January 2026 - Concept

[PATCH 00/19] ext4l: Reduce ext4_uboot.h size by moving code to include/linux
by Simon Glass 17 Jan '26

17 Jan '26

From: Simon Glass <simon.glass(a)canonical.com> This series reduces fs/ext4l/ext4_uboot.h from 2584 to 1980 lines (a 23% reduction) by moving definitions to appropriate include/linux/ header files. The ext4l subsystem requires many Linux kernel types and functions. Rather than keeping everything in one large file, this series creates proper stub headers in include/linux/ matching Linux kernel organisation: - ktime.h: Time functions and ktime_t type - rcupdate.h: RCU stubs (single-threaded no-ops) - shrinker.h: Memory-reclaim stubs - mm_types.h: VM stubs (address_space, vm_area_struct) - pagemap.h: Page cache and folio stubs - bio.h: Block I/O stubs - fs_context.h/fs_parser.h: Filesystem context types - dcache.h: Dentry and qstr types - uuid.h: UUID/GUID types - smp.h: SMP memory barriers - refcount.h: Reference counting type - kobject.h: Kobject stubs - xarray.h: XArray stubs - mempool.h: Memory pool stubs - kdev_t.h: Device number macros - fsmap.h: Filesystem map definitions Additionally, cmpxchg(), try_cmpxchg(), and xchg() are moved to asm-generic/atomic.h, and blk_plug is added to linux/blkdev.h. All implementations are appropriate stubs for U-Boot's single-threaded, no-VM, no-page-cache environment. Simon Glass (19): linux: Add ktime.h header with time functions linux: Add rcupdate.h header with RCU stubs linux: Add shrinker.h header with memory-reclaim stubs linux: Add mm_types.h header with VM stubs linux: Add pagemap.h header with folio and page cache stubs linux: Add bio.h header with block I/O stubs linux: Add fs_context.h and fs_parser.h headers linux: Add dcache.h header with dentry stubs linux: Add uuid.h header with UUID/GUID types asm-generic: atomic: Add cmpxchg() macro linux: Add smp.h header with SMP stubs linux: Add refcount.h header with refcount_t type linux: Add kobject.h header with kobject stubs linux: Add xarray.h header with XArray stubs linux: Add mempool.h header with memory pool stubs linux: Add kdev_t.h header with device number macros linux: blkdev: Add blk_plug structure and operations linux: Add fsmap.h header with filesystem map definitions asm-generic: atomic: Add xchg() and try_cmpxchg() macros fs/ext4l/ext4_fscrypt.h | 9 +- fs/ext4l/ext4_uboot.h | 738 ++++------------------------------- include/asm-generic/atomic.h | 55 +++ include/linux/bio.h | 117 ++++++ include/linux/blkdev.h | 25 ++ include/linux/dcache.h | 94 +++++ include/linux/fs_context.h | 154 ++++++++ include/linux/fs_parser.h | 110 ++++++ include/linux/fsmap.h | 46 +++ include/linux/kdev_t.h | 56 +++ include/linux/kobject.h | 34 ++ include/linux/ktime.h | 70 ++++ include/linux/mempool.h | 57 +++ include/linux/mm_types.h | 116 ++++++ include/linux/pagemap.h | 201 ++++++++++ include/linux/rcupdate.h | 54 +++ include/linux/refcount.h | 70 ++++ include/linux/shrinker.h | 83 ++++ include/linux/smp.h | 60 +++ include/linux/uuid.h | 120 ++++++ include/linux/xarray.h | 50 +++ 21 files changed, 1640 insertions(+), 679 deletions(-) create mode 100644 include/linux/bio.h create mode 100644 include/linux/dcache.h create mode 100644 include/linux/fs_context.h create mode 100644 include/linux/fs_parser.h create mode 100644 include/linux/fsmap.h create mode 100644 include/linux/kdev_t.h create mode 100644 include/linux/kobject.h create mode 100644 include/linux/ktime.h create mode 100644 include/linux/mempool.h create mode 100644 include/linux/mm_types.h create mode 100644 include/linux/pagemap.h create mode 100644 include/linux/rcupdate.h create mode 100644 include/linux/refcount.h create mode 100644 include/linux/shrinker.h create mode 100644 include/linux/smp.h create mode 100644 include/linux/uuid.h create mode 100644 include/linux/xarray.h -- 2.43.0 base-commit: 3f7c0f2d98ec4d41d0f15578358219ae2fdf8b88 branch: extr

