Pali Rohár [Mon, 23 May 2022 20:50:36 +0000 (22:50 +0200)]
ubifs: Add missing dependency on GZIP
GZIP option can be manually de-selected when UBIFS is enabled. This cause
following compile error because ubifs calls gzip functions.
/tmp/ccxVrh2c.ltrans1.ltrans.o: in function `gzip_decompress.lto_priv.566':
<artificial>:(.text+0x768): undefined reference to `zunzip'
collect2: error: ld returned 1 exit status
make: *** [Makefile:1813: u-boot] Error 1
Rasmus Villemoes [Fri, 20 May 2022 11:19:08 +0000 (13:19 +0200)]
common/board_r.c: drop legacy and unused bi_enetaddr
The bi_enetaddr field in struct bd_info is write-only; nothing ever
reads back the value.
Moreover, the value we write is more or less random, and certainly not
something one can rely on: If the board has a writable environment and
the mac address has been stored there, we fetch that value. But if the
board doesn't, this code runs before initr_net() -> eth_initialize(),
and thus before the code in eth-uclass which fetches MAC addresses
from eeprom, fuses or whatnot and populates the (run-time) environment
with those values.
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk> Reviewed-by: Tom Rini <trini@konsulko.com>
Sean Anderson [Thu, 19 May 2022 22:26:05 +0000 (18:26 -0400)]
bootm: Fix Linux silent console on newer kernels
Linux determines its console based on several sources:
1. the console command line parameter
2. device tree (e.g. /chosen/stdout-path)
3. various other board- and arch-specific sources
If the console parameter specifies a real console (e.g. ttyS0) then that is
used as /dev/console. However, if it does not specify a real console (e.g.
ttyDoesntExist) then *nothing* will be used as /dev/console.
Reading/writing it will return ENODEV. Additionally, no other source will
be used as a console source.
Linux commit ab4af56ae250 ("printk/console: Allow to disable console output
by using console="" or console=null") recently changed the semantics of the
parameter. Previously, specifying console="" would be treated like
specifying some other bad console. This commit changed things so that it
added /dev/ttynull as a console (if available). However, it also allows
for other console sources. If the device tree specifies a console (such as
if U-Boot and Linux share a device tree), then it will be used in addition
to /dev/ttynull. This can result in a non-silent console.
To avoid this, explicitly set ttynull as the console. This will disable
other console sources. If CONFIG_NULL_TTY is disabled, then this will have
the same behavior as in the past (no output, and writing /dev/console
returns ENODEV).
[1] and [2] have additional background on this kernel change.
Add a check for calloc() failing to allocate the requested memory.
Make decode_regions() return an error code.
Cc: Bin Meng <bmeng.cn@gmail.com> Cc: Simon Glass <sjg@chromium.org> Cc: Stefan Roese <sr@denx.de> Signed-off-by: Pierre-Clément Tosi <ptosi@google.com> Reviewed-by: Stefan Roese <sr@denx.de>
Rasmus Villemoes [Thu, 19 May 2022 09:10:43 +0000 (11:10 +0200)]
fdtdec: drop needlessly convoluted CONFIG_PHANDLE_CHECK_SEQ
Asking if the alias we found actually points at the device tree node
we passed in (in the guise of its offset from blob) can be done simply
by asking if the fdt_path_offset() of the alias' path is identical to
offset.
In fact, the current method suffers from the possibility of false
negatives: dtc does not necessarily emit a phandle property for a node
just because it is referenced in /aliases; it only emits a phandle
property for a node if it is referenced in <angle brackets>
somewhere. So if both the node we passed in and the alias node we're
considering don't have phandles, fdt_get_phandle() returns 0 for both.
Since the proper check is so simple, there's no reason to hide that
behind a config option (and if one really wanted that, it should be
called something else because there's no need to involve phandle in
the check).
Judy Wang [Tue, 3 May 2022 06:04:40 +0000 (14:04 +0800)]
drivers:optee:rpmb: initialize drivers of mmc devices in UCLASS_BLK for rpmb access
CONFIG_MMC only initializes drivers for devices in UCLASS_MMC, we need
to initialize drivers for devices of type IF_TYPE_MMC in UCLASS_BLK as
well because they are the child devices of devices in UCLASS_MMC. This
is required for feature RPMB since it will access eMMC in optee-os.
