September 2024 - U-Boot

[PATCH] cmd: sf: prevent overwriting the reserved memory
by Prasad Kummari 01 Sep '24

01 Sep '24

Added LMB API to prevent SF command from overwriting reserved memory areas. The current SPI code does not use LMB APIs for loading data into memory addresses. To resolve this, LMB APIs were added to check the load address of an SF command and ensure it does not overwrite reserved memory addresses. Similar checks are used in TFTP, serial load, and boot code to prevent overwriting reserved memory. Signed-off-by: Prasad Kummari <prasad.kummari(a)amd.com> --- UT: Tested on Versal NET board Versal NET> sf read 0xf000000 0x0 0x100 device 0 offset 0x0, size 0x100 ERROR: trying to overwrite reserved memory... Versal NET> sf read 0x79ea9000 0x 0x100 device 0 offset 0x0, size 0x100 ERROR: trying to overwrite reserved memory... Versal NET> sf write 0x79ea9000 0x 0x100 device 0 offset 0x0, size 0x100 ERROR: trying to overwrite reserved memory... Versal NET> sf write 0x79eaf000 0x20000 0x100 device 0 offset 0x20000, size 0x100 ERROR: trying to overwrite reserved memory... cmd/sf.c | 38 +++++++++++++++++++++++++++++++++++++- 1 file changed, 37 insertions(+), 1 deletion(-) diff --git a/cmd/sf.c b/cmd/sf.c index f43a2e08b3..d2ce1af8a0 100644 --- a/cmd/sf.c +++ b/cmd/sf.c @@ -10,6 +10,7 @@ #include <div64.h> #include <dm.h> #include <log.h> +#include <lmb.h> #include <malloc.h> #include <mapmem.h> #include <spi.h> @@ -272,6 +273,35 @@ static int spi_flash_update(struct spi_flash *flash, u32 offset, return 0; } +#ifdef CONFIG_LMB +static int do_spi_read_lmb_check(ulong start_addr, loff_t len) +{ + struct lmb lmb; + phys_size_t max_size; + ulong end_addr; + + lmb_init_and_reserve(&lmb, gd->bd, (void *)gd->fdt_blob); + lmb_dump_all(&lmb); + + max_size = lmb_get_free_size(&lmb, start_addr); + if (!max_size) { + printf("ERROR: trying to overwrite reserved memory...\n"); + return CMD_RET_FAILURE; + } + + end_addr = start_addr + max_size; + if (!end_addr) + end_addr = ULONG_MAX; + + if ((start_addr + len) > end_addr) { + printf("ERROR: trying to overwrite reserved memory...\n"); + return CMD_RET_FAILURE; + } + + return 0; +} +#endif + static int do_spi_flash_read_write(int argc, char *const argv[]) { unsigned long addr; @@ -315,7 +345,13 @@ static int do_spi_flash_read_write(int argc, char *const argv[]) ret = spi_flash_update(flash, offset, len, buf); } else if (strncmp(argv[0], "read", 4) == 0 || strncmp(argv[0], "write", 5) == 0) { - int read; + int read, ret; + +#ifdef CONFIG_LMB + ret = do_spi_read_lmb_check(addr, len); + if (ret) + return ret; +#endif read = strncmp(argv[0], "read", 4) == 0; if (read) -- 2.25.1

6 19

[PATCH] efi_loader: fix memory freeing in efi_get_dp_from_boot()
by Ilias Apalodimas 01 Sep '24

01 Sep '24

efi_get_var() allocates memory which must be freed after the variable is used. Since the device path is duplicated after we deserialize the load options free the memory used for the variable payload Fixes: db61e70e0d2a efi_loader: efi_dp_from_lo() should skip VenMedia node Signed-off-by: Ilias Apalodimas <ilias.apalodimas(a)linaro.org> --- lib/efi_loader/efi_helper.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/lib/efi_loader/efi_helper.c b/lib/efi_loader/efi_helper.c index 65d2116381ae..916f43da26c7 100644 --- a/lib/efi_loader/efi_helper.c +++ b/lib/efi_loader/efi_helper.c @@ -74,6 +74,7 @@ out: */ struct efi_device_path *efi_get_dp_from_boot(const efi_guid_t *guid) { + struct efi_device_path *file_path = NULL; struct efi_load_option lo; void *var_value; efi_uintn_t size; @@ -92,11 +93,11 @@ struct efi_device_path *efi_get_dp_from_boot(const efi_guid_t *guid) if (ret != EFI_SUCCESS) goto err; - return efi_dp_from_lo(&lo, guid); + file_path = efi_dp_from_lo(&lo, guid); err: free(var_value); - return NULL; + return file_path; } /** -- 2.45.2

