- Devel - libvirt List Archives

[libvirt] [PATCH] proxy: Fix use of uninitalized memory
by Rasputin 28 Jan '09

28 Jan '09

On short read, members of packet header are checked before actually read. If uninitialized values can pass the test, they can be set to arbitrary values while reading remaining portion of a packet. Buffer overflow is possible. libvirt_proxy is suid-root. diff -urp libvirt-0.5.1/proxy/libvirt_proxy.c libvirt-dev/proxy/libvirt_proxy.c --- libvirt-0.5.1/proxy/libvirt_proxy.c 2008-11-20 08:58:43.000000000 +0100 +++ libvirt-dev/proxy/libvirt_proxy.c 2009-01-25 12:51:33.000000000 +0100 @@ -385,7 +385,8 @@ retry: fprintf(stderr, "read %d bytes from client %d on socket %d\n", ret, nr, pollInfos[nr].fd); - if ((req->version != PROXY_PROTO_VERSION) || + if ((ret != sizeof(virProxyPacket)) || + (req->version != PROXY_PROTO_VERSION) || (req->len < sizeof(virProxyPacket)) || (req->len > sizeof(virProxyFullPacket))) goto comm_error;

2 3

[libvirt] PATCH: Detect SASL library on Solaris
by Daniel P. Berrange 28 Jan '09

28 Jan '09

I did a test build of CVS on Solaris and noticed it didn't detect SASL, even though SASL was available. It turns out the library on SOlaris is called libsasl.so, instead of libsasl2.so. It is still API compatible with the Linux one, so there does not seem to be a reason not to use it So this patch extends the configure test to make this work Daniel Index: configure.in =================================================================== RCS file: /data/cvs/libvirt/configure.in,v retrieving revision 1.202 diff -u -p -u -p -r1.202 configure.in --- configure.in 27 Jan 2009 15:29:53 -0000 1.202 +++ configure.in 28 Jan 2009 11:38:25 -0000 @@ -517,18 +517,26 @@ if test "x$with_sasl" != "xno"; then fail=1 fi]) if test "x$with_sasl" != "xno" ; then - AC_CHECK_LIB([sasl2], [sasl_client_init],[with_sasl=yes],[ - if test "x$with_sasl" = "xcheck" ; then + AC_CHECK_LIB([sasl2], [sasl_client_init],[ + SASL_LIBS="$SASL_LIBS -lsasl2" + with_sasl=yes + ],[ + AC_CHECK_LIB([sasl], [sasl_client_init],[ + SASL_LIBS="$SASL_LIBS -lsasl" + with_sasl=yes + ],[ + if test "x$with_sasl" = "xcheck" ; then with_sasl=no - else + else fail=1 - fi]) + fi + ]) + ]) fi test $fail = 1 && AC_MSG_ERROR([You must install the Cyrus SASL development package in order to compile libvirt]) CFLAGS="$old_cflags" LIBS="$old_libs" - SASL_LIBS="$SASL_LIBS -lsasl2" if test "x$with_sasl" = "xyes" ; then AC_DEFINE_UNQUOTED([HAVE_SASL], 1, [whether Cyrus SASL is available for authentication]) -- |: Red Hat, Engineering, London -o- http://people.redhat.com/berrange/ :| |: http://libvirt.org -o- http://virt-manager.org -o- http://ovirt.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: GnuPG: 7D3B9505 -o- F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|

1 0

[libvirt] PATCH: Reset global error object at end of test
by Daniel P. Berrange 28 Jan '09