Signed-off-by: Judy Wang <wangjudy@microsoft.com>
[trini: Add my SoB line and adjust Judy's name in git, having emailed
off-list] Signed-off-by: Tom Rini <trini@konsulko.com>
Sean Anderson [Tue, 17 May 2022 17:55:07 +0000 (13:55 -0400)]
serial: smh: Fake tstc
ARM semihosting provides no provisions for determining if there is
pending input. The only way to determine if there is console input is to
do a read (and block until the user types something). For this reason,
we always return true for tstc (since you will always get input if you
try). However, this behavior can cause problems for code which expects
tstc to eventually be empty. In query_console_serial, there is the
following construct:
/* empty input buffer */
while (tstc())
getchar();
with the current implementation, this effectively turns into an infinite
loop. To avoid this, fake tstc by returning false half of the time. This
is generally OK because the other common construct looks like
do {
if (tstc())
process(getchar());
} while (!timeout());
so it's fine if we only read a new character every other loop. This will
break things like CYGACC_COMM_IF_GETC_TIMEOUT, but that could be
reworked to test on the timeout instead of calling tstc again (and
ymodem over semihosted serial is not that useful in the first place).
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Sean Anderson [Mon, 16 May 2022 20:11:08 +0000 (16:11 -0400)]
mkimage: Support signing 'auto' FITs
This adds support for signing images in auto-generated FITs. To do this,
we need to add a signature node. The algorithm name property already has
its own option, but we need one for the key name hint. We could have
gone the -G route and added an explicit name for the public key (like
what is done for the private key). However, many places assume the
public key can be constructed from the key dir and hint, and I don't
want to do the refactoring necessary.
As a consequence of this, it is now easier to add public keys to an
existing image without signing something. This could be done all along,
but now you don't have to create an its just to do it. Ideally, we
wouldn't create a FIT at the end. This could be done by calling
fit_image_setup_sig/info.crypto->add_verify_data directly.
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Ovidiu Panait [Sun, 15 May 2022 18:40:29 +0000 (21:40 +0300)]
event: fix static events for CONFIG_NEEDS_MANUAL_RELOC
Static events do not currently work post-relocation for boards that enable
CONFIG_NEEDS_MANUAL_RELOC. Relocate event handler pointers for all event
spies to fix this.
Pali Rohár [Fri, 13 May 2022 20:24:51 +0000 (22:24 +0200)]
mtd: mtdpart: Change size type from fdt_addr_t to fdt_size_t
Set correct type for 3rd argument of ofnode_get_addr_size_index_notrans()
function. It expects fdt_size_t * and not fdt_addr_t *.
When these two types do not have same size then U-Boot throw compile
warning:
drivers/mtd/mtdpart.c: In function ‘add_mtd_partitions_of’:
drivers/mtd/mtdpart.c:906:57: warning: passing argument 3 of ‘ofnode_get_addr_size_index_notrans’ from incompatible pointer type [-Wincompatible-pointer-types]
offset = ofnode_get_addr_size_index_notrans(child, 0, &size);
^~~~~
In file included from include/dm/device.h:13,
from include/linux/mtd/mtd.h:26,
from include/ubi_uboot.h:28,
from drivers/mtd/mtdpart.c:27:
include/dm/ofnode.h:530:25: note: expected ‘fdt_size_t *’ {aka ‘long long unsigned int *’} but argument is of type ‘fdt_addr_t *’ {aka ‘long unsigned int *’}
fdt_size_t *size);
~~~~~~~~~~~~^~~~
Signed-off-by: Pali Rohár <pali@kernel.org> Reviewed-by: Marek Behún <marek.behun@nic.cz>
Tom Rini [Tue, 10 May 2022 18:36:59 +0000 (14:36 -0400)]
zlib: Port fix for CVE-2018-25032 to U-Boot
While our copy of zlib is missing upstream commit 263b1a05b04e ("Allow
deflatePrime() to insert bits in the middle of a stream.") we do have
Z_FIXED support, and so the majority of the code changes in 5c44459c3b28
("Fix a bug that can crash deflate on some input when using Z_FIXED.")
apply here directly and cleanly. As this has been assigned a CVE, lets
go and apply these changes.
