Andrei Borzenkov [Fri, 28 Nov 2014 18:12:00 +0000 (21:12 +0300)]
grub-core/disk/luks.c: fix use after free and memory leaks
configure_ciphers:
- several memory leaks where allocated ciphers were not freed. CID: 73813,
73710
- use after free. It is probably quite innocent as grub is single threaded,
but could potentially be a problem with memory allocator debugger turned on.
CID: 73730
Michael Chang [Tue, 18 Nov 2014 08:03:08 +0000 (16:03 +0800)]
send router solicitation for ipv6 address autoconf v2
Many routers have long router advertisment interval configured by
default. The Neighbor Discovery protocol (RFC4861) has defined default
MaxRtrAdvInterval value as 600 seconds and
MinRtrAdvInterval as 0.33*MaxRtrAdvInterval. This makes
net_ipv6_autoconf fails more often than not as currently it passively
listens the RA message to perfom address autoconfiguration.
This patch tries to send router solicitation to overcome the problem of
long RA interval.
v2:
use cpu_to_be macro for network byte order conversion
add missing error handling
Leif Lindholm [Mon, 17 Nov 2014 14:11:01 +0000 (14:11 +0000)]
efi: check *path non-null before grub_strrchr
The EFI version of grub_machine_get_bootlocation crops the boot image
name back to the last / in order to get a directory path. However, it
does not check that *name is actually set before calling grub_strrchr
to do this, and neither does grub_strrchr before dereferencing a NULL
pointer.
Parent function, grub_set_prefix_and_root, does check the pointer
before using.
Michael Chang [Mon, 3 Nov 2014 08:27:45 +0000 (16:27 +0800)]
icmp6 fix no respond to neighbor solicit message
The structure size used in grub_netbuff_pull to get the pointer to
option header is apparently wrong, which leads to subsequent range check
failed and therefore not responding to any neighbor solicit message in my
testing.
Andrei Borzenkov [Tue, 14 Oct 2014 16:12:15 +0000 (20:12 +0400)]
Use full initializer for initrd_ctx to avoid fatal warnings with older GCC
struct ... foo = { 0, } is valid initializer, but older GCC emits
warning which is fatal error due to -Werror=missing-field-initializer.
So simply use full initializer to avoid these errors. This was fixed
probably in GCC 4.7.
See https://gcc.gnu.org/bugzilla/show_bug.cgi?id=36750
Colin Watson [Tue, 23 Sep 2014 11:06:30 +0000 (12:06 +0100)]
Add a new "none" platform that only builds utilities
This makes it possible to build generally-useful utilities such as
grub-mount even if the rest of GRUB has not been ported to the target
CPU.
* configure.ac: Add "none" platform. Default to it for unsupported
CPUs rather than stopping with a fatal error. Don't downgrade
x86_64-none to i386. Define COND_real_platform Automake conditional
if the platform is anything other than "none". Don't do any include
directory linking for "none".
* Makefile.am: Skip building grub-core and all bootcheck targets if
!COND_real_platform.
* include/grub/time.h: Don't include <grub/cpu/time.h> if GRUB_UTIL
is defined.
The AML parser implements only a small subset of possible AML
opcodes. On the Fujitsu Lifebook E744 this and another bug in
the parser (incorrect handling of TermArg data types) would lead
to the laptop not turning off (_S5 not found).
* grub-core/commands/acpihalt.c: Support OpAlias in the AML parser;
in skip_ext_op(), handle some Type2Opcodes more correctly (TermArgs
aren't always simply strings!); Add function to skip TermArgs
* include/grub/acpi.h: Add new opcodes
Peter Jones [Wed, 17 Sep 2014 20:30:11 +0000 (16:30 -0400)]
Initialized initrd_ctx so we don't free a random pointer from the stack.
Currently, if "linux" fails, the "goto fail;" in grub_cmd_initrd sends us
into grub_initrd_close() without grub_initrd_init() being called, and thus
it never clears initrd_ctx->components. grub_initrd_close() then frees that
address, which is stale data from the stack. If the stack happens to have a
stale *address* there that matches a recent allocation, then you'll get a
double free later.
Michael Chang [Thu, 14 Aug 2014 10:17:45 +0000 (18:17 +0800)]
Fix incorrect address reference in btrfs
We encountered a weird random kernel initrd unpacking error on btrfs
and finally found it was caused by incorrect address reference in range
check for type GRUB_BTRFS_EXTENT_REGULAR and the entire result is
unpredictable.
This is a quick fix to make the address reference to the
grub_btrfs_extent_data structure correctly, not the pointer variable
to it.
Colin Watson [Sun, 7 Sep 2014 22:04:35 +0000 (23:04 +0100)]
Support grub-emu on x32 (ILP32 but with x86-64 instruction set)
* configure.ac: Remove -m64 from checks for -mcmodel=large and
-mno-red-zone. These are always either unnecessary (x86_64-emu) or
already in TARGET_CFLAGS at this point, and they produce incorrect
results when building for x32.