28 Jan '09

There statstest test case is expected to raise a number of libvirt errors. Since these are now stored in thread locals, it is expected that this won't be free'd automatically at system shutdown. This patch adds a call to virResetError() at the end to release the memory associated with the stored error object. There is still one block of memory alocated though, which is the thread local error error object itself. We can't free this easily, so I just add a valgrind suppression rule instead. This lets 'make valgrind' pass again Also since the test case doesn't call virInitialize()as a normal app would do, we need to explicitly initialize the thread local storage to ensure things work correctly. Daniel diff -r 93337ec227d6 tests/.valgrind.supp --- a/tests/.valgrind.supp Mon Jan 26 15:25:53 2009 +0000 +++ b/tests/.valgrind.supp Tue Jan 27 11:15:47 2009 +0000 @@ -238,3 +238,18 @@ fun:virtTestRun fun:mymain } +{ + ignoreThreadLocalErrorObject + Memcheck:Leak + fun:calloc + fun:virAlloc + fun:virLastErrorObject + fun:virRaiseError + fun:statsErrorFunc + fun:xenLinuxDomainDeviceID + fun:testDeviceHelper + fun:virtTestRun + fun:mymain + fun:virtTestMain + fun:main +} diff -r 93337ec227d6 tests/testutils.c --- a/tests/testutils.c Mon Jan 26 15:25:53 2009 +0000 +++ b/tests/testutils.c Tue Jan 27 11:15:47 2009 +0000 @@ -26,6 +26,8 @@ #include "internal.h" #include "memory.h" #include "util.h" +#include "threads.h" +#include "virterror_internal.h" #if TEST_OOM_TRACE #include <execinfo.h> @@ -319,8 +321,8 @@ int virtTestMain(int argc, int (*func)(int, char **)) { char *debugStr; + int ret; #if TEST_OOM - int ret; int approxAlloc = 0; int n; char *oomStr = NULL; @@ -330,6 +332,10 @@ int virtTestMain(int argc, int worker = 0; #endif + if (virThreadInitialize() < 0 || + virErrorInitialize() < 0) + return 1; + if ((debugStr = getenv("VIR_TEST_DEBUG")) != NULL) { if (virStrToLong_ui(debugStr, NULL, 10, &testDebug) < 0) testDebug = 0; @@ -349,8 +355,10 @@ int virtTestMain(int argc, if (getenv("VIR_TEST_MP") != NULL) { mp = sysconf(_SC_NPROCESSORS_ONLN); fprintf(stderr, "Using %d worker processes\n", mp); - if (VIR_ALLOC_N(workers, mp) < 0) - return EXIT_FAILURE; + if (VIR_ALLOC_N(workers, mp) < 0) { + ret = EXIT_FAILURE; + goto cleanup; + } } if (testOOM) @@ -359,7 +367,7 @@ int virtTestMain(int argc, /* Run once to count allocs, and ensure it passes :-) */ ret = (func)(argc, argv); if (ret != EXIT_SUCCESS) - return EXIT_FAILURE; + goto cleanup; #if TEST_OOM_TRACE if (testDebug) @@ -431,9 +439,11 @@ int virtTestMain(int argc, else fprintf(stderr, " FAILED\n"); } +cleanup: +#else + ret = (func)(argc, argv); +#endif + + virResetLastError(); return ret; - -#else - return (func)(argc, argv); -#endif } -- |: Red Hat, Engineering, London -o- http://people.redhat.com/berrange/ :| |: http://libvirt.org -o- http://virt-manager.org -o- http://ovirt.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: GnuPG: 7D3B9505 -o- F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|

2 1

[libvirt] [PATCH] Fix stderr spewage from docs/examples install
by john.levon＠sun.com 28 Jan '09