Ovidiu Panait [Sun, 8 May 2022 10:01:42 +0000 (13:01 +0300)]
cmd: dm: migrate dm command to use U_BOOT_CMD_WITH_SUBCMDS()
Migrate dm command to use U_BOOT_CMD_WITH_SUBCMDS() helper macro, to reduce
duplicated code. We can also drop the CONFIG_NEEDS_MANUAL_RELOC exception,
as the command list is updated post relocation in board_r.c initcall
initr_manual_reloc_cmdtable().
Tom Rini [Mon, 6 Jun 2022 16:09:41 +0000 (12:09 -0400)]
Merge branch '2022-06-06-finish-SPL-Kconfig-migration' into next
- Bring in a number of series of patches that migrate all remaining
CONFIG_SPL symbols to Kconfig, remove some dead code that this
uncovered and then start to tighten the dependencies in Kconfig now
that everything is migrated and these relationships can be clearly
expressed.
Tom Rini [Tue, 31 May 2022 14:24:55 +0000 (10:24 -0400)]
spl: Rework and tighten some dependencies
- In a few places, add missing "depends on" that can be implied from the
option name (i.e. SPL_DM_xxx depends on SPL_DM).
- Make less use of "if SPL_xxx ... endif" clauses as most of the time
this reads better as depends on. In the case of UBI however, move it
all to a sub-menu.
- Rework SPL_NO_CPU_SUPPORT as it's very specific to the
non-SPL_FRAMEWORK implementation used on those platforms, and a
tangent to how CONFIG_SPL_START_S_PATH was used.
Tom Rini [Mon, 30 May 2022 21:01:22 +0000 (17:01 -0400)]
spl: Move all VPL, TPL and PowerPC specific CONFIG options to separate files
- Move all PowerPC (and some shared with Layerscape) options to
common/spl/Kconfig.nxp
- Move all other TPL related options to common/spl/Kconfig.tpl
- Move all VPL related options to common/spl/Kconfig.vpl
This makes the whole of common/spl/Kconfig slightly more readable.
Chris Packham [Sat, 28 May 2022 23:13:18 +0000 (11:13 +1200)]
arm: mvebu: Remove CONFIG_SPL_BOOT_DEVICE
CONFIG_SPL_BOOT_DEVICE was made obsolete by
CONFIG_MVEBU_SPL_BOOT_DEVICE_{SPI,MMC,SATA,UART}.
CONFIG_MVEBU_SPL_BOOT_DEVICE_SPI is the default so existing users of
CONFIG_SPL_BOOT_DEVICE can simply have the option removed.
Signed-off-by: Chris Packham <judge.packham@gmail.com> Reviewed-by: Marek Behún <marek.behun@nic.cz> Reviewed-by: Stefan Roese <sr@denx.de>
Chris Packham [Sat, 28 May 2022 23:13:17 +0000 (11:13 +1200)]
Convert CONFIG_FIXED_SDHCI_ALIGNED_BUFFER to Kconfig
CONFIG_FIXED_SDHCI_ALIGNED_BUFFER is needed on some Marvell SoCs when
booting from MMC. All existing usages of this have the same value so
make this the default and have the Kconfig option depend on SPL &&
MVEBU_SPL_BOOT_DEVICE_MMC.
Signed-off-by: Chris Packham <judge.packham@gmail.com> Reviewed-by: Marek Behún <marek.behun@nic.cz> Reviewed-by: Stefan Roese <sr@denx.de>
Looking at the git history and values used for the raw kernel/args
location, it's clear these platforms only ever did Falcon Mode via
filesystem images and not raw MMC/SD locations. Disable
CONFIG_SPL_FALCON_BOOT_MMCSD.
Tom Rini [Sat, 28 May 2022 16:40:40 +0000 (12:40 -0400)]
spl: Remove CONFIG_SPL_START_S_PATH and rework the logic behind it
In some cases, when we don't use CONFIG_SPL_FRAMEWORK nor are we on
PowerPC using their specific SPL/TPL framework, we need to specify the
start.S file to use for these typically very constrained systems. Do
this within the Makefile logic, rather than introducing a string-based
CONFIG option, as this would get slightly complex to do in Kconfig for a
very limited number of users.
Tom Rini [Sat, 28 May 2022 13:13:59 +0000 (09:13 -0400)]
ax25-ae350: Move CONFIG_SYS_FDT_BASE to Kconfig
The address where the device tree will be passed in to U-Boot at is now
moved to the Kconfig file. If this is user configurable, it needs to be
exposed rather than hidden, and should probably be renamed as well.
