7.org/linux/man-pages/man2/memfd_create.2.html says:
>
>         int memfd_create(const char *name, unsigned int flags);
>
>      The name supplied in name is used as a filename and will be
>      displayed as the target of the corresponding symbolic link in the
>      directory /proc/self/fd/.  The displayed name is always prefixed
>      with memfd: and serves only for debugging purposes.
>
> So I would propose something like the following (untested) patch:
>
> diff --git a/coregrind/m_debuginfo/debuginfo.c b/coregrind/m_debuginfo/debuginfo.c
> index 15686fc5c46f..6c4a6926560b 100644
> --- a/coregrind/m_debuginfo/debuginfo.c
> +++ b/coregrind/m_debuginfo/debuginfo.c
> @@ -1196,7 +1196,9 @@ ULong VG_(di_notify_mmap)( Addr a, Bool allow_SkFileV, Int use_fd )
>      if (sr_isError(statres)) {
>         DebugInfo fake_di;
>         Bool quiet = VG_(strstr)(filename, "/var/run/nscd/") != NULL
> -                   || VG_(strstr)(filename, "/dev/shm/") != NULL;
> +                   || VG_(strstr)(filename, "/dev/shm/") != NULL
> +                   || VG_(strncmp)("/memfd:", filename,
> +                                   VG_(strlen)("/memfd:")) == 0;
>         if (!quiet && VG_(clo_verbosity) > 1) {
>            VG_(memset)(&fake_di, 0, sizeof(fake_di));
>            fake_di.fsm.filename = ML_(dinfo_strdup)("di.debuginfo.nmm", filename);
>
> Does that work for you?
>
> Thanks,
>
> Mark

Re: [Valgrind-developers] [valgrind] coregrind/m_debuginfo: don't try to examine zero sized mmapped files

[Mark W. <ma...@kl...>]

Hi Chiaki,

On Mon, Mar 10, 2025 at 03:32:16AM +0900, ISHIKAWA,chiaki wrote:
> Does this help reduce the occurences of presumably non-critical
> error messages like the following?

No. These are indeed related messages, but caused by using memfd
allocated memory.

> I get this when I try to run mozilla's thunderbird mail client under
> valgrind.
> They appear many times.
> 
>   23:44.69 GECKO(2640779) --2640783-- WARNING: Serious error when
> reading debug info
>   23:44.69 GECKO(2640779) --2640783-- When reading debug info from
> /memfd:mozilla-ipc (deleted):
>   23:44.69 GECKO(2640779) --2640783-- failed to stat64/stat this file
>   23:44.69 GECKO(2640779) --2640783-- WARNING: Serious error when
> reading debug info
>   23:44.69 GECKO(2640779) --2640783-- When reading debug info from
> /memfd:mozilla-ipc (deleted):
>   23:44.69 GECKO(2640779) --2640783-- failed to stat64/stat this file
> 
> I have no idea why valgrind wants try to to read debug info from
> memory region.

I presume because valgrind thinks it might be a real memory mapped
file. The code triggering this warning is:

   /* Don't let the stat call fail silently.  Filter out some known
      sources of noise before complaining, though. */
   if (sr_isError(statres)) {
      DebugInfo fake_di;
      Bool quiet = VG_(strstr)(filename, "/var/run/nscd/") != NULL
                   || VG_(strstr)(filename, "/dev/shm/") != NULL;
      if (!quiet && VG_(clo_verbosity) > 1) { 
         VG_(memset)(&fake_di, 0, sizeof(fake_di));
         fake_di.fsm.filename = ML_(dinfo_strdup)("di.debuginfo.nmm", filename);
         ML_(symerr)(&fake_di, True, "failed to stat64/stat this file");
      }
      return 0;
   }

So we probably should add "/memfd:" to the patterns for which not to
warn.

https://www.man7.org/linux/man-pages/man2/memfd_create.2.html says:

       int memfd_create(const char *name, unsigned int flags);

    The name supplied in name is used as a filename and will be
    displayed as the target of the corresponding symbolic link in the
    directory /proc/self/fd/.  The displayed name is always prefixed
    with memfd: and serves only for debugging purposes.

So I would propose something like the following (untested) patch:

diff --git a/coregrind/m_debuginfo/debuginfo.c b/coregrind/m_debuginfo/debuginfo.c
index 15686fc5c46f..6c4a6926560b 100644
--- a/coregrind/m_debuginfo/debuginfo.c
+++ b/coregrind/m_debuginfo/debuginfo.c
@@ -1196,7 +1196,9 @@ ULong VG_(di_notify_mmap)( Addr a, Bool allow_SkFileV, Int use_fd )
    if (sr_isError(statres)) {
       DebugInfo fake_di;
       Bool quiet = VG_(strstr)(filename, "/var/run/nscd/") != NULL
-                   || VG_(strstr)(filename, "/dev/shm/") != NULL;
+                   || VG_(strstr)(filename, "/dev/shm/") != NULL
+                   || VG_(strncmp)("/memfd:", filename,
+                                   VG_(strlen)("/memfd:")) == 0;
       if (!quiet && VG_(clo_verbosity) > 1) {
          VG_(memset)(&fake_di, 0, sizeof(fake_di));
          fake_di.fsm.filename = ML_(dinfo_strdup)("di.debuginfo.nmm", filename);

Does that work for you?