2 1

[PATCH 1/3] efi_loader: shorten efi_bootmgr_release_uridp_resource()
by Ilias Apalodimas 01 Sep '24

01 Sep '24

We use this function to clean up leftover resources when booting an EFI HTTP boot image, but the name is unnecessary long. Shorten it to efi_bootmgr_release_uridp() Signed-off-by: Ilias Apalodimas <ilias.apalodimas(a)linaro.org> --- lib/efi_loader/efi_bootmgr.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/lib/efi_loader/efi_bootmgr.c b/lib/efi_loader/efi_bootmgr.c index 589d3996b680..181fc8775b96 100644 --- a/lib/efi_loader/efi_bootmgr.c +++ b/lib/efi_loader/efi_bootmgr.c @@ -380,12 +380,12 @@ err: } /** - * efi_bootmgr_release_uridp_resource() - cleanup uri device path resource + * efi_bootmgr_release_uridp() - cleanup uri device path resource * * @ctx: event context * Return: status code */ -efi_status_t efi_bootmgr_release_uridp_resource(struct uridp_context *ctx) +efi_status_t efi_bootmgr_release_uridp(struct uridp_context *ctx) { efi_status_t ret = EFI_SUCCESS; @@ -432,7 +432,7 @@ static void EFIAPI efi_bootmgr_image_return_notify(struct efi_event *event, efi_status_t ret; EFI_ENTRY("%p, %p", event, context); - ret = efi_bootmgr_release_uridp_resource(context); + ret = efi_bootmgr_release_uridp(context); EFI_EXIT(ret); } @@ -544,7 +544,7 @@ static efi_status_t try_load_from_uri_path(struct efi_device_path_uri *uridp, return ret; err: - efi_bootmgr_release_uridp_resource(ctx); + efi_bootmgr_release_uridp(ctx); return ret; } -- 2.45.2

