[Crash-utility] [ANNOUNCE] crash-7.0.2 is available

Download from: http://people.redhat.com/anderson Changelog: - Added "bison" to the BuildRequires line of the crash.spec file. Without the patch, the build of the embedded gdb-7.6 module will fail unless either /usr/bin/bison or /usr/bin/yacc are available. The failure will result in a stream of error messages from different files that indicate: multiple definition of 'main' undefined reference to 'c_parse_escape' undefined reference to 'ada_parse' undefined reference to 'ada_error' undefined reference to 'c_parse' undefined reference to 'c_error' undefined reference to 'cp_demangled_name_to_comp' undefined reference to 'cp_demangled_name_parse_free' undefined reference to 'cp_comp_to_string' undefined reference to 'cp_new_demangle_parse_info' and the build fails like so: collect2: ld returned 1 exit status make[4]: *** [gdb] Error 1 crash build failed If building with rpmbuild, the new BuildRequires "bison" entry will prevent the build from initiating unless the bison package has been installed. If building with the tar.gz file, the build attempt will proceed and fail unless either the bison or byacc (Berkeley Yacc) package is installed. (anderson(a)redhat.com) - Fix the S390X initialization sequence on kernels that are configured with CONFIG_STRICT_DEVMEM to automatically try /proc/kcore if: (1) the /dev/crash driver is not available, and (2) the initial /dev/mem access fails. Without the patch, if /dev/mem is selected as the memory source and it is restricted, the crash session will fail during initialization with the error message "crash: read error: kernel virtual address: <address> type: cpu_possible_mask". (anderson(a)redhat.com) - When checking whether a argument on the crash command line is a dumpfile that may be in makedumpfile's "flattened" format, do not bother checking character device files. (anderson(a)redhat.com) - Fix for the PPC64 virtual-to-physical virtual address translation mechanism for vmalloc and user-space virtual addresses on Linux 3.10 and later kernels. Without the patch, the message "WARNING: cannot access vmalloc'd module memory" is displayed during initialization, and during the crash session, if a command attempts to translate or read a vmalloc or user-space virtual address, it will fail. (anderson(a)redhat.com) - Clean up all files that emit "warning: format not a string literal and no format arguments" when compiled with -Wformat-security warning option. All instances of fprintf, sprintf and snprintf using the format "fprintf(fp, buf)" are replaced with "fprintf(fp, "%s", buf)". Also, the -Wformat-security warning option has been added to the option list used when compiling with "make warn". (stefan.bader(a)canonical.com, anderson(a)redhat.com) - Fix a build failure when compiling with very old gcc-3.4.6 version on a 2.6.9-based RHEL4 IA64 host. The bfd library in gdb-7.6 is compiled with the -Werror option, and it fails with the message "elflink.c:4733: warning: 'idx' might be used uninitialized in this function". (anderson(a)redhat.com) - Fix a build failure when compiling with very old gcc-3.4.6 version on a 2.6.9-based RHEL4 S390 or S390X hosts. The embedded gdb-7.6 fails to compile with the error message "s390-nat.c:364: error: storage size of 'iov' isn't known". (anderson(a)redhat.com) - Fix to properly store two-digit kernel version numbers. (timo.lindfors(a)iki.fi) - Fix to provide hugepage address translation for the "vtop" command on the PPC64 architecture. (hbathini(a)linux.vnet.ibm.com) - Fix for the "log" command to account for the kernel data structure name change from "log" to "printk_log" in Linux 3.11-rc4 and later kernels. Without the patch, the message "WARNING: log buf data structure(s) have changed" will be displayed during initialization and by the "log" command. (anatol.pomozov(a)gmail.com) - Fix to add a linefeed after the description of the "kmem -I" option in the "help kmem" output, which was recently added in crash-7.0.0. (anderson(a)redhat.com) - Document the "-s" command line option in the "crash -h|--help" output and in the crash.8 man page to also indicate that runtime command scrolling is turned off by default. (anderson(a)redhat.com) - Fix for the "irq -d" option on 2.6.25 and later X86_64 kernels to display the Intel interrupt descriptor table contents. Without the patch, those kernel versions would display "irq: -d option not supported or applicable on this architecture or kernel". (anatol.pomozov(a)gmail.com) - Fix for the "kmem -[sS]" options on Linux 3.11-rc1 and later kernels that are configured with CONFIG_SLAB. Without the patch, the command fails with the error message "kmem: invalid structure member offset: kmem_cache_s_lists". (anatol.pomozov(a)gmail.com) - Fix for the "kmem <address>" and the "bt -F" options on Linux 3.8 and later kernels that are configured with CONFIG_SLUB. Without the patch, the command would fail with the error message "kmem: invalid structure member offset: page_slab". (anderson(a)redhat.com) - Fix misspellings in the "bt" and "search" help page output. (anatol.pomozov(a)gmail.com) - Fix for the determination of the base of the kernel's unity-mapped virtual address region on recent ARM kernels whose "_stext" variable address has changed from 0xc0008000 to 0xc0100000. Without the patch, the crash session fails during invocation with the error message "crash: vmlinux and vmcore do not match!". (Jan.Karlsson(a)sonymobile.com) - When printing data structures, prevent the embedded gdb from symbolically translating pointers that are not kernel virtual addresses. Kernel or module symbols that are not virtual addresses can be mistaken for virtual addresses, leading to NULL pointers being invalidly translated into a symbol name from the vmlinux or module object file. For example, in X86_64 kernels, NULL pointers are translated into the symbol "irq_stack_union", whose value is not a virtual address, but rather a per-cpu offset value of 0. (anderson(a)redhat.com) - Fix for the "kmem -s <address>" or "kmem <address>" options on Linux 3.11 and later kernels configured with CONFIG_SLAB. Without the patch, both commands fail with the error message "kmem: cannot resolve cache_cache". (anderson(a)redhat.com) - Fix to prevent the "bt" command from generating a segmentation violation in a case where the per-cpu "current_task" variable and the runqueue's "curr" variable did not agree, and the panic task had overflowed its kernel stack. This led to the selection of the a starting RSP address which belonged to the other task; without the patch, the command generated a segmentation violation after printing the first frame of the backtrace. (anderson(a)redhat.com)