1 19

[PATCH 00/18] Refactor vidconsole context for dynamic allocation
by Simon Glass 17 Jan '26

17 Jan '26

From: Simon Glass <simon.glass(a)canonical.com> This series refactors the vidconsole subsystem to support dynamic context allocation. This is groundwork for allowing multiple independent contexts per vidconsole device, which will enable each expo textline to have its own context without needing save/restore logic. The main changes are: 1. Move fields from vidconsole_priv into vidconsole_ctx (patches 2-5) - curs, xstart_frac, tab_width_frac, xsize_frac - This consolidates per-context state into a single structure 2. Add infrastructure for dynamic context allocation (patches 8-10) - Add vidconsole_uc_plat with ctx_size for per-driver context sizes - Implement vidconsole_ctx_new/dispose for context lifecycle - Allocate vidconsole_ctx dynamically using ctx_size 3. Clean up truetype driver (patches 1, 11-12) - Rename vc_ctx to com for consistency - Remove embedded ctx from console_tt_priv - Use consistent ctx/com pattern throughout 4. Move truetype-specific fields to console_tt_ctx (patches 13-14) - cur_met, cur_fontdata now in console_tt_ctx - These are per-context rather than per-device 5. Simplify save/restore (patches 15-17) - Remove redundant store.cur field - Remove console_tt_store and console_store structs - Save/restore entire context directly 6. Add tests (patch 18) - Add dm_test_video_entry_save() for bitmap fonts - Add dm_test_video_entry_save_tt() for truetype fonts The next step (not in this series) is to add a ctx parameter to driver methods like putc_xy(), allowing callers to specify which context to use. This will enable expo textlines to each maintain their own cursor position and state. Simon Glass (18): video: truetype: Rename vc_ctx to com video: Move curs into vidconsole_ctx video: Move xstart_frac into vidconsole_ctx video: Move tab_width_frac into vidconsole_ctx video: Move xsize_frac into vidconsole_ctx video: Use vidconsole_ctx_from_priv() in expo_test video: Update cli_index_adjust() to use vidconsole_ctx video: Add vidconsole_uc_plat for per-device platform data video: Set ctx_size in console_truetype driver video: Allocate vidconsole_ctx dynamically using ctx_size video: truetype: Remove ctx from console_tt_priv video: truetype: Use consistent ctx/com pattern video: Move cur_met to console_tt_ctx video: Move cur_fontdata to console_tt_ctx video: truetype: Remove redundant store.cur field video: truetype: Remove console_tt_store struct video: normal: Remove console_store struct video: Add test for entry save/restore board/atmel/common/video_display.c | 4 +- boot/expo_test.c | 12 +- drivers/video/console_core.c | 8 +- drivers/video/console_normal.c | 29 +--- drivers/video/console_rotate.c | 6 +- drivers/video/console_truetype.c | 221 ++++++++++++----------------- drivers/video/vidconsole-uclass.c | 73 ++++++---- include/video_console.h | 50 ++++--- test/dm/video.c | 70 ++++++++- 9 files changed, 260 insertions(+), 213 deletions(-) -- 2.43.0 base-commit: 1f26d803f26bdf6a03494caaf5ae5b5df18c7905 branch: expc

1 18

[PATCH 00/14] video: Add per-client context to vidconsole
by Simon Glass 16 Jan '26

