On Tue, Apr 15, 2025 at 03:36:53PM +0200, Jerome Forissier wrote: > Hi Tom, > > On 4/11/25 22:29, Tom Rini wrote: > > On Mon, Mar 31, 2025 at 02:30:54PM +0200, Jerome Forissier wrote: > > > >> This series introduces threads and uses them for two things: > >> - Improve the performance of the USB bus scanning code (scan multiple > >> busses simultaneously), > >> - Implement background jobs in the shell via two new commands: 'spawn' > >> and 'wait'. > >> > >> The threading framework is called 'uthread' and is inspired from the > >> barebox threads [2]. setjmp() and longjmp() are used to save and > >> restore contexts, as well as a non-standard extension called initjmp(). > >> This new function is added in several patches, one for each > >> architecture that supports HAVE_SETJMP. A new symbol is defined: > >> HAVE_INITJMP. Two tests, one for initjmp() and one for the uthread > >> scheduling, are added to the lib suite. > >> > >> The schedule() and udelay() function are updated to become thread > >> re-scheduling points, and mutex objects are introduced. With that the > >> USB stack initialization is modified to benefit from concurrency when > >> UTHREAD is enabled, where uthreads are used in usb_init() to initialize > >> and scan multiple busses at the same time. The code was tested on arm64 > >> and arm QEMU with 4 simulated XHCI buses and some devices. On this > >> platform the USB scan takes 2.2 s instead of 5.6 s. Tested on i.MX93 EVK > >> with two USB hubs, one ethernet adapter and one webcam on each, "usb > >> start" takes 2.4 s instead of 4.6 s. > >> > >> Finally, the spawn and wait commands are introduced, allowing the use of > >> threads from the shell. Tested on the i.MX93 EVK with a spinning HDD > >> connected to USB1 and the network connected to ENET1. The USB plus DHCP > >> init sequence "spawn usb start; spawn dhcp; wait" takes 4.5 seconds > >> instead of 8 seconds for "usb start; dhcp". > >> > >> [1] https://patchwork.ozlabs.org/project/uboot/list/?series=446674 > >> [2] https://github.com/barebox/barebox/blob/master/common/bthread.c > > > > So, we're nearly there. But there's some CI problems. First, CI isn't > > testing this because nothing enables UTHREADS. > > Thanks for looking into this. > > I do have a TESTING patch [0] that enables this for various QEMU plaforms > (arm, arm64, riscv32, riscv64) and for sandbox64. Please refer to the > following CI pipeline [1]. > > [0] https://source.denx.de/u-boot/custodians/u-boot-net/-/commit/8460f6abefff > [1] https://source.denx.de/u-boot/custodians/u-boot-net/-/pipelines/25706
Oh good, thanks. > > I don't know if we should > > enable it by default on qemu_arm64 for example or not. > > Good question. Perhaps we could have a separate defconfig for the somewhat > experimental stuff? Well, personally, I'd like to see the literal qemu targets (ie qemu_arm64 but not sifive_unleashed) enable UNIT_TEST and a few other things so that those features are tested somewhere, and I don't know the use case for U-Boot for QEMU virtual machine target for production, where the final use case isn't "... for testing $X". Happy to be informed otherwise of course. And then yes, we could just enable UTHREAD for at least some if not all qemu targets that support it, and test it there. Especially since it's not just a single flag to enable. > > But for testing, > > I added this to .gitlab-ci.yml (and something similar would work for > > Azure): > > qemu_arm64_uthread test.py: > > variables: > > TEST_PY_BD: "qemu_arm64" > > TEST_PY_TEST_SPEC: "not sleep" > > OVERRIDE: "-a CONFIG_UTHREAD=y -a CONFIG_CMD_SPAWN=y -a > > CONFIG_UNIT_TEST=y" > > <<: *buildman_and_testpy_dfn > > > > This however leads first to unrelated tests failing: > > https://source.denx.de/u-boot/u-boot/-/jobs/1099921 > > Unrelated indeed. It turns out UNIT_TEST=y on qemu_arm64 is broken in master. > I just posted a fix [2]. > > [2] https://lists.denx.de/pipermail/u-boot/2025-April/586897.html Ah thanks, nice. > > And I didn't see if the new tests ran either. > > It seems you need to enable CMD_TIME too. In my testing patch I added -v > (verbose) to the test.py command so that we can clearly see that > lib_initjmp [3], lib_uthread [4] and lib_uthread_mutex [5] are run in CI. > > [3] https://source.denx.de/u-boot/custodians/u-boot-net/-/jobs/1103031#L457 > [4] https://source.denx.de/u-boot/custodians/u-boot-net/-/jobs/1103031#L504 > [5] https://source.denx.de/u-boot/custodians/u-boot-net/-/jobs/1103031#L505 Ah very good, I was worried about hitting further issues. So I think in sum for v6, we just need to respin v5 such that some representative sets of qemu enable UTHREAD, and then a stand alone series (v2 of your series in [2]) to enable UNIT_TEST more widely in qemu. Thanks! -- Tom
signature.asc
Description: PGP signature
