An internal security audit of the flashrom project by
Carl-Daniel Hailfinger found a buffer overflow bug present in all
flashrom versions since the year 2005.
This bug was independently found and reported to flashrom.org by
Cosmin Gorgovan a few days ago.
A buffer on the stack and a buffer on the heap are affected by the
overflow caused by an incorrect fscanf format string.
The buffer overflow can only be triggered if the optional layout feature
is used and if the user manually specifies a specially crafted layout
file on the command line. Command line parsing and flash image handling
do not trigger the buggy code path.
Most usage of flashrom does not involve layout files.
The fix in this commit (changed fscanf format string) can be applied to
layout.c of all past flashrom versions.
Corresponding to flashrom svn r1953.
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Stefan Tauner <stefan.tauner@alumni.tuwien.ac.at>
Even if fwrite() succeeds the data is not necessarily out of the clib's buffers
and writing it eventually could fail. Even if the data is flushed out (explicitly by
fflush() or implicitly by fclose()) the kernel might still hold a buffer.
Previously we have ignored this to a large extent - even in important cases
like writing the flash contents to a file. The results can be truncated
images that would brick the respective machine if written back as is (though
flashrom would not allow that due to a size mismatch). flashrom would not
indicate the problem in any output - so far we only check the return value
of fwrite() that is not conclusive.
This patch checks the return values of all related system calls like fclose()
unless we only read the file and are not really interested in output errors.
In the latter case the return value is casted to void to document this fact.
Additionally, this patch explicitly calls fflush() and fsync() (on regular files only)
to do the best we can to guarantee the read image reaches the disk safely
and at least inform the user if it did not work.
Corresponding to flashrom svn r1902.
Signed-off-by: Stefan Tauner <stefan.tauner@alumni.tuwien.ac.at>
Acked-by: Urja Rannikko <urjaman@gmail.com>
- Introduce a variable in doit() that allows to influence
read-before-write and its consequences.
- Modify build_new_image so that it still works even if the old content
is not read before.
- Add copy_old_content() to ease the pain for future patches.
Corresponding to flashrom svn r1851.
Signed-off-by: Stefan Tauner <stefan.tauner@alumni.tuwien.ac.at>
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Stefan Tauner <stefan.tauner@alumni.tuwien.ac.at>
This fixes a SEGFAULT if a layout entry is included that addresses memory
outside the current chip's address range. flashrom will only abort if the
offending region(s) is/are included else it will just warn.
It will print warnings for regions with negative or zero-length address ranges
and bail out after checking all of them.
Also, abort for non-write operations if a layout file is given because there is
no layout support for other operations yet.
Corresponding to flashrom svn r1751.
Signed-off-by: Stefan Tauner <stefan.tauner@alumni.tuwien.ac.at>
Acked-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
The type describes one entry of the whole layout actually.
Using layout_entry_t or something similar would be more correct,
but due to it length we will use "rom" instead of "layout" here and
in upcoming code.
Corresponding to flashrom svn r1732.
Signed-off-by: Stefan Tauner <stefan.tauner@alumni.tuwien.ac.at>
Acked-by: Stefan Tauner <stefan.tauner@alumni.tuwien.ac.at>
Since we are planning to support image files for rom entries, rename the
variable used to count the number of known rom entries to avoid confusion.
There is already num_include_args with similar semantics, hence we use
num_rom_entries.
Corresponding to flashrom svn r1731.
Signed-off-by: Stefan Tauner <stefan.tauner@alumni.tuwien.ac.at>
Acked-by: Stefan Tauner <stefan.tauner@alumni.tuwien.ac.at>
The only caller is able to check the return code and handle it
correctly.
Corresponding to flashrom svn r1545.
Signed-off-by: Niklas Söderlund <niso@kth.se>
Acked-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
NOTE:
The --list-supported-wiki output changed to use -p internal:mainboard=
instead of -m
The --list-supported output changed the heading of the mainboard list
from
Vendor Board Status Required option
to
Vendor Board Status Required value for
-p internal:mainboard=
Fix lb_vendor_dev_from_string() not to write to the supplied string.
Corresponding to flashrom svn r1483.