16 Jan '26

From: Simon Glass <simon.glass(a)canonical.com> This series refactors the vidconsole interface to support multiple clients with their own context. This will enables features like expo to maintain separate text-entry state without interfering with each other. The approach is to create a new struct vidconsole_ctx that holds per-client state, initially embedded in vidconsole_priv for the default context. Helper functions vidconsole_ctx() and vidconsole_ctx_from_priv() provide access to the context. Fields moved into vidconsole_ctx: - rows, cols, x_charsize, y_charsize (display metrics) - xcur_frac, ycur (cursor position) - last_ch (for kerning) - cli_index, xmark_frac, ymark (CLI entry state) - ansi, utf8_buf (parsing state) The series also adds ctx_new() and ctx_dispose() operations for drivers to allocate and free per-client context, with implementations for both the normal and truetype console drivers. Simon Glass (14): video: Add empty vidconsole_ctx struct video: Add vidconsole context creation and disposal video: truetype: Restore pos_start in entry_restore() video: truetype: Move per-client context into ctx struct video: Add a helper to obtain the vidconsole context video: Move rows and cols into vidconsole_ctx video: Move x_charsize and y_charsize into vidconsole_ctx video: Create struct vidconsole_ansi for ANSI state video: Move xcur_frac into vidconsole_ctx video: Move ycur into vidconsole_ctx video: Move last_ch into vidconsole_ctx video: Move cli_index into vidconsole_ctx video: Move xmark_frac and ymark into vidconsole_ctx video: Move ansi and utf8_buf into vidconsole_ctx board/atmel/common/video_display.c | 3 +- boot/expo_test.c | 10 +- common/console.c | 6 +- drivers/video/console_core.c | 3 +- drivers/video/console_normal.c | 66 +++++++-- drivers/video/console_rotate.c | 9 +- drivers/video/console_truetype.c | 186 +++++++++++++++---------- drivers/video/vidconsole-uclass.c | 214 ++++++++++++++++++----------- drivers/video/video-uclass.c | 4 +- include/video_console.h | 167 ++++++++++++++++------ lib/efi_loader/efi_console.c | 10 +- test/dm/video.c | 22 +++ 12 files changed, 473 insertions(+), 227 deletions(-) -- 2.43.0 base-commit: 26606c3849cb974cba6037c79edd0329f7b27a32 branch: expb

1 14

[PATCH 00/34] ext4l: Clean up ext4_uboot.h by moving definitions to standard headers
by Simon Glass 14 Jan '26

14 Jan '26

From: Simon Glass <simon.glass(a)canonical.com> This series reorganises fs/ext4l/ext4_uboot.h by moving definitions to their proper locations in the Linux compatibility headers. The goal is to reduce duplication and make these definitions available to other parts of U-Boot. The file is reduced from 3129 lines to 2584 lines (17% reduction). Changes fall into several categories: 1. Moving standard Linux definitions to include/linux/*.h: - Types: pgoff_t, timespec64, time64_t, blk_opf_t - Filesystem: dir_context, FMODE_*, DT_*, S_*, ATTR_*, SB_*, FALLOC_FL_*, RENAME_* - Buffer: buffer_migrate_folio, noop_dirty_folio - Misc: jiffies, printk variants, module stubs, seq_file 2. Creating new Linux compatibility headers: - linux/lockdep.h: lock dependency stubs - linux/percpu.h: per-CPU variable stubs 3. Consolidating ext4l-specific stubs: - fs/ext4l/ext4_trace.h: tracing stubs - fs/ext4l/ext4_fscrypt.h: fscrypt stubs 4. Using existing headers instead of duplicating: - linux/blk_types.h for blk_opf_t - linux/pagevec.h for folio_batch - linux/completion.h for completion stubs - linux/build_bug.h and linux/bug.h for assertion macros 5. Removing dead code and duplicate definitions This cleanup makes the ext4l code more maintainable and brings the Linux compatibility layer closer to the actual Linux kernel structure. There is still more which could be done, but the goal is not to empty the header file, since most of its contents is of no use to the rest of U-Boot. Simon Glass (34): ext4l: Move timespec64 and time64_t to linux/time.h ext4l: Move rol32 to linux/bitops.h ext4l: Move atomic operations to asm-generic/atomic.h ext4l: Consolidate trace stubs into ext4_trace.h ext4l: Remove duplicate macro definitions ext4l: Move flush_workqueue to linux/workqueue.h ext4l: Consolidate fscrypt stubs into ext4_fscrypt.h ext4l: Remove duplicate percpu_counter stubs ext4l: Use linux/completion.h for completion stubs ext4l: Convert dquot_suspend from function to macro ext4l: Remove dead iomap definitions ext4l: Create linux/percpu.h for per-CPU stubs ext4l: Remove redundant forward declarations ext4l: Move buffer operation stubs to linux/buffer_head.h ext4l: Move address_space_operations to linux/fs.h ext4l: Use linux/build_bug.h and linux/bug.h for macros ext4l: Move seq_file definitions to linux/seq_file.h ext4l: Use linux/pagevec.h for folio_batch ext4l: Move printk variants to linux/printk.h ext4l: Move jiffies definitions to linux/jiffies.h ext4l: Move module stubs to linux/module.h ext4l: Create linux/lockdep.h for lock dependency stubs ext4l: Move string helpers to standard locations ext4l: Move FMODE hash flags to linux/fs.h ext4l: Move struct dir_context to linux/fs.h ext4l: Move pgoff_t to linux/types.h ext4l: Use linux/blk_types.h for blk_opf_t ext4l: Move DT_* directory entry types to linux/fs.h ext4l: Move S_* inode flags to linux/fs.h ext4l: Move S_IRUGO to linux/fs.h ext4l: Move FALLOC_FL_* and RENAME_* flags to linux/fs.h ext4l: Move ATTR_* iattr flags to linux/fs.h ext4l: Move SB_* superblock flags to linux/fs.h ext4l: Move buffer migration stubs to linux/buffer_head.h fs/ext4l/ext4_fscrypt.h | 144 +++++++ fs/ext4l/ext4_trace.h | 172 ++++++++ fs/ext4l/ext4_uboot.h | 787 ++++++----------------------------- fs/ext4l/stub.c | 54 +-- include/asm-generic/atomic.h | 53 +++ include/linux/bitops.h | 20 + include/linux/buffer_head.h | 36 ++ include/linux/completion.h | 24 +- include/linux/fs.h | 125 +++++- include/linux/jiffies.h | 14 + include/linux/lockdep.h | 36 ++ include/linux/module.h | 5 + include/linux/percpu.h | 54 +++ include/linux/printk.h | 11 + include/linux/seq_file.h | 13 + include/linux/slab.h | 16 + include/linux/string.h | 41 ++ include/linux/time.h | 8 + include/linux/types.h | 3 + include/linux/workqueue.h | 4 + lib/string.c | 25 ++ 21 files changed, 920 insertions(+), 725 deletions(-) create mode 100644 fs/ext4l/ext4_fscrypt.h create mode 100644 fs/ext4l/ext4_trace.h create mode 100644 include/linux/lockdep.h create mode 100644 include/linux/percpu.h -- 2.43.0 base-commit: a5a45ea2b2f3503462ab0a4ac2163fdbc5ba2c2e branch: extq