* grub-core/kern/x86_64/dl.c (grub_arch_dl_relocate_symbols): Cast
pointers to Elf64_Xword via grub_addr_t, in order to work on x32.
* include/grub/x86_64/types.h (GRUB_TARGET_SIZEOF_VOID_P,
GRUB_TARGET_SIZEOF_LONG): Define to 4 on x32.
Colin Watson [Thu, 10 Apr 2014 15:54:33 +0000 (16:54 +0100)]
Tolerate devices with no filesystem UUID returned by os-prober
* util/grub.d/30_os-prober.in: Tolerate devices with no filesystem
UUID. Other parts of grub-mkconfig tolerate these, they were
previously allowed here up to commit 55e706c918922def17f5012c23cfe88c4c645208, and they can arise in
practice when the system has active LVM snapshots.
Fixes Ubuntu bug #1287436.
Colin Watson [Mon, 31 Mar 2014 13:48:33 +0000 (14:48 +0100)]
Fix partmap, cryptodisk, and abstraction handling in grub-mkconfig.
Commit 588744d0dc655177d5883bdcb8f72ff5160109ed caused grub-mkconfig
no longer to be forgiving of trailing spaces on grub-probe output
lines, which among other things means that util/grub.d/10_linux.in
no longer detects LVM. To fix this, make grub-probe's output
delimiting more consistent. As a bonus, this improves the coverage
of the -0 option.
is_qemu is not being set lead to disabling of feature like
GRUB_IEEE1275_FLAG_HAS_CURSORONOFF. This resulted in cursor not being
displayed during the grub-menu edit.
Signed-off-by: Nikunj A Dadhania <nikunj@linux.vnet.ibm.com>
It enables net boot even when there is no bootp/dhcp server.
* grub-core/net/drivers/ieee1275/ofnet.c: Add grub_ieee1275_parse_bootpath and
call it at grub_ieee1275_net_config_real.
* grub-core/kern/ieee1275/init.c: Add bootpath to grub_ieee1275_net_config.
* include/grub/ieee1275/ieee1275.h: Likewise.
* grub-core/term/serial.c (grub_serial_register): Fix invalid free.
Ensure that pointers are inited to NULL and that pointers are not
accessed after free.
Andrey Borzenkov [Sat, 25 Jan 2014 17:49:41 +0000 (21:49 +0400)]
fix Mingw W64-32 cross compile failure due to printf redefinition in libintl.h
In file included from util/misc.c:36:0:
./include/grub/emu/misc.h:56:1: error: 'libintl_printf' is an unrecognized format function type [-Werror=format=]
char * EXPORT_FUNC(xasprintf) (const char *fmt, ...) __attribute__ ((format (printf, 1, 2))) WARN_UNUSED_RESULT;
^
./include/grub/emu/misc.h:58:1: error: 'libintl_printf' is an unrecognized format function type [-Werror=format=]
The reason is libintl.h which redefines printf as libintl_printf. The problem
is not present in native MinGW build which avoids redefinition. Use
(format (__printf__) instead which is valid replacement in GCC.
v2: add grub-core/lib/libgcrypt/src/g10lib.h
v3: modify g10lib.h during import
Andrey Borzenkov [Tue, 21 Jan 2014 16:54:09 +0000 (20:54 +0400)]
add GRUB_WINDOWS_EXTRA_DIST to allow shipping runtime files
Not all toolkits provide static libraries. This patch enables creation of self
contained distribution that does not require pre-existing runtime libraries.
Intended usage is
export GRUB_WINDOWS_EXTRA_DIST="/path/to/liblzma.dll /path/to/libintl.dll"
make
make windowszip
As those libraries and locations are dependent on toolchain in use, trying
to autodetect them is likely impossible. So just provide a simple way to
package everything in one step.
Also remove $(windowsdir) after ZIP was created same as other "make dist"
targets.
In net/net.c there is a while (1) that only exits if there is a stop
condition and more then 10 packages or if there is no package received.
If GRUB is idle and enter in this loop, the only condition to leave is
if it doesn't have incoming packages. In a network with heavy traffic
this never happens.
Colin Watson [Fri, 17 Jan 2014 02:28:46 +0000 (02:28 +0000)]
Ignore EPERM when modifying kern.geom.debugflags
Many tests fail when run as a non-root user on FreeBSD. The failures
all amount to an inability to open files using grub_util_fd_open,
because we cannot set the kern.geom.debugflags sysctl. This sysctl is
indeed important to allow us to do such things as installing GRUB to the
MBR, but if we need to do that and can't then we will get an error
later. Enforcing it here is unnecessary and prevents otherwise
perfectly reasonable operations.
Andrey Borzenkov [Sat, 18 Jan 2014 18:48:04 +0000 (22:48 +0400)]
fix removal of {cpu,machine} links on mingw/msys
At least on Windows 2003 using "ln -s dir1 dir2" in msys shell succeeds,
but results in what looks like hard link. Subsequent "rm -f dir2" (e.g.
during second config.status invocation) fails. Check that we also can
remove link to directory.
Make it more clear in message that we are checking "ln -s".
projects / thirdparty / grub.git / log