Reviewed-by: Rick Chen <rick@andestech.com> Signed-off-by: Tom Rini <trini@konsulko.com>
Tom Rini [Thu, 26 May 2022 17:46:32 +0000 (13:46 -0400)]
imx7: Update CONFIG_SPL_STACK defaults in Kconfig
Update the Kconfig entry to have the correct defaults for i.MX7
platforms, and move the existing large comment from imx7_spl.h to
doc/imx/common/imx7.txt so that it's not lost.
Tom Rini [Thu, 26 May 2022 17:36:17 +0000 (13:36 -0400)]
imx6: Update CONFIG_SPL_STACK defaults in Kconfig
Update the Kconfig entry to have the correct defaults for i.MX6
platforms, and move the existing large comment from imx6_spl.h to
doc/imx/common/imx6.txt so that it's not lost.
Tom Rini [Wed, 25 May 2022 16:16:03 +0000 (12:16 -0400)]
Migrate CUSTOM_SYS_INIT_SP_ADDR to Kconfig using system-constants.h
- Make all users of CUSTOM_SYS_INIT_SP_ADDR reference SYS_INIT_SP_ADDR
- Introduce HAS_CUSTOM_SYS_INIT_SP_ADDR to allow for setting the stack
pointer directly, otherwise we use the common calculation.
- On some platforms that were using the standard calculation but did not
set CONFIG_SYS_INIT_RAM_SIZE / CONFIG_SYS_INIT_RAM_ADDR, set them.
- On a small number of platforms that were not subtracting
GENERATED_GBL_DATA_SIZE do so now via the standard calculation.
- CONFIG_SYS_INIT_SP_OFFSET is now widely unused, so remove it from most
board config header files.
Tom Rini [Wed, 25 May 2022 14:16:18 +0000 (10:16 -0400)]
Introduce include/system-constants.h
We have a number of CONFIG symbols today that are of the form:
SYM1 = CONST1 + CONST2
or other static math operations (shifts, etc). The issue is that by
moving these to Kconfig we no longer have the ability to calculate these
values, so they become less flexible and useful. It's also the case
that sometimes a platform will just define SYM1 directly or perform a
slightly different set of calculations. We introduce this header now to
have a place to start to handle these cases.
Tom Rini [Tue, 24 May 2022 18:14:02 +0000 (14:14 -0400)]
powerpc: Switch to using CONFIG_SYS_INIT_SP_OFFSET from CONFIG_SYS_GBL_DATA_OFFSET
In the places where PowerPC references CONFIG_SYS_GBL_DATA_OFFSET it
does so as (CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_GBL_DATA_OFFSET). And
it defines CONFIG_SYS_GBL_DATA_OFFSET in the same manner that other
architectures define CONFIG_SYS_INIT_SP_OFFSET. Other architectures
define CONFIG_SYS_INIT_SP_ADDR as (CONFIG_SYS_INIT_RAM_ADDR +
CONFIG_SYS_INIT_SP_OFFSET) typically. Rename things within PowerPC for
consistency with other architectures.
Tom Rini [Tue, 24 May 2022 17:49:56 +0000 (13:49 -0400)]
mpc85xx: Switch to setting the initial stack pointer more clearly
Currently, since we know that in the combination of
CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_GBL_DATA_OFFSET all of the "high"
bits are in CONFIG_SYS_INIT_RAM_ADDR and "low" bits are in
CONFIG_SYS_GBL_DATA_OFFSET we reference this separately in start.S, but
added together everywhere else. For clarity consistency, reference the
combined value here instead.
Tom Rini [Tue, 24 May 2022 17:11:41 +0000 (13:11 -0400)]
arm: Use CONFIG_SPL_STACK or CONFIG_SYS_INIT_SP_ADDR directly.
In some cases, we define CONFIG_SYS_INIT_SP_ADDR differently for SPL or
full U-Boot. This case should be making use of CONFIG_SPL_STACK, as
that's what that variable is for. In a few other cases we define
CONFIG_SPL_STACK directly to CONFIG_SYS_INIT_SP_ADDR, but do not need to
as the code handles this correctly, normally.