28 Jan '09

# HG changeset patch # User John Levon <john.levon(a)sun.com> # Date 1232970961 28800 # Node ID 839f1721d6d9c2cfdbc9327814a4c4a9564ff814 # Parent 26d337992a98a00d98f83247eb113ca3d729d8ef Fix stderr spewage from docs/examples install *.res don't exist, so shouldn't be in the generated Makefile.am Signed-off-by: John Levon <john.levon(a)sun.com> diff --git a/docs/examples/Makefile.am b/docs/examples/Makefile.am --- a/docs/examples/Makefile.am +++ b/docs/examples/Makefile.am @@ -20,7 +20,7 @@ install-data-local: install-data-local: $(mkinstalldirs) $(DESTDIR)$(HTML_DIR) -@INSTALL@ -m 0644 $(srcdir)/*.html $(srcdir)/*.c $(srcdir)/*.xml \ - $(srcdir)/*.xsl $(srcdir)/*.res $(DESTDIR)$(HTML_DIR) + $(srcdir)/*.xsl $(DESTDIR)$(HTML_DIR) EXTRA_DIST=examples.xsl index.py examples.xml diff --git a/docs/examples/index.py b/docs/examples/index.py --- a/docs/examples/index.py +++ b/docs/examples/index.py @@ -241,7 +241,7 @@ install-data-local: install-data-local: $(mkinstalldirs) $(DESTDIR)$(HTML_DIR) -@INSTALL@ -m 0644 $(srcdir)/*.html $(srcdir)/*.c $(srcdir)/*.xml \\ - $(srcdir)/*.xsl $(srcdir)/*.res $(DESTDIR)$(HTML_DIR) + $(srcdir)/*.xsl $(DESTDIR)$(HTML_DIR) """ EXTRA_DIST= string.join(extras, ' ')

2 1

[libvirt] [PATCH] Fix misuse of PF_UNIX
by john.levon＠sun.com 28 Jan '09

28 Jan '09

# HG changeset patch # User john.levon(a)sun.com # Date 1233105335 28800 # Node ID b3a2537e2f3d5ccb055df1820c112b469a26efdb # Parent 35f5d4f77a3f50cadacf32100fdbd992394f6002 Fix misuse of PF_UNIX PF_UNIX is a protocol familay, not a valid protocol, so isn't suitable for the 3rd socket() argument. Solaris should be ignoring the invalid protocol anyway, but this fix is correct regardless. Signed-off-by: John Levon <john.levon(a)sun.com> diff --git a/src/xend_internal.c b/src/xend_internal.c --- a/src/xend_internal.c +++ b/src/xend_internal.c @@ -752,7 +752,11 @@ xenDaemonOpen_unix(virConnectPtr conn, c memset(&priv->addr, 0, sizeof(priv->addr)); priv->addrfamily = AF_UNIX; - priv->addrprotocol = PF_UNIX; + /* + * This must be zero on Solaris at least for AF_UNIX (which should + * really be PF_UNIX, but doesn't matter). + */ + priv->addrprotocol = 0; priv->addrlen = sizeof(struct sockaddr_un); addr = (struct sockaddr_un *)&priv->addr;

3 3

[libvirt] [PATCH] Install schemas into correct location for Solaris
by john.levon＠sun.com 28 Jan '09

28 Jan '09

# HG changeset patch # User john.levon(a)sun.com # Date 1233119659 28800 # Node ID cc848242fa810e8d0126b651849de715c7ef32e4 # Parent f91041f0c607be72c4b5695f081d20716521f6c5 Install schemas into correct location for Solaris Signed-off-by: John Levon <john.levon(a)sun.com> diff --git a/configure.in b/configure.in --- a/configure.in +++ b/configure.in @@ -141,6 +141,10 @@ if test -n "$UDEVSETTLE"; then AC_DEFINE_UNQUOTED([UDEVSETTLE],["$UDEVSETTLE"], [Location or name of the udevsettle program]) fi + +SCHEMA_DIR=${pkgdatadir}/schemas +test `uname -s` = SunOS && SCHEMA_DIR=${datadir}/lib/xml/rng/libvirt +AC_SUBST([SCHEMA_DIR]) dnl Specific dir for HTML output ? AC_ARG_WITH([html-dir], [AC_HELP_STRING([--with-html-dir=path], diff --git a/docs/schemas/Makefile.am b/docs/schemas/Makefile.am --- a/docs/schemas/Makefile.am +++ b/docs/schemas/Makefile.am @@ -1,6 +1,7 @@ -schemadir = $(pkgdatadir)/schemas +schemadir = $(SCHEMA_DIR) + schema_DATA = \ domain.rng \ network.rng \

3 2

[libvirt] [PATCH] avoid printf format-mismatch warnings
by Jim Meyering 28 Jan '09