2 5

My labgrid setup
by Tom Rini 01 Sep '24

01 Sep '24

Hey all, So now that I've posted my u-boot-test-hooks for labgrid: https://patchwork.ozlabs.org/project/uboot/patch/20240829185620.3179866-1-t… the next relevant parts would be how I use that. There's three scripts, first of which is "uboot-hw-testcycle.sh" : ------------------------ >8 ------------------------ #!/bin/bash set -e # Save output LOG=$HOME/logs/`git describe --abbrev=0`/`date +"%Y-%m-%d-%H%M"`-hw.txt # Make a base directory B=`mktemp -d` chmod a+rwX ${B} cleanup() { # Clean up random outputs [ -d test/py/.pytest_cache ] && find -type d -name __pycache__ | xargs rm -r sudo rm -rf ${B} rm -rf test/py/.pytest_cache } # Cleanups trap cleanup INT trap cleanup ERR trap cleanup EXIT # Find our arguments while test $# -ne 0; do if [ "$1" == "--no-32bit" ]; then NO32BIT="$1" shift elif [ "$1" == "--no-64bit" ]; then NO64BIT="$1" shift else echo "Unknown argument $1" exit 1 fi done # Run the build if [ -f .gitlab-ci.yml ]; then IMG=`grep ci_runner_image: .azure-pipelines.yml | cut -d ' ' -f 4` else IMG=`git grep ci_runner_image: origin/master -- .azure-pipelines.yml | cut -d ' ' -f 4` fi # Build all platforms touch $LOG && chmod a+rw $LOG sudo docker run --rm -it --privileged \ -v /home/trini/bin:/home/uboot/bin \ -v /home/trini/u-boot:/home/uboot/u-boot \ -v /home/trini/logs:/home/uboot/logs \ -v ${B}:${B} \ -w /home/uboot/u-boot/u-boot \ -e PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/home/uboot/u-boot/u-boot-test-hooks/bin \ $IMG \ /home/uboot/bin/uboot-local-ci-in-docker.sh ${B} $NO32BIT $NO64BIT --log $LOG # Test all targets /home/trini/bin/uboot-local-ci-outside-docker.sh $B $NO32BIT $NO64BIT --log $LOG less $LOG ------------------------ >8 ------------------------ Which is still a little debug-y, but is basically just build everything I want in the same container as CI, and then run pytest outside of that container. Could I do it inside the container? Maybe. This has grown from when I used to build everything on the host and used "flashair" SD cards to update my hardware. With labgrid everything might be abstracted enough that yes, really, I can stop doing that. I'll likely look in to that next just to clean things up more. Second is "uboot-local-ci-in-docker.sh" as invoked above: ------------------------ >8 ------------------------ #!/bin/bash # Build everything we are going to test. set -e # Let git stop being mad about permissions. git config --global --add safe.directory /home/uboot/u-boot/u-boot buildaboard() { # Board is first, all fragments follow. BRD=$1 shift 1 # Second would be to use clang if [ "$1" == "--clang" ]; then echo "Building $BRD with clang" | tee -a $LOG OBJ=${B}/${BRD}_clang export HOSTCC=clang-16 export CC=clang-16 shift 1 else echo "Building $BRD" | tee -a $LOG OBJ=${B}/${BRD} fi # All that remains are fragments. FRAGS=$@ make O=${OBJ} ${BRD}_defconfig -s 2>&1 | tee -a $LOG pushd ${OBJ} >/dev/null [ ! -z "${FRAGS}" ] && ${S}/scripts/kconfig/merge_config.sh ${S}/configs/${BRD}_defconfig ${FRAGS} 2>&1 | tee -a $LOG make -sj$(nproc) 2>&1 | tee -a $LOG popd >/dev/null } # Prepare virtualenv if [ -f test/py/requirements.txt ]; then virtualenv -p /usr/bin/python3 /tmp/venv . /tmp/venv/bin/activate pip install --quiet -r test/py/requirements.txt pip install --quiet -r doc/sphinx/requirements.txt [ -f tools/buildman/requirements.txt ] && pip install --quiet -r tools/buildman/requirements.txt fi # Find our arguments while test $# -ne 0; do if [ "$1" == "--no-32bit" ]; then NO32BIT="true" shift elif [ "$1" == "--no-64bit" ]; then NO64BIT="true" shift elif [ "$1" == "--log" ]; then LOG=${2/trini/uboot} shift 2 else B=$1 shift fi done if [ -z "$LOG" ]; then echo "Must pass --log /path/to/log/file" exit 1 fi S=`pwd` # Common to all newer TI platforms export BINMAN_INDIRS=/home/uboot/u-boot/ti-linux-firmware # Prepare defconfigs with addons echo 'CONFIG_UNIT_TEST=y' > ${B}/unittest.config echo 'CONFIG_CMD_BOOTMENU=y' > ${B}/general.config echo 'CONFIG_CMD_LOG=y' >> ${B}/general.config echo '# CONFIG_CMD_BOOTEFI_SELFTEST is not set' > ${B}/noefist.config echo 'CONFIG_CMD_BOOTEFI_HELLO=y' > ${B}/bootefi.config echo 'CONFIG_CMD_BOOTEFI_SELFTEST=y' >> ${B}/bootefi.config echo 'CONFIG_IPV6=y' > ${B}/net.config echo 'CONFIG_IPV6_ROUTER_DISCOVERY=y' >> ${B}/net.config echo 'CONFIG_CMD_TFTPPUT=y' >> ${B}/net.config echo 'CONFIG_FIT=y' >> ${B}/net.config echo 'CONFIG_FIT_SIGNATURE=y' >> ${B}/net.config echo '# CONFIG_AUTOBOOT_KEYED is not set' > ${B}/noautobootkeyed.config echo 'CONFIG_BOOTSTAGE_STASH_ADDR=0x02400000' > ${B}/bootstage-pi.config echo 'CONFIG_BOOTSTAGE=y' >> ${B}/bootstage-pi.config echo 'CONFIG_BOOTSTAGE_STASH=y' >> ${B}/bootstage-pi.config echo 'CONFIG_CMD_BOOTSTAGE=y' >> ${B}/bootstage-pi.config ## ARM if [ -d /opt/gcc-13.2.0-nolibc/arm-linux-gnueabi ]; then export