1 34

[PATCH 0/3] codman: Add category system and CSV output
by Simon Glass 13 Jan '26

13 Jan '26

From: Simon Glass <simon.glass(a)canonical.com> This series adds a category system to codman for classifying source files by functional area (boot, drivers, networking, etc.). Categories and features are defined in a TOML configuration file (category.cfg). The category system is used by a new CSV output option, which generates machine-readable reports suitable for spreadsheet analysis. Other output formats (terminal, HTML) do not yet use categories. Simon Glass (3): codman: Add category module for file classification codman: Set up categories for the U-Boot codebase codman: Add CSV output with category support tools/codman/category.cfg | 854 ++++++++++++++++++++++++++++++++++ tools/codman/category.py | 147 ++++++ tools/codman/codman.py | 19 +- tools/codman/codman.rst | 89 ++++ tools/codman/output.py | 209 ++++++++- tools/codman/test_category.py | 279 +++++++++++ tools/codman/test_output.py | 212 +++++++++ 7 files changed, 1807 insertions(+), 2 deletions(-) create mode 100644 tools/codman/category.cfg create mode 100644 tools/codman/category.py create mode 100644 tools/codman/test_category.py create mode 100644 tools/codman/test_output.py -- 2.43.0 base-commit: c6b889d3fa570424ac3db8f555d1ca0373ec2145 branch: codmana

1 3

[PATCH 0/5] buildman: Improve toolchain selection and config adjustment
by Simon Glass 12 Jan '26