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Stefan Tauner <stefan.tauner@alumni.tuwien.ac.at>
The general idea and most of the code are based on the following
commits in the chromiumos flashrom tree:
Corresponding to flashrom svn r1482.
8fc0740356ca15d02fb1c65ab43b10844f148c3b
bb9049c66ca55e0dc621dd2c70b5d2cb6e5179bf
Signed-off-by: Louis Yung-Chieh Lo <yjlou@chromium.org>
and the main part:
d0ea9ed71e7f86bb8e8db2ca7c32a96de25343d8
Signed-off-by: David Hendricks <dhendrix@chromium.org>
This implementation does not defer the processing until doit(), but after the
argument parsing loop only (doit() should not contain argument checks).
This allows to specify -i and -l parameters in any order.
Signed-off-by: Stefan Tauner <stefan.tauner@alumni.tuwien.ac.at>
Acked-by: David Hendricks <dhendrix@google.com>
- rename from find_next_included_romentry to get_next_included_romentry
- return a pointer to a rom_entry instead of just its index.
this relieves the (single existing) caller from directly accessing the
data structure holding the entries hence improving segregation and
readability.
Corresponding to flashrom svn r1481.
Signed-off-by: Stefan Tauner <stefan.tauner@alumni.tuwien.ac.at>
Acked-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Struct flashchip is used only for the flashchips array and for
operations which do not access hardware, e.g. printing a list of
supported flash chips.
struct flashctx (flash context) contains all data available in
struct flashchip, but it also contains runtime information like
mapping addresses. struct flashctx is expected to grow additional
members over time, a prime candidate being programmer info.
struct flashctx contains all of struct flashchip with identical
member layout, but struct flashctx has additional members at the end.
The separation between struct flashchip/flashctx shrinks the memory
requirement of the big flashchips array and allows future extension
of flashctx without having to worry about bloat.
Corresponding to flashrom svn r1473.
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Michael Karcher <flashrom@mkarcher.dialup.fu-berlin.de>
If a layout file was specified, all regions not mentioned in the layout
file were taken from the new image instead of being preserved.
If regions overlap, the non-included regions won.
New behaviour:
If a layout file is specified, only the regions explicitly requested for
inclusion will be taken from the new image.
If regions overlap, the included regions win.
Corresponding to flashrom svn r1223.
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: David Hendricks <dhendrix@google.com>
Focus is on partial write and layout functionality.
Minor modifications by Carl-Daniel Hailfinger.
Corresponding to flashrom svn r1218.
Signed-off-by: David Hendricks <dhendrix@google.com>
Acked-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
This will allow flashrom to skip erase of already-erased blocks and to
skip write of blocks which already have the wanted contents.
Avoid emergency messages by checking if the chip contents after a failed
write operation (erase/write) are unchanged.
Keep the emergency messages after a failed pure erase. That part is
debatable because if someone wants erase, he pretty sure doesn't care
about the flash contents anymore.
Please note that this introduces additional overhead of a full chip read
before write. This is frowned upon by people with slow programmers. A
followup patch will make this configurable.
Corresponding to flashrom svn r1215.
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Stefan Reinauer <stepan@coreboot.org>
This doesn't include changes to the frontend which must be
done separately, so this won't work out of the box.
This code was tested on hardware.
Corresponding to flashrom svn r1184.
Signed-off-by: Patrick Georgi <patrick.georgi@coresystems.de>
Acked-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Programmer specific functions are of absolutely no interest to any file
except those dealing with programmer specific actions (special SPI
commands and the generic core).
The new header structure is as follows (and yes, improvements are
possible):
flashchips.h flash chip IDs
chipdrivers.h chip-specific read/write/... functions
flash.h common header for all stuff that doesn't fit elsewhere
hwaccess.h hardware access functions
programmer.h programmer specific functions
coreboot_tables.h header from coreboot, internal programmer only
spi.h SPI command definitions
Corresponding to flashrom svn r1112.
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Uwe Hermann <uwe@hermann-uwe.de>
Carl-Daniel Hailfinger
Stefan Tauner
Mark Marshall
Stefan Reinauer
Niklas Söderlund
Louis Yung-Chieh Lo
David Hendricks
Patrick Georgi