CROSS_COMPILE=/opt/gcc-13.2.0-nolibc/arm-linux-gnueabi/bin/arm-linux-gnueabi- else echo "Toolchain missing for ARM" exit 1 fi if [ -z "$NO32BIT" ]; then # Raspberry Pi boards buildaboard rpi_4_32b ${B}/bootefi.config ${B}/net.config ${B}/general.config ${B}/bootstage-pi.config buildaboard rpi_4_32b --clang ${B}/bootefi.config ${B}/net.config ${B}/general.config ${B}/bootstage-pi.config fi if [ -z "$NO64BIT" ]; then ## Build r5 cores for the TI platforms that need both ## Need to backport the kernel fix for cpp with clang and passing the right ## flags. Then need to figure out pip / venv. buildaboard am64x_evm_r5 buildaboard am62x_evm_r5 buildaboard am62x_beagleplay_r5 fi ## AARCH64 if [ -d /opt/gcc-13.2.0-nolibc/aarch64-linux ]; then export CROSS_COMPILE=/opt/gcc-13.2.0-nolibc/aarch64-linux/bin/aarch64-linux- else echo "Toolchain missing for ARM64" exit 1 fi if [ -z "$NO64BIT" ]; then # Raspberry Pi platforms buildaboard rpi_4 ${B}/bootefi.config ${B}/net.config ${B}/general.config ${B}/bootstage-pi.config buildaboard rpi_4 --clang ${B}/bootefi.config ${B}/net.config ${B}/general.config ${B}/bootstage-pi.config buildaboard rpi_arm64 ${B}/bootefi.config ${B}/net.config ${B}/general.config ${B}/bootstage-pi.config buildaboard rpi_arm64 --clang ${B}/bootefi.config ${B}/net.config ${B}/general.config ${B}/bootstage-pi.config ## These TI platforms require an R5 core and then Axx core ## Need to backport the kernel fix for cpp with clang and passing the right ## flags. export BL31=/home/uboot/u-boot/external-binaries/am64x-u-boot-binaries/bl31.bin export TEE=/home/uboot/u-boot/external-binaries/am64x-u-boot-binaries/tee-raw.bin buildaboard am64x_evm_a53 ${B}/noefist.config ${B}/net.config ${B}/general.config # The AM62x platforms share BL31/TEE. export BL31=/home/uboot/u-boot/external-binaries/am62x-u-boot-binaries/bl31.bin export TEE=/home/uboot/u-boot/external-binaries/am62x-u-boot-binaries/tee-raw.bin buildaboard am62x_evm_a53 ${B}/noefist.config ${B}/net.config ${B}/general.config buildaboard am62x_beagleplay_a53 ${B}/noautobootkeyed.config ${B}/noefist.config ${B}/net.config fi echo "Building complete on `git rev-parse --short HEAD`" ------------------------ >8 ------------------------ Which is a little debug-y but more so comments to self. I setup the additional changes to a platform that I want to test with, and then build them. For required additional binaries I point them at known good copies so I don't have to worry if a failure is because of U-Boot or some other project. Finally we have "uboot-local-ci-outside-docker.sh" which is where pytest is actually run. ------------------------ >8 ------------------------ #!/bin/bash set -e # Find our arguments while test $# -ne 0; do if [ "$1" == "--no-32bit" ]; then NO32BIT="true" shift elif [ "$1" == "--no-64bit" ]; then NO64BIT="true" shift elif [ "$1" == "--log" ]; then LOG=$2 shift 2 else B=$1 shift fi done if [ -z "$LOG" ]; then echo "Must pass --log /path/to/log/file" exit 1 fi # Env export PATH=${PATH}:/home/trini/u-boot/u-boot-test-hooks/bin:/sbin export PYTHONPATH=/home/trini/u-boot/u-boot-test-hooks/py/lootbox:${PYTHONPATH} # Setup Labgrid variables and use the special virtualenv for now export LG_ENV=/home/trini/u-boot/lg_env.yaml export LG_CROSSBAR=ws://lootbox.k.g:20408/ws export PATH=$PATH:/home/trini/u-boot-test-hooks/bin source /home/trini/labgrid-venv/bin/activate runtestpy() { BT=$1 shift 1 if [ "$1" == "--clang" ]; then echo "Testing $BT built with clang" shift 1 BD=${B}/${BT} RD=${B}/${BT}_clang_results PD=${B}/${BT}_clang_persistent_data else echo "Testing $BT" BD=${B}/${BT} RD=${B}/${BT}_results PD=${B}/${BT}_persistent_data fi # If we need special pytest arguments do that here. while getopts "sk:" options; do case $options in k) PA="-k $2" shift 2 ;; esac done labgrid-client -c $LG_ENV -p $LG_PLACE release >/dev/null 2>&1 || true labgrid-client -c $LG_ENV -p $LG_PLACE acquire ./test/py/test.py -ra --bd $BT --build-dir $BD --result-dir $RD \ --persistent-data-dir $PD --exitfirst $PA >> $LOG 2>&1 labgrid-client -c $LG_ENV -p $LG_PLACE release } if [ -z "$NO64BIT" ]; then export LG_PLACE=rpi4 runtestpy rpi_4 runtestpy rpi_arm64 runtestpy rpi_4 --clang runtestpy rpi_arm64 --clang export LG_PLACE=am64-sk runtestpy am64x_evm_a53 export LG_PLACE=am62-sk runtestpy am62x_evm_a53 export LG_PLACE=beagleplay runtestpy am62x_beagleplay_a53 fi if [ -z "$NO32BIT" ]; then export LG_PLACE=rpi4 runtestpy rpi_4_32b runtestpy rpi_4_32b --clang fi echo "Tests complete on `git rev-parse --short HEAD`" ------------------------ >8 ------------------------ I will say, if there's one thing I don't quite like with how my integrations with labgrid work right now it's that acquire/release is done the way it is. This is where I think Simon's adding another hook is the right direction, if there's not some other hook that can be used and I'm just missing. -- Tom