28 Jan '09

autobuild.sh fails like this on at least RHEL5.3: cc1: warnings being treated as errors qemud.c: In function 'qemudClientReadBuf': qemud.c:1470: warning: format '%d' expects type 'int', but argument 7 has type 'ssize_t' qemud.c: In function 'qemudClientWriteBuf': qemud.c:1695: warning: format '%d' expects type 'int', but argument 7 has type 'ssize_t' make[2]: *** [libvirtd-qemud.o] Error 1 so I'm about to apply the following fix: >From 162177b69fec410b7d940cf9242cba9b147f0bdb Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Wed, 28 Jan 2009 12:08:26 +0100 Subject: [PATCH] avoid printf format-mismatch warnings * qemud/qemud.c (qemudClientReadBuf, qemudClientWriteBuf): Use %lld and a (long long int) cast to print a ssize_t value. --- qemud/qemud.c | 6 ++++-- 1 files changed, 4 insertions(+), 2 deletions(-) diff --git a/qemud/qemud.c b/qemud/qemud.c index eb91533..fa5e17d 100644 --- a/qemud/qemud.c +++ b/qemud/qemud.c @@ -1467,7 +1467,8 @@ static ssize_t qemudClientReadBuf(struct qemud_client *client, ssize_t ret; if (len < 0) { - VIR_ERROR(_("unexpected negative length request %d"), len); + VIR_ERROR(_("unexpected negative length request %lld"), + (long long int) len); qemudDispatchClientFailure(client); return -1; } @@ -1692,7 +1693,8 @@ static ssize_t qemudClientWriteBuf(struct qemud_client *client, ssize_t ret; if (len < 0) { - VIR_ERROR(_("unexpected negative length request %d"), len); + VIR_ERROR(_("unexpected negative length request %lld"), + (long long int) len); qemudDispatchClientFailure(client); return -1; } -- 1.6.1.1.374.g0d9d7

1 0

[libvirt] use virReportOOMError, not VIR_ERR_NO_MEMORY
by Jim Meyering 27 Jan '09