12 Jan '26

From: Simon Glass <simon.glass(a)canonical.com> This little series provides a few fixes for problems reported quite some time ago. Firstly, when downloading a toolchain via --fetch-arch for an architecture that matches the host (e.g., aarch64 on aarch64), two problems occur: 1. The kernel.org toolchain tarballs contain symlinks with doubled prefixes like 'x86_64-linux-x86_64-linux-gcc', causing an "ambiguous toolchains" warning. 2. The downloaded toolchain may be ignored in favour of a system-installed one (e.g., from a distribution package) because both have the same calculated priority. This series fixes these issues by: - Adding a new PRIORITY_DOWNLOADED level so downloaded toolchains are preferred over system-installed ones - Filtering out doubled-prefix binaries during toolchain scanning This series also improves the --adjust-cfg feature to use merge_config.sh for proper Kconfig dependency resolution, including 'imply' statements. Simon Glass (5): buildman: Update test_reproducible for real Kconfig resolution buildman: Add a board parameter to check_command() buildman: Use merge_config.sh for --adjust-cfg buildman: Prioritise downloaded toolchains over system ones buildman: Filter out doubled-prefix toolchain binaries test/Kconfig | 14 +++ tools/buildman/builderthread.py | 12 +-- tools/buildman/buildman.rst | 29 +++++- tools/buildman/cfgutil.py | 88 +++++++++++++++++ tools/buildman/func_test.py | 165 ++++++++++++++++++++++++++++---- tools/buildman/test.py | 67 +++++++++++++ tools/buildman/test_cfgutil.py | 31 ++++++ tools/buildman/toolchain.py | 49 +++++++++- 8 files changed, 422 insertions(+), 33 deletions(-) -- 2.43.0 base-commit: 9316673236b9cea174ebc03b8e43d4ab63536f0d branch: bms

1 5

[PATCH 00/14] buildman: Clean up pylint warnings in func_test
by Simon Glass 10 Jan '26

10 Jan '26

From: Simon Glass <simon.glass(a)canonical.com> This series addresses pylint warnings in tools/buildman/func_test.py, improving the code quality score from ~9.06 to 10.0/10, although with some disabling. Changes include: - Rename constants to UPPER_CASE and methods to snake_case (C0103) - Add missing module and function docstrings (C0114/C0116) - Fix unused variables and arguments (W0612/W0613) - Use isinstance() instead of type() for type checking (C0123) - Fix implicit string concatenation bug (W1404) - Use tools.read_file() instead of open() without encoding - Initialise attributes in setUp() instead of dynamically (W0201) - Break long lines to stay within 80 characters (C0301) - Refactor _handle_command() to use single return - Suppress unavoidable structural warnings (too-many-lines, etc.) Simon Glass (14): buildman: Rename module constants to UPPER_CASE in func_test buildman: Use snake_case for method names in func_test buildman: Remove unnecessary semicolons in func_test buildman: Use identity comparison for False in func_test buildman: Remove unused import in func_test buildman: Fix indentation in func_test buildman: Convert % formatting to f-strings in func_test buildman: Add missing docstrings in func_test buildman: Fix unused variable warnings in func_test buildman: Fix pylint style warnings in func_test buildman: Fix attribute and line-length warnings in func_test buildman: Fix misc pylint warnings in func_test buildman: Refactor _handle_command() to use single return buildman: Suppress pylint warnings for test structure tools/buildman/func_test.py | 432 ++++++++++++++++++++---------------- 1 file changed, 238 insertions(+), 194 deletions(-) -- 2.43.0 base-commit: 09a384a8c40462c8bbc521998c1c1967507a21d8 branch: bmr

1 14

[PATCH 00/26] boot: pxe: Add three-phase API and fix memory leaks
by Simon Glass 10 Jan '26