2 3

[PATCH v2 0/3] efi: Start tidying up memory management
by Simon Glass 01 Sep '24

01 Sep '24

We have been discussing the state of EFI memory management for some years so I thought it might be best to send a short series showing some of the issues we have talked about. This one just deals with memory allocation. It updates EFI to use U-Boot memory allocation for the pool where possible. Most functions use that anyway and it is much more efficient. It also avoids allocating things 'in space' in conflict with the loaded images. For v2 I have dropped patches which are not germane to the main idea. Note that this series is independent from Sugosh's lmb series[1], although I believe it will point the way to simplifying some of the later patches in that series. Overall, some issues which should be resolved in future are: - allocation inconsistency, e.g. efi_add_protocol() uses malloc() but efi_dp_dup() does not (this series makes a start) - change efi_bl_init() to register events later, when the devices are actually used - rather than efi_set_bootdev(), provide a bootstd way to take note of the device images came from and allow EFI to query that when needed - EFI_LOADER_BOUNCE_BUFFER - can use U-Boot bounce buffers? Minor questions on my mind: - is unaligned access useful? Is there a performance penalty? - API confusion about whether something is an address or a pointer [1] https://patchwork.ozlabs.org/project/uboot/list/?series=416441 Changes in v2: - Drop patch 'Show more information in efi index' - Drop patch 'Avoid pool allocation in efi_get_memory_map_alloc()' - Add the word 'warning', use log_warning() and show the end address - Reorder patches a little Simon Glass (3): efi: Allow use of malloc() for the EFI pool efi: Convert device_path allocation to use malloc() efi: Show the location of the bounce buffer common/dlmalloc.c | 7 ++ include/efi_loader.h | 18 ++++ include/malloc.h | 7 ++ lib/efi_loader/efi_bootbin.c | 11 +++ lib/efi_loader/efi_device_path.c | 43 +++++---- lib/efi_loader/efi_device_path_to_text.c | 2 +- lib/efi_loader/efi_memory.c | 110 +++++++++++++++++------ 7 files changed, 148 insertions(+), 50 deletions(-) -- 2.34.1

3 21

[PATCH 1/3] doc: board: stm32mp1: add missing newline at EOF
by Grzegorz Szymaszek 01 Sep '24

01 Sep '24

Signed-off-by: Grzegorz Szymaszek <gszymaszek(a)short.pl> --- doc/board/st/stm32mp1.rst | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/doc/board/st/stm32mp1.rst b/doc/board/st/stm32mp1.rst index 63b44776ffc..239e18b5e17 100644 --- a/doc/board/st/stm32mp1.rst +++ b/doc/board/st/stm32mp1.rst @@ -838,4 +838,4 @@ Arm TrustZone technology + https://www.op-tee.org/ + https://optee.readthedocs.io/en/latest/ + https://optee.readthedocs.io/en/latest/building/devices/stm32mp1.html - + https://github.com/OP-TEE/optee_os \ No newline at end of file + + https://github.com/OP-TEE/optee_os -- 2.43.0

2 7