27 Jan '09

I've changed over 200 uses of <some_error_function> with the VIR_ERR_NO_MEMORY flag to call virReportOOMError with whatever "conn"-like argument (if any) was in the original call. I don't claim to have tested these changes (OOM is a pain to simulate, and besides, testing so many failure points would take more infrastructure than we have right now). However, I did automate most of the process. most of the changes were performed automatically via this: perl -0x0 -pi -e \ 's/\w+( ?)$([\w>-]+), *(?:(?:VIR_FROM_NONE|domcode|vm|dom|net|NULL), ?){0,3}VIR_ERR_NO_MEMORY,.*?$;/virReportOOMError$1($2);/sg' \ $(g grep -l VIR_ERR_NO_MEMORY|grep '\.c$'|grep -vF virterror.c) Exceptions: ====================== Exceptions (as noted in ChangeLog) were to add the now-required definitions of VIR_FROM_THIS, as needed, e.g., #define VIR_FROM_THIS VIR_FROM_OPENVZ in openvz-related files. Also, I replaced uses of qparam_report_oom automatically: perl -pi -e 's/qparam_report_oom;/virReportOOMError(NULL);/' \ src/qparams.c and removed the definition of that macro manually. I spotted and manually adjusted one misuse of VIR_ERR_NO_MEMORY: remote_internal.c: fix typo that would mistakenly report OOM * src/remote_internal.c (addrToString): Report VIR_ERR_UNKNOWN_HOST, not VIR_ERR_NO_MEMORY. ctxt in src/conf.c is not usable: perl -pi -e 's/virReportOOMError$ctxt$;/virReportOOMError(NULL);/' \ src/conf.c I manually converted two uses of virSexprError(VIR_ERR_NO_MEMORY in src/sexpr.c. And of course, copyright dates were updated via an emacs write hook. The first iteration involved applying the perl-based xforms more of less file by file, and making sure everything still compiled. Once done, I repeated the process on a new branch, automating as much as possible, and then diff'd the branches. The diff was nearly empty, modulo copyright dates and added VIR_FROM_THIS definitions. I'm posting two small sort-of-different changes here. I'll post this big first one separately. >From 099536470ae2cbe9503ed471d391e403fb2587a4 Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Tue, 27 Jan 2009 12:20:06 +0100 Subject: [PATCH 1/3] error-reporting calls using VIR_ERR_NO_MEMORY: use virReportOOMError instead * src/uml_conf.c (VIR_FROM_THIS): Define to VIR_FROM_UML. * src/xs_internal.c (VIR_FROM_THIS): Define to VIR_FROM_XEN. * src/xml.c (VIR_FROM_THIS): Define to VIR_FROM_XML. * src/stats_linux.c (VIR_FROM_THIS): Define to VIR_FROM_STATS_LINUX. * src/datatypes.c (VIR_FROM_THIS): Define to VIR_FROM_NONE. * src/lxc_conf.c (VIR_FROM_THIS): Define to VIR_FROM_LXC. * src/libvirt.c (VIR_FROM_THIS): Define to VIR_FROM_NONE. * src/node_device_conf.c (VIR_FROM_THIS): Define to VIR_FROM_NODEDEV. * src/openvz_conf.c (VIR_FROM_THIS): Define to VIR_FROM_OPENVZ. * src/openvz_driver.c (VIR_FROM_THIS): Define to VIR_FROM_OPENVZ. * src/conf.c (VIR_FROM_THIS): Define to VIR_FROM_CONF. Note: this loses config_filename:config_lineno diagnostics, but that's ok. * src/node_device.c (VIR_FROM_THIS): Define to VIR_FROM_NODEDEV. * src/sexpr.c (VIR_FROM_THIS): Define to VIR_FROM_SEXPR. --- ... >From a223b8f6064e867d304df9a0a1498ef2940631ef Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Tue, 27 Jan 2009 15:58:07 +0100 Subject: [PATCH 2/3] qparams.c: Use virReportOOMError(NULL), not qparam_report_oom() * src/qparams.c (VIR_FROM_THIS): Define to VIR_FROM_NONE. (qparam_report_oom): Remove definition. Replace all uses. --- src/qparams.c | 22 ++++++++++------------ 1 files changed, 10 insertions(+), 12 deletions(-) diff --git a/src/qparams.c b/src/qparams.c index 22c5853..d0a84b3 100644 --- a/src/qparams.c +++ b/src/qparams.c @@ -1,4 +1,4 @@ -/* Copyright (C) 2007 Red Hat, Inc. +/* Copyright (C) 2007, 2009 Red Hat, Inc. * * This library is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public @@ -33,9 +33,7 @@ #include "memory.h" #include "qparams.h" -#define qparam_report_oom(void) \ - virReportErrorHelper(NULL, VIR_FROM_NONE, VIR_ERR_NO_MEMORY, \ - __FILE__, __FUNCTION__, __LINE__, NULL) +#define VIR_FROM_THIS VIR_FROM_NONE struct qparam_set * new_qparam_set (int init_alloc, ...) @@ -47,14 +45,14 @@ new_qparam_set (int init_alloc, ...) if (init_alloc <= 0) init_alloc = 1; if (VIR_ALLOC(ps) < 0) { - qparam_report_oom(); + virReportOOMError(NULL); return NULL; } ps->n = 0; ps->alloc = init_alloc; if (VIR_ALLOC_N(ps->p, ps->alloc) < 0) { VIR_FREE (ps); - qparam_report_oom(); + virReportOOMError(NULL); return NULL; } @@ -98,7 +96,7 @@ grow_qparam_set (struct qparam_set *ps) { if (ps->n >= ps->alloc) { if (VIR_REALLOC_N(ps->p, ps->alloc * 2) < 0) { - qparam_report_oom(); + virReportOOMError(NULL); return -1; } ps->alloc *= 2; @@ -115,14 +113,14 @@ append_qparam (struct qparam_set *ps, pname = strdup (name); if (!pname) { - qparam_report_oom(); + virReportOOMError(NULL); return -1; } pvalue = strdup (value); if (!pvalue) { VIR_FREE (pname); - qparam_report_oom(); + virReportOOMError(NULL); return -1; } @@ -156,7 +154,7 @@ qparam_get_query (const struct qparam_set *ps) } if (virBufferError(&buf)) { - qparam_report_oom(); + virReportOOMError(NULL); return NULL; } @@ -184,7 +182,7 @@ qparam_query_parse (const char *query) ps = new_qparam_set (0, NULL); if (!ps) { - qparam_report_oom(); + virReportOOMError(NULL); return NULL; } @@ -257,7 +255,7 @@ qparam_query_parse (const char *query) return ps; out_of_memory: - qparam_report_oom(); + virReportOOMError(NULL); free_qparam_set (ps); return NULL; } -- 1.6.1.1.363.g2a3bd >From dd9477353e930ba75dea1e06ccbced5b1cb61e09 Mon Sep 17 00:00:00 2001 From: Jim Meyering <meyering(a)redhat.com> Date: Tue, 27 Jan 2009 15:53:27 +0100 Subject: [PATCH 3/3] remote_internal.c: fix typo that would mistakenly report OOM * src/remote_internal.c (addrToString): Report VIR_ERR_UNKNOWN_HOST, not VIR_ERR_NO_MEMORY. --- src/remote_internal.c | 11 +++++------ 1 files changed, 5 insertions(+), 6 deletions(-) diff --git a/src/remote_internal.c b/src/remote_internal.c index 8ed52bc..5c2e705 100644 --- a/src/remote_internal.c +++ b/src/remote_internal.c @@ -2,7 +2,7 @@ * remote_internal.c: driver to provide access to libvirtd running * on a remote machine * - * Copyright (C) 2007-2008 Red Hat, Inc. + * Copyright (C) 2007-2009 Red Hat, Inc. * * This library is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public @@ -4949,10 +4949,10 @@ static char *addrToString(struct sockaddr_storage *sa, socklen_t salen) port, sizeof(port), NI_NUMERICHOST | NI_NUMERICSERV)) != 0) { virRaiseError (NULL, NULL, NULL, VIR_FROM_REMOTE, - VIR_ERR_NO_MEMORY, VIR_ERR_ERROR, - NULL, NULL, NULL, 0, 0, - _("Cannot resolve address %d: %s"), - err, gai_strerror(err)); + VIR_ERR_UNKNOWN_HOST, VIR_ERR_ERROR, + NULL, NULL, NULL, 0, 0, + _("Cannot resolve address %d: %s"), + err, gai_strerror(err)); return NULL; } @@ -6836,4 +6836,3 @@ remoteRegister (void) return 0; } - -- 1.6.1.1.363.g2a3bd