Tom Rini [Tue, 24 May 2022 13:57:18 +0000 (09:57 -0400)]
mvebu: Use CONFIG_SPL_STACK + 4 directly for bootparam location
The definition of CONFIG_SPL_BOOTROM_SAVE is always a fixed
CONFIG_SPL_STACK + 4, while CONFIG_SPL_STACK is not constant. This
change will make it clear where the location is still, once
CONFIG_SPL_STACK moves to Kconfig.
Cc: Stefan Roese <sr@denx.de> Signed-off-by: Tom Rini <trini@konsulko.com>
Tom Rini [Sat, 21 May 2022 15:26:27 +0000 (11:26 -0400)]
ppc / layerscape: Clean up CONFIG_SYS_CCSR_DO_NOT_RELOCATE usage
A number of PowerPC platforms define this, for SPL. To move this to
Kconfig, it needs to be CONFIG_SPL_SYS_CCSR_DO_NOT_RELOCATE, so use
CONFIG_IS_ENABLED() to check for usage. A number of layerscape
platforms bring this logic from PowerPC, but only need a small part of
it, for the fman driver. Remove their unused portion at least.
Tom Rini [Fri, 13 May 2022 17:37:30 +0000 (13:37 -0400)]
spl: Remove CONFIG_SPL_SATA_BOOT_DEVICE
This is only referenced in non-SPL_DM cases, of which there are
currently none. Remove this option and slightly re-organize the code is
there is now never an if/else at the start of spl_sata_load_image()
Tom Rini [Thu, 12 May 2022 21:22:26 +0000 (17:22 -0400)]
arm: omap2plus: Move CONFIG_SYS_PTV out of CONFIG namespace
This is always defined to 2, and referenced in two places. Move the
define to <asm/omap_common.h> and make sure the code that uses this
includes that file. Make <asm/arch-omap*/clock.h> not include that
file, as we don't need to be doing so.
Two defconfigs were missed when transitioning the SYS_FMAN_FW_ADDR
symbol to Kconfig. CONFIG_SYS_FMAN_FW_ADDR is currently initialized to
0 by default on these builds, which prevents the firmware from loading.
Add the correct symbols to these defconfigs.
Fixes: a97a071d10d2b ("configs: fsl: migrate FMAN/QE specific defines to Kconfig") Signed-off-by: Camelia Groza <camelia.groza@nxp.com>
Vincent Stehlé [Tue, 31 May 2022 07:55:34 +0000 (09:55 +0200)]
efi: test/py: authenticate fit capsules
Add support for the authentication of UEFI capsules containing FIT images.
The authentication code is moved out of the function handling raw images
into a new function efi_firmware_capsule_authenticate(). The special case
for the FMP header coming from edk2 tools is preserved. There is no
functional change for capsules containing raw images.
The python test for signed capsules with raw images is renamed with no
functional change and a new test is added for signed capsules containing
FIT images.
This can be tested with sandbox64_defconfig or sandbox_flattree_defconfig,
plus CONFIG_EFI_CAPSULE_AUTHENTICATE=y.
Signed-off-by: Vincent Stehlé <vincent.stehle@arm.com> Cc: Heinrich Schuchardt <xypron.glpk@gmx.de>
Vincent Stehlé [Tue, 31 May 2022 07:55:33 +0000 (09:55 +0200)]
test/py: efi_capsule: repair image authentication test
Repair the python tests for authenticated EFI capsules, which can be run
with sandbox_defconfig plus CONFIG_EFI_CAPSULE_AUTHENTICATE=y.
- Account for the reset changes done by commit 3e6f81000672 ("efi_loader:
test/py: Reset system after capsule update on disk").
- Fix the capsule GUID typo introduced by commit 2e9c3c6965ba ("test:
capsule: Modify the capsule tests to use GUID values for sandbox").
Signed-off-by: Vincent Stehlé <vincent.stehle@arm.com> Cc: Heinrich Schuchardt <xypron.glpk@gmx.de>
Sughosh Ganu [Tue, 31 May 2022 07:15:35 +0000 (12:45 +0530)]
EFI: Update the documentation to reflect the correct value of OsIndications
The OsIndications is a 64 bit variable, and the current code expects
the value of the variable to be 64 bit. Update the documentation to
reflect this fact.