10 Jan '26

From: Simon Glass <simon.glass(a)canonical.com> This series adds a callback-free API for PXE/extlinux parsing that eparates the process into three distinct phases: parse, load, and boot. The traditional API uses callbacks during parsing to load files immediately. The new API instead collects file information (kernel, initrd, FDT, overlays) into a list during parsing, allowing callers to load files manually before booting. Include files are handled in a similar way. Rather than requiring the config files to be nul-termiated, the size of the file is provided and used to determine the end of the configuration file. New functions: - pxe_parse() - Parse config and return menu with file list - pxe_load() - Load files for a selected label (optional helper) - pxe_boot() - Boot using loaded files (renamed from pxe_do_boot) - pxe_cleanup() - Clean up context and menu The series also fixes several memory leaks in the parser found through malloc dump analysis. Size growth is unfortunately significant. On Thumb2 the memory-leak fixes add 116 bytes; removing the need for nul-termination takes 116 bytes and the alists and and general logic take another 60 Simon Glass (26): lib: abuf: Add abuf_init_addr() helper boot: pxe: Free bmp in pxe_menu_uninit() boot: pxe: Fix token memory leaks in parser boot: pxe: Free menu label in label_destroy() boot: pxe: Fix memory leak in parse_fdtoverlays() boot: pxe: Make pxe_prepare() static and document it boot: pxe: Simplify pxe_parse_include() to take only address boot: pxe: Add struct pxe_file for file loading boot: pxe: Replace pxe_fdtoverlay with pxe_file boot: pxe: Use the files list instead of fdtoverlays boot: pxe: Add a helper to add to the file list boot: pxe: Add the kernel to the files list when parsing boot: pxe: Add the initrd to the files list when parsing boot: pxe: Add the FDT to the files list when parsing boot: pxe: Set pxe_file_size in get_pxe_file() boot: pxe: Set pxe_file_size in extlinux_read_all() boot: pxe: Add size parameter to parse_pxefile() boot: pxe: Add pxe_process_str() for nul-terminated strings boot: pxe: Update parse_pxefile() to take struct abuf boot: pxe: Remove unused base parameter from parse_menu() boot: pxe: Remove unused base parameter from parse_pxefile_top() boot: pxe: Add a limit parameter to parser functions boot: pxe: Add a callback-free PXE API test: pxe: Add a test for the callback-free files API test: pxe: Add memory-leak check to files-API test doc: bootstd: Add PXE parser API documentation boot/ext_pxe_common.c | 5 +- boot/pxe_parse.c | 242 +++++++++++++++++++---------- boot/pxe_utils.c | 127 ++++++++++++--- cmd/pxe.c | 2 +- cmd/sysboot.c | 2 +- doc/develop/bootstd/extlinux.rst | 2 +- doc/develop/bootstd/index.rst | 1 + doc/develop/bootstd/pxe_api.rst | 194 +++++++++++++++++++++++ doc/develop/bootstd/pxelinux.rst | 2 +- include/abuf.h | 12 ++ include/pxe_utils.h | 176 ++++++++++++++++----- lib/abuf.c | 6 + test/boot/pxe.c | 255 +++++++++++++++++++++++++++---- test/lib/abuf.c | 23 +++ test/py/tests/test_pxe_parser.py | 11 ++ 15 files changed, 876 insertions(+), 184 deletions(-) create mode 100644 doc/develop/bootstd/pxe_api.rst -- 2.43.0 base-commit: c0b7ed1ad434eabf50bc8bdf7ef7f31423f4977d branch: fite

1 26

[PATCH 00/18] buildman: Split up the enormous Builder class
by Simon Glass 10 Jan '26

10 Jan '26

From: Simon Glass <simon.glass(a)canonical.com> This series extracts the ~800 lines of result-display code from builder.py into a new ResultHandler class. Builder handles building and file I/O, while ResultHandler handles all display of build results, progress, errors, sizes, and config changes. This is a complex and tedious refactor which is why have not done it for the past 5 years, but the time is now. Along the way, this: - Creates a DisplayOptions namedtuple for result-display settings - Adds col (terminal.Color) as a required Builder parameter - Creates a ResultHandler class with all display-related methods - Cleans up naming conventions with underscore prefixes for private members After this series, builder.py is reduced from ~2200 to ~1200 lines, making it more focused and maintainable. Unsurprisingly resulthandler.py ends up at just over 1000 lines. Simon Glass (18): buildman: Move Config class to cfgutil buildman: Move process_config() function to cfgutil buildman: Move cfgutil tests to test_cfgutil.py buildman: Move Outcome class to its own module buildman: Move OUTCOME_* constants to outcome module buildman: Move BoardStatus and ErrLine to outcome module buildman: Add a namedtuple for result-display settings buildman: Add col parameter to Builder buildman: Create a class for handling results buildman: Move size-display methods to ResultHandler buildman: Move error/warning display methods to ResultHandler buildman: Move board classification methods to ResultHandler buildman: Move error delta calculation to ResultHandler buildman: Move result summary methods to ResultHandler buildman: Move print_build_summary() to ResultHandler buildman: Move show_summary() and produce_result_summary() to ResultHandler buildman: Use underscore prefix for private methods buildman: Use underscore prefix for private member variables tools/buildman/builder.py | 1230 +++---------------------------- tools/buildman/cfgutil.py | 70 +- tools/buildman/control.py | 35 +- tools/buildman/main.py | 5 +- tools/buildman/outcome.py | 68 ++ tools/buildman/resulthandler.py | 1028 ++++++++++++++++++++++++++ tools/buildman/test.py | 290 ++------ tools/buildman/test_builder.py | 155 ++-- tools/buildman/test_cfgutil.py | 234 ++++++ 9 files changed, 1655 insertions(+), 1460 deletions(-) create mode 100644 tools/buildman/outcome.py create mode 100644 tools/buildman/resulthandler.py create mode 100644 tools/buildman/test_cfgutil.py -- 2.43.0 base-commit: c0b7ed1ad434eabf50bc8bdf7ef7f31423f4977d branch: bmq