Thanks,

Mark

Re: [Valgrind-developers] [valgrind] coregrind/m_debuginfo: don't try to examine zero sized mmapped files

[ISHIKAWA,chiaki <ish...@yk...>]

Hi,

Does this help reduce the occurences of presumably non-critical error 
messages like the following?
I get this when I try to run mozilla's thunderbird mail client under 
valgrind.
They appear many times.

   23:44.69 GECKO(2640779) --2640783-- WARNING: Serious error when 
reading debug info
   23:44.69 GECKO(2640779) --2640783-- When reading debug info from 
/memfd:mozilla-ipc (deleted):
   23:44.69 GECKO(2640779) --2640783-- failed to stat64/stat this file
   23:44.69 GECKO(2640779) --2640783-- WARNING: Serious error when 
reading debug info
   23:44.69 GECKO(2640779) --2640783-- When reading debug info from 
/memfd:mozilla-ipc (deleted):
   23:44.69 GECKO(2640779) --2640783-- failed to stat64/stat this file

I have no idea why valgrind wants try to to read debug info from memory 
region.

TIA

Chiaki

On 2025/03/10 0:05, Mark Wielaard wrote:
> https://sourceware.org/git/gitweb.cgi?p=valgrind.git;h=86ac4f2b004f57fa11224efafc1cd1c8fa8ded84
>
> commit 86ac4f2b004f57fa11224efafc1cd1c8fa8ded84
> Author: Mark Wielaard <ma...@kl...>
> Date:   Sun Mar 9 15:59:29 2025 +0100
>
>      coregrind/m_debuginfo: don't try to examine zero sized mmapped files
>      
>      When run on an nfs filesystem memcheck/tests/pointer-trace fails
>      because it generates warnings "connection to image failed". This is
>      caused by trying to mmap a deleted file which the nfs file system
>      represents as a (hidden) regular file. This is normally not a problem
>      except when that file is empty.
>      
>      Fix this by not trying to check whether a file is an ELF or MACHO
>      against an empty (regular) file in di_notify_mmap. An empty file is
>      never a valid ELF or MACHO file (and cannot be represented as
>      DiImage).
>      
>      https://bugs.kde.org/show_bug.cgi?id=501119
>
> Diff:
> ---
>   NEWS                              |  1 +
>   coregrind/m_debuginfo/debuginfo.c | 10 ++++++++--
>   2 files changed, 9 insertions(+), 2 deletions(-)
>
> diff --git a/NEWS b/NEWS
> index 0fcbc5d0e2..16cfeef2a4 100644
> --- a/NEWS
> +++ b/NEWS
> @@ -56,6 +56,7 @@ are not entered into bugzilla tend to get forgotten about or ignored.
>   498492  none/tests/amd64/lzcnt64 crashes on FreeBSD compiled with clang
>   499183  FreeBSD: differences in avx-vmovq output
>   499212  mmap() with MAP_ALIGNED() returns unaligned pointer
> +501119  memcheck/tests/pointer-trace fails when run on NFS filesystem
>   501194  Fix ML_(check_macho_and_get_rw_loads) so that it is correct for any number of segment commands
>   
>   
> diff --git a/coregrind/m_debuginfo/debuginfo.c b/coregrind/m_debuginfo/debuginfo.c
> index 612833a997..15686fc5c4 100644
> --- a/coregrind/m_debuginfo/debuginfo.c
> +++ b/coregrind/m_debuginfo/debuginfo.c
> @@ -1206,8 +1206,14 @@ ULong VG_(di_notify_mmap)( Addr a, Bool allow_SkFileV, Int use_fd )
>      }
>   
>      /* Finally, the point of all this stattery: if it's not a regular file,
> -      don't try to read debug info from it. */
> -   if (! VKI_S_ISREG(statbuf.mode))
> +      don't try to read debug info from it. Also if it is a "regular file"
> +      but has a zero size then skip it. Having a zero size will definitely
> +      fail when trying to create an DiImage and wouldn't be a valid elf or
> +      macho file. This can happen when mmapping a deleted file, which
> +      would normally fail in the check above, because the stat call will
> +      fail.  But if the deleted file is on an NFS file system then a fake
> +      (regular) empty file might be returned.  */
> +   if (! VKI_S_ISREG(statbuf.mode) || statbuf.size == 0)
>         return 0;
>   
>      /* no uses of statbuf below here. */
>
>
> _______________________________________________
> Valgrind-developers mailing list
> Val...@li...
> https://lists.sourceforge.net/lists/listinfo/valgrind-developers

[Valgrind-developers] [valgrind] docs/Makefile.am: Make sure xml catalog file exists for xmllint check

[Mark W. <ma...@so...>]

https://sourceware.org/git/gitweb.cgi?p=valgrind.git;h=9f956db3e5eb0afb0d60987f3658b66646a0ac81

commit 9f956db3e5eb0afb0d60987f3658b66646a0ac81
Author: Mark Wielaard <ma...@kl...>
Date:   Sun Mar 9 16:46:50 2025 +0100

    docs/Makefile.am: Make sure xml catalog file exists for xmllint check
    
    When XML_CATALOG_FILES don't exist on the system xmllint will have to
    query those files through various websites. When there is a network
    error xmllint will fail. So make sure to only run the validity tests
    when both xmllint and XML_CATALOG_FILES exists.