2 1

Re: [libvirt] Clock problems with Windows guests and Xen
by Matthew Donovan 27 Jan '09

27 Jan '09

> On Tue, Jan 27, 2009 at 01:33:21PM -0500, Matthew Donovan wrote: > > When I start Windows guests with libvirt my clocks are 5 hours off. When I > > use a Xen configuration file, I can specify an offset (rtc_timeoffset = > > -18000) to get my clock to sync with the host. Is there a way to do specify > > this offset in XML with libvirt? > > > > Right now, I'm using version 0.4.4 but I downloaded the source for 0.5.1 and > > it looks like that part of domain_conf.c hasn't changed. > Sounds like you have the wrong clock offset setting, in the XML. There > are two options: > <clock offset="localtime"/> > <clock offset="utc"/> > Usually for Windows you want the 'localtime' one since it is dumb at time > keeping & DST shifts. > Daniel That seems to have done it. Thanks! -matthew

1 0

[libvirt] Clock problems with Windows guests and Xen
by Matthew Donovan 27 Jan '09

27 Jan '09

When I start Windows guests with libvirt my clocks are 5 hours off. When I use a Xen configuration file, I can specify an offset (rtc_timeoffset = -18000) to get my clock to sync with the host. Is there a way to do specify this offset in XML with libvirt? Right now, I'm using version 0.4.4 but I downloaded the source for 0.5.1 and it looks like that part of domain_conf.c hasn't changed. Thanks for any help! -matthew

2 1