Signed-off-by: Sughosh Ganu <sughosh.ganu@linaro.org> Reviewed-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Sughosh Ganu [Tue, 31 May 2022 07:15:33 +0000 (12:45 +0530)]
EFI: Populate descriptor_count value only when image_info_size is not zero
The GetImageInfo function of the Firmware Mangement Protocol(FMP) gets
called initially to query the size of the image descriptor array that
would have to be allocated. During this call, the rest of the function
arguments, specifically pointers might be passed as NULL. Do not
populate the descriptor_count value before it is known that the call
to GetImageInfo has been made with the allocated buffer for the image
descriptors.
Signed-off-by: Sughosh Ganu <sughosh.ganu@linaro.org> Reviewed-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Chris Packham [Wed, 25 May 2022 01:08:51 +0000 (13:08 +1200)]
doc: environment: Fix typo
"valu" should be "value".
Signed-off-by: Chris Packham <judge.packham@gmail.com> Reviewed-by: Bin Meng <bmeng.cn@gmail.com> Reviewed-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Add a new device tree property "u-boot,version" in the chosen node to
pass the U-Boot version to the operating system.
This can be useful to implement a firmware upgrade procedure from the
operating system.
Signed-off-by: Francesco Dolcini <francesco.dolcini@toradex.com> Reviewed-by: Tom Rini <trini@konsulko.com>
Fabio Estevam [Thu, 26 May 2022 14:14:37 +0000 (11:14 -0300)]
net: Check for the minimum IP fragmented datagram size
Nicolas Bidron and Nicolas Guigo reported the two bugs below:
"
----------BUG 1----------
In compiled versions of U-Boot that define CONFIG_IP_DEFRAG, a value of
`ip->ip_len` (IP packet header's Total Length) higher than `IP_HDR_SIZE`
and strictly lower than `IP_HDR_SIZE+8` will lead to a value for `len`
comprised between `0` and `7`. This will ultimately result in a
truncated division by `8` resulting value of `0` forcing the hole
metadata and fragment to point to the same location. The subsequent
memcopy will overwrite the hole metadata with the fragment data. Through
a second fragment, this can be exploited to write to an arbitrary offset
controlled by that overwritten hole metadata value.
This bug is only exploitable locally as it requires crafting two packets
the first of which would most likely be dropped through routing due to
its unexpectedly low Total Length. However, this bug can potentially be
exploited to root linux based embedded devices locally.
/* payload starts after IP header, this fragment is in there */
payload = (struct hole *)(pkt_buff + IP_HDR_SIZE);
offset8 = (ip_off & IP_OFFS);
thisfrag = payload + offset8;
start = offset8 * 8;
len = ntohs(ip->ip_len) - IP_HDR_SIZE;
```
The last line of the previous excerpt from `u-boot/net/net.c` shows how
the attacker can control the value of `len` to be strictly lower than
`8` by issuing a packet with `ip_len` between `21` and `27`
(`IP_HDR_SIZE` has a value of `20`).
Also note that `offset8` here is `0` which leads to `thisfrag = payload`.
```C
} else if (h >= thisfrag) {
/* overlaps with initial part of the hole: move this hole */
newh = thisfrag + (len / 8);
*newh = *h;
h = newh;
if (h->next_hole)
payload[h->next_hole].prev_hole = (h - payload);
if (h->prev_hole)
payload[h->prev_hole].next_hole = (h - payload);
else
first_hole = (h - payload);
} else {
```
Lower down the same function, execution reaches the above code path.
Here, `len / 8` evaluates to `0` leading to `newh = thisfrag`. Also note
that `first_hole` here is `0` since `h` and `payload` point to the same
location.
```C
/* finally copy this fragment and possibly return whole packet */
memcpy((uchar *)thisfrag, indata + IP_HDR_SIZE, len);
```
Finally, in the above excerpt the `memcpy` overwrites the hole metadata
since `thisfrag` and `h` both point to the same location. The hole
metadata is effectively overwritten with arbitrary data from the
fragmented IP packet data. If `len` was crafted to be `6`, `last_byte`,
`next_hole`, and `prev_hole` of the `first_hole` can be controlled by
the attacker.
Finally the arbitrary offset write occurs through a second fragment that
only needs to be crafted to write data in the hole pointed to by the
previously controlled hole metadata (`next_hole`) from the first packet.