Diff:
---
 docs/Makefile.am | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)

diff --git a/docs/Makefile.am b/docs/Makefile.am
index 12d13db1da..68a3da5617 100644
--- a/docs/Makefile.am
+++ b/docs/Makefile.am
@@ -186,14 +186,20 @@ all-docs: FAQ.txt man-pages html-docs print-docs
 
 valid-manual: $(MANUAL_XML_FILES)
 	if type $(XMLLINT) 2>/dev/null; then \
-	  $(XMLLINT) $(XMLLINT_FLAGS) $(myxmldir)/index.xml && \
-	  touch $@; \
+	  if [ -f $(XML_CATALOG_FILES) ]; then \
+	    export XML_CATALOG_FILES=$(XML_CATALOG_FILES) && \
+	    $(XMLLINT) $(XMLLINT_FLAGS) $(myxmldir)/index.xml && \
+	    touch $@; \
+	  fi \
 	fi
 
 valid-manpages: $(MANPAGES_XML_FILES)
 	if type $(XMLLINT) 2>/dev/null; then \
-	  $(XMLLINT) $(XMLLINT_FLAGS) $(myxmldir)/manpages-index.xml && \
-	  touch $@; \
+	  if [ -f $(XML_CATALOG_FILES) ]; then \
+	    export XML_CATALOG_FILES=$(XML_CATALOG_FILES) && \
+	    $(XMLLINT) $(XMLLINT_FLAGS) $(myxmldir)/manpages-index.xml && \
+	    touch $@; \
+	  fi \
 	fi
 
 check-local: valid-manual valid-manpages

[Valgrind-developers] [valgrind] coregrind/m_debuginfo: don't try to examine zero sized mmapped files

[Mark W. <ma...@so...>]

https://sourceware.org/git/gitweb.cgi?p=valgrind.git;h=86ac4f2b004f57fa11224efafc1cd1c8fa8ded84

commit 86ac4f2b004f57fa11224efafc1cd1c8fa8ded84
Author: Mark Wielaard <ma...@kl...>
Date:   Sun Mar 9 15:59:29 2025 +0100

    coregrind/m_debuginfo: don't try to examine zero sized mmapped files
    
    When run on an nfs filesystem memcheck/tests/pointer-trace fails
    because it generates warnings "connection to image failed". This is
    caused by trying to mmap a deleted file which the nfs file system
    represents as a (hidden) regular file. This is normally not a problem
    except when that file is empty.
    
    Fix this by not trying to check whether a file is an ELF or MACHO
    against an empty (regular) file in di_notify_mmap. An empty file is
    never a valid ELF or MACHO file (and cannot be represented as
    DiImage).
    
    https://bugs.kde.org/show_bug.cgi?id=501119

Diff:
---
 NEWS                              |  1 +
 coregrind/m_debuginfo/debuginfo.c | 10 ++++++++--
 2 files changed, 9 insertions(+), 2 deletions(-)

diff --git a/NEWS b/NEWS
index 0fcbc5d0e2..16cfeef2a4 100644
--- a/NEWS
+++ b/NEWS
@@ -56,6 +56,7 @@ are not entered into bugzilla tend to get forgotten about or ignored.
 498492  none/tests/amd64/lzcnt64 crashes on FreeBSD compiled with clang
 499183  FreeBSD: differences in avx-vmovq output
 499212  mmap() with MAP_ALIGNED() returns unaligned pointer
+501119  memcheck/tests/pointer-trace fails when run on NFS filesystem
 501194  Fix ML_(check_macho_and_get_rw_loads) so that it is correct for any number of segment commands
 
 
diff --git a/coregrind/m_debuginfo/debuginfo.c b/coregrind/m_debuginfo/debuginfo.c
index 612833a997..15686fc5c4 100644
--- a/coregrind/m_debuginfo/debuginfo.c
+++ b/coregrind/m_debuginfo/debuginfo.c
@@ -1206,8 +1206,14 @@ ULong VG_(di_notify_mmap)( Addr a, Bool allow_SkFileV, Int use_fd )
    }
 
    /* Finally, the point of all this stattery: if it's not a regular file,
-      don't try to read debug info from it. */
-   if (! VKI_S_ISREG(statbuf.mode))
+      don't try to read debug info from it. Also if it is a "regular file"
+      but has a zero size then skip it. Having a zero size will definitely
+      fail when trying to create an DiImage and wouldn't be a valid elf or
+      macho file. This can happen when mmapping a deleted file, which
+      would normally fail in the check above, because the stat call will
+      fail.  But if the deleted file is on an NFS file system then a fake
+      (regular) empty file might be returned.  */
+   if (! VKI_S_ISREG(statbuf.mode) || statbuf.size == 0)
       return 0;
 
    /* no uses of statbuf below here. */