Re: [PATCH 0/5] arm: Remove the 'musicpal' machine

[Jan Kiszka <jan.kiszka@web.de>]

On 19.08.26 20:54, Peter Maydell wrote:
> The 'musicpal' machine emulates the Freecom MusicPal, which was an
> "internet radio" released in the mid-2000s.  The stock system used a
> 2.6.16 Linux kernel, but as far as I'm aware, support for the
> hardware was never upstreamed to the mainline kernel.

Right. I once talked to the developer at Freecom behind the software,
but they were already in maintenance mode and apparently had neither the
power nor the will to influence Marvell on this.

> 
> There are a number of reasons for deprecating this board:
>  * The hardware is now solidly obsolete
>  * It was never supported in mainline Linux
>  * The SoC is a Marvell one with no public datasheet
>  * The only QEMU changes to it have been the usual tree wide
>    updates and refactoring work
>  * We have no test image for it that we can distribute
>  * The only bug report we've ever had about it has been one of
>    the usual "bug found by static analysis" kind (#4135)
> 
> But in fact we can just drop it immediately, because in commit
> 59754f85ed35 ("target/arm: Do memory type alignment check when
> translation disabled") in 2024 we broke the boot of the standard
> firmware image for this board.  (The image's u-boot does a LDRH to an
> unaligned address, which is UNPREDICTABLE; we used to let it through,
> but that commit makes it take a Data Abort, so the image crashes
> before it boots the kernel.) Since nobody complained, we can deduce
> that nobody's actually using this machine.
> 
> This patchset first fixes up a couple of places (notably the
> pflash_cfi02 tests) that were using or referring to the musicpal
> board; then it removes the board, followed by the devices used only by
> that board.
> 
> Strictly speaking it's possible that somebody is using this machine
> with some other image that doesn't have the same "does an unaligned
> LDRH" problem the stock firmware image does; but it seems unlikely
> to me. We can take the standard "deprecate for two releases, then
> drop" route if people prefer that.
> 
> I do also have a patch which reverts the "take a Data Abort on
> unaligned access" change for v5 and v6-with-SCTLR.U=0 CPUs, which I'll
> post separately.
> 
> thanks
> -- PMM
> 
> Peter Maydell (5):
>   tests/qtest/pflash-cfi02-test: Don't use musicpal machine for testing
>   hw/arm: Remove spurious WM8750 dependency from Realview Kconfig
>   hw/arm: Remove the 'musicpal' machine
>   hw: Remove Marvell 88W8618 devices
>   hw/audio: Remove wm8750 I2C audio device
> 
>  MAINTAINERS                             |   10 -
>  configs/devices/arm-softmmu/default.mak |    1 -
>  docs/about/removed-features.rst         |   10 +
>  docs/system/arm/musicpal.rst            |   19 -
>  docs/system/target-arm.rst              |    1 -
>  hw/arm/Kconfig                          |   16 -
>  hw/arm/meson.build                      |    1 -
>  hw/arm/musicpal.c                       | 1388 -----------------------
>  hw/audio/Kconfig                        |    4 -
>  hw/audio/marvell_88w8618.c              |  315 -----
>  hw/audio/meson.build                    |    2 -
>  hw/audio/wm8750.c                       |  738 ------------
>  hw/net/meson.build                      |    1 -
>  hw/net/mv88w8618_eth.c                  |  403 -------
>  include/hw/audio/wm8750.h               |   30 -
>  include/hw/net/mv88w8618_eth.h          |   13 -
>  tests/qtest/meson.build                 |    6 +-
>  tests/qtest/pflash-cfi02-test.c         |   38 +-
>  18 files changed, 36 insertions(+), 2960 deletions(-)
>  delete mode 100644 docs/system/arm/musicpal.rst
>  delete mode 100644 hw/arm/musicpal.c
>  delete mode 100644 hw/audio/marvell_88w8618.c
>  delete mode 100644 hw/audio/wm8750.c
>  delete mode 100644 hw/net/mv88w8618_eth.c
>  delete mode 100644 include/hw/audio/wm8750.h
>  delete mode 100644 include/hw/net/mv88w8618_eth.h
> 

RIP - no problems with this. At least it was allowed to turn 18 years
in-tree. ;)

Acked-by: Jan Kiszka <jan.kiszka@web.de>

Fun fact: My real device that once motivated this reverse-engineered
machine model is still alive and basically working. Its hard-coded
web-radio service is dead by now, but manually entered streams still
work. Only the power supply had to be replace a while ago.

Jan