### Recommendation
Handle cases where `len` is strictly lower than 8 by preventing the
overwrite of the hole metadata during the memcpy of the fragment. This
could be achieved by either:
* Moving the location where the hole metadata is stored when `len` is
lower than `8`.
* Or outright rejecting fragmented IP datagram with a Total Length
(`ip_len`) lower than 28 bytes which is the minimum valid fragmented IP
datagram size (as defined as the minimum fragment of 8 octets in the IP
Specification Document:
[RFC791](https://datatracker.ietf.org/doc/html/rfc791) page 25).
----------BUG 2----------
In compiled versions of U-Boot that define CONFIG_IP_DEFRAG, a value of
`ip->ip_len` (IP packet header's Total Length) lower than `IP_HDR_SIZE`
will lead to a negative value for `len` which will ultimately result in
a buffer overflow during the subsequent `memcpy` that uses `len` as it's
`count` parameter.
This bug is only exploitable on local ethernet as it requires crafting
an invalid packet to include an unexpected `ip_len` value in the IP UDP
header that's lower than the minimum accepted Total Length of a packet
(21 as defined in the IP Specification Document:
[RFC791](https://datatracker.ietf.org/doc/html/rfc791)). Such packet
would in all likelihood be dropped while being routed to its final
destination through most routing equipment and as such requires the
attacker to be in a local position in order to be exploited.
/* payload starts after IP header, this fragment is in there */
payload = (struct hole *)(pkt_buff + IP_HDR_SIZE);
offset8 = (ip_off & IP_OFFS);
thisfrag = payload + offset8;
start = offset8 * 8;
len = ntohs(ip->ip_len) - IP_HDR_SIZE;
```
The last line of the previous excerpt from `u-boot/net/net.c` shows
where the underflow to a negative `len` value occurs if `ip_len` is set
to a value strictly lower than 20 (`IP_HDR_SIZE` being 20). Also note
that in the above excerpt the `pkt_buff` buffer has a size of
`CONFIG_NET_MAXDEFRAG` which defaults to 16 KB but can range from 1KB to
64 KB depending on configurations.
```C
/* finally copy this fragment and possibly return whole packet */
memcpy((uchar *)thisfrag, indata + IP_HDR_SIZE, len);
```
In the above excerpt the `memcpy` overflows the destination by
attempting to make a copy of nearly 4 gigabytes in a buffer that's
designed to hold `CONFIG_NET_MAXDEFRAG` bytes at most which leads to a DoS.
### Recommendation
Stop processing of the packet if `ip_len` is lower than 21 (as defined
by the minimum length of a data carrying datagram in the IP
Specification Document:
[RFC791](https://datatracker.ietf.org/doc/html/rfc791) page 34)."
Add a check for ip_len lesser than 28 and stop processing the packet
in this case.
Such a check covers the two reported bugs.
Reported-by: Nicolas Bidron <nicolas.bidron@nccgroup.com> Signed-off-by: Fabio Estevam <festevam@denx.de>
Andre Przywara [Mon, 9 May 2022 16:08:49 +0000 (17:08 +0100)]
armv8: Fix TCR 64-bit writes
The AArch64 TCR_ELx register is a 64-bit register, and many newer
architecture features use bits in the upper half. So far U-Boot was
igorant of those bits, trying to leave them alone.
However, in an effort to set bit 31 to 1, it failed doing so, because
the compiler sign-extended "1 << 31", so that all bits[63:31] got set.
Older ARMv8.0 cores don't define anything dangerous up there, but newer
architecture revisions do, and setting all those bits will end badly:
=================
$ qemu-system-aarch64 -cpu max ....
U-Boot 2022.07-rc1 (May 09 2022 - 15:21:00 +0100)
DRAM: 1.5 GiB
================= (hangs here)
Defining TCR_ELx_RSVD to "1U << 31" avoids the sign-extension, so all
upper bits stay at a safe 0 value. This means no more surprises when
U-Boot runs on a more capable CPU core.
Reported-by: Balaji Anandapadmanaban <Balaji.Anandapadmanaban@arm.com> Signed-off-by: Andre Przywara <andre.przywara@arm.com> Reviewed-by: Peng Fan <peng.fan@nxp.com> Tested-by: Peter Collingbourne <pcc@google.com> Reviewed-by: Peter Collingbourne <pcc@google.com>