Good day,
Is someone firm with the details around git send-mail? I explicitly
disabled threading (i.e. "git config sendemail.thread" yields "false"),
and the original patch file does not contain any "In-Reply-To" header.
Yet I can see it being added referring to the cover letter of a
differen
For whom it may interest, this was caught with the help of this patch:
https://edk2.groups.io/g/devel/message/79318
Best regards,
Marvin
On 21/08/2021 21:55, Marvin Häuser wrote:
The CLANG38 toolchain aligns 64-bit data types on an 4-Byte boundary
for IA32, which is the default for the GNU ABI
Update the control flow to take the same actions for failed
fixed-address loading as if the feature was disabled. This allows
Images to still be loaded to their preferred address in the case of
a mismatch between fixed-address and preferred address, and also
ensures correct handling of stripped rel
GDB symbols are currently loaded by specifying the .text section
address. It is assumed to be the value of the PE/COFF SizeOfHeaders
field. This may not be the case for various reasons, including a
sufficiently strict Image section alignment. Use the "-o" parameter
to specify the Image base address
The SMM stack does not support loading Images to preferred addresses
in any way. Add checks that Image relocations have not been stripped.
Cc: Jian J Wang
Cc: Hao A Wu
Cc: Eric Dong
Cc: Ray Ni
Cc: Vitaly Cheptsov
Signed-off-by: Marvin Häuser
---
MdeModulePkg/Core/PiSmmCore/Dispatcher.c | 12
The SMM stack does not support loading Images to preferred addresses
in any way. Add checks that Image relocations have not been stripped.
Cc: Jian J Wang
Cc: Hao A Wu
Cc: Eric Dong
Cc: Ray Ni
Cc: Vitaly Cheptsov
Signed-off-by: Marvin Häuser
---
MdeModulePkg/Core/PiSmmCore/PiSmmIpl.c | 9 ++
The initial import of the RVCT debug code used 0 for the address of
the first Image section [1]. This commit has been made in 2010, when
"--ro-base 0" was still passed as a build argument [2]. In 2015, this
was changed to mirror the PE/COFF memory layout the same way that ELF
does [3]. Update the c
Especially in the relative mode, fixed-address loading may not target
the preferred Image base address. In this case, Image relocations are
required to load the Image. Add the necessary check for this.
Cc: Jian J Wang
Cc: Hao A Wu
Cc: Dandan Bi
Cc: Liming Gao
Cc: Vitaly Cheptsov
Signed-off-by
Update the control flow to take the same actions for failed
fixed-address loading as if the feature was disabled. This
primarily removes code duplication.
Cc: Jian J Wang
Cc: Hao A Wu
Cc: Dandan Bi
Cc: Liming Gao
Cc: Debkumar De
Cc: Harry Han
Cc: Catharine West
Cc: Vitaly Cheptsov
Signed-o
PiSmmCoreMemoryAllocationLib duplicates private definitions of
PiSmmCore, namely the SMM_CORE_PRIVATE_DATA structure. Move this code
into PiSmmCore, so that the struct definition can be consumed
directly instead.
Cc: Jian J Wang
Cc: Hao A Wu
Cc: Eric Dong
Cc: Ray Ni
Cc: Vitaly Cheptsov
Signed
GDB symbols are currently loaded by specifying the .text section
address. It is assumed to be the value of the PE/COFF SizeOfHeaders
field. This may not be the case for various reasons, including a
sufficiently strict Image section alignment. Use the "-o" parameter
to specify the Image base address
The CLANG38 toolchain aligns 64-bit data types on an 4-Byte boundary
for IA32, which is the default for the GNU ABI. For the Microsoft ABI
used by UEFI however, natural alignment holds for all data types on
all architectures. Add the "-malign-double" flag already used by the
GCC toolchains to ensur
Use the correct X64 path for CLANGPDB RCPATH.
Cc: Bob Feng
Cc: Liming Gao
Cc: Yuwei Chen
Cc: Vitaly Cheptsov
Signed-off-by: Marvin Häuser
---
BaseTools/Conf/tools_def.template | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/BaseTools/Conf/tools_def.template
b/BaseTools/C
Both GDB and RVCT debugging libraries currently use .text address
based symbols loading for different reasons. As the current code is
making assumptions that are not guaranteed, update both usages to
use the Image base address instead.
Marvin Häuser (3):
EmulatorPkg: Use Image base address for G
This should close up some possible exploits using crafted
filesystem images.
Cc: Leif Lindholm
Cc: Michael D Kinney
Cc: Bret Barkelew
Signed-off-by: Pedro Falcato
---
Features/Ext4Pkg/Ext4Dxe/Directory.c | 90
1 file changed, 51 insertions(+), 39 deletions(-)
di
This commit adds support for EFI_FILE_SYSTEM_VOLUME_LABEL requests
in GetInfo().
Cc: Leif Lindholm
Cc: Michael D Kinney
Cc: Bret Barkelew
Signed-off-by: Pedro Falcato
---
Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.c | 1 -
Features/Ext4Pkg/Ext4Dxe/Ext4Dxe.h | 17
Features/Ext4Pkg/Ext4Dxe/File.
This helps us track directories and directory entries,
which helps us getting a general idea of how the filesystem
looks. In the future, it might serve as a directory cache.
Right now, it only lets us know which name "." and ".." may refer
to, which fixes a EFI_FILE_PROTOCOL::GetInfo() bug that got
This makes it so callers that may expect FAT32 filesystems (most do)
have more normal looking ReadDir() results.
This commit also presents a better filename for files opened through
Open(".").
Cc: Leif Lindholm
Cc: Michael D Kinney
Cc: Bret Barkelew
Signed-off-by: Pedro Falcato
---
Features/
Fixes bug triggered by ShellPkg code, in usage of EFI_FILE_PROTOCOL's
Open().
Cc: Leif Lindholm
Cc: Michael D Kinney
Cc: Bret Barkelew
Signed-off-by: Pedro Falcato
---
Features/Ext4Pkg/Ext4Dxe/File.c | 29 ++---
1 file changed, 22 insertions(+), 7 deletions(-)
diff -
This patch-series addresses bugs found when testing the filesystem with
more complex usage of the file protocol, particularly through the shell
itself.
This is version 2 of the patch series and addresses feedback received
from the community. This version also adds two new patches to further
improv
20 matches
Mail list logo