1 18

[PATCH 00/32] boot: pxe: Refactor into separate load/setup phases
by Simon Glass 09 Jan '26

09 Jan '26

From: Simon Glass <simon.glass(a)canonical.com> This series refactors the PXE/extlinux boot path to separate file-loading from boot-setup. This is working toward a clean three-phase API: parse, load, boot. The refactoring makes the code more modular and testable, and allows callers to inspect loaded files before committing to boot. It builds on the previous series which added tests. This one adds a few more tests but mostly focuses on reorganising the code without changing behaviour. Key changes: - Split pxe_load_label() into load and setup phases, allowing flexible use by callers that need to load files without immediate boot setup - Improve FDT overlay handling: convert from string to alist, load to sequential addresses, and split loading from applying overlays - Move all file loading (kernel, initrd, FDT, overlays) into pxe_load_files() for better organisation - Add test for FIT images with embedded FDT (documents a pre-existing limitation where conf_fdt_str is NULL in this case) - Add comprehensive tests for the new APIs and overlay loading - Various cleanups: use IS_ENABLED(), use global working_fdt, export set_working_fdt_addr_quiet() Simon Glass (32): test: pxe: Preserve filesystem images when persist is enabled sandbox: bootz: Fix incorrect pointer passed to unmap_sysmem() test: pxe: Add helper to check menu output lines test: pxe: Test devicetree alias keywords boot: pxe: Add separate APIs for label selection and file loading boot: pxe: Update tests to use the new API test: pxe: Add a test for missing fdtoverlay_addr_r test: pxe: Add test for FIT images with embedded FDT boot: pxe: Add a flag to suppress file-retrieval messages boot: pxe: Print 'say' message when label is booted boot: pxe: Rename destroy_pxe_menu() to pxe_menu_uninit() boot: pxe: Add pxe_menu_init() function boot: pxe: Defer processing of include files boot: pxe: Extract pxe_load_files() from pxe_load_label() boot: pxe: Split filename handling from label_process_fdt() boot: pxe: Add FDT pointer to context for kaslrseed boot: pxe: Move FIT handling from label_process_fdt() to caller boot: pxe: Move fdt_addr_r check to caller of label_load_fdt() boot: pxe: Use early return in label_load_fdt() boot: pxe: Use ctx->fdt in label_boot_fdtoverlay() boot: pxe: Move label_get_fdt_path() call to pxe_load_label() boot: pxe: Load FDT files in pxe_load_files() boot: pxe: Use IS_ENABLED() for CONFIG_OF_LIBFDT_OVERLAY boot: pxe: Drop fdtoverlay_addr_r check in label_boot_fdtoverlay() boot: pxe: Convert fdtoverlays from string to alist boot: pxe: Load overlays to sequential addresses boot: pxe: Split overlay loading and applying into two passes boot: pxe: Split label_boot_fdtoverlay() into load and apply functions boot: pxe: Move overlay loading to pxe_load_files() boot: pxe: Add test for FDT overlay loading boot: pxe: Split pxe_load_label() into load and setup phases boot: pxe: Extract boot preparation into separate function arch/sandbox/lib/bootm.c | 2 +- boot/pxe_parse.c | 98 ++-- boot/pxe_utils.c | 859 +++++++++++++++++++------------ include/pxe_utils.h | 152 +++++- test/boot/pxe.c | 559 ++++++++++++++++---- test/py/tests/test_pxe_parser.py | 107 +++- 6 files changed, 1300 insertions(+), 477 deletions(-) -- 2.43.0 base-commit: bb4eacf4d4aa58b13ce091f68b6ef9ee6624ae93 branch: fitd

1 32