Re: [PATCH v3 07/11] hw/sh4/r2d: Realize IDE controller before accessing it

[Guenter Roeck]

On 10/12/24 07:06, Bernhard Beschow wrote:
> Am 12. Oktober 2024 09:40:27 UTC schrieb Thomas Huth <thuth@redhat.com>:
> > On 12/10/2024 00.48, Philippe Mathieu-Daudé wrote:
> > > On 11/10/24 05:23, Thomas Huth wrote:
> > > > On 03/05/2024 23.34, Guenter Roeck wrote:
> > > > > Hi,
> > > > >
> > > > > On Thu, Feb 08, 2024 at 07:12:40PM +0100, Philippe Mathieu-Daudé wrote:
> > > > > > We should not wire IRQs on unrealized device.
> > > > > >
> > > > > > Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> > > > > > Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
> > > > > > Reviewed-by: Yoshinori Sato <ysato@users.sourceforge.jp>
> > > > >
> > > > > qemu 9.0 fails to boot Linux from ide/ata drives with the sh4
> > > > > and sh4eb emulations. Error log is as follows.
> > > > >
> > > > > ata1.00: ATA-7: QEMU HARDDISK, 2.5+, max UDMA/100
> > > > > ata1.00: 16384 sectors, multi 16: LBA48
> > > > > ata1.00: configured for PIO
> > > > > scsi 0:0:0:0: Direct-Access     ATA      QEMU HARDDISK    2.5+ PQ: 0 ANSI: 5
> > > > > sd 0:0:0:0: [sda] 16384 512-byte logical blocks: (8.39 MB/8.00 MiB)
> > > > > sd 0:0:0:0: [sda] Write Protect is off
> > > > > sd 0:0:0:0: [sda] Write cache: enabled, read cache: enabled, doesn't support 
> > > > > DPO or FUA
> > > > > ata1: lost interrupt (Status 0x58)
> > > > >
> > > > > [ and more similar errors ]
> > > > >
> > > > > qemu command line:
> > > > >
> > > > > qemu-system-sh4eb -M r2d -kernel arch/sh/boot/zImage \
> > > > >      -snapshot -drive file=rootfs.ext2,format=raw,if=ide \
> > > > >      -append "root=/dev/sda console=ttySC1,115200 noiotrap" \
> > > > >      -serial null -serial stdio -monitor null -nographic -no-reboot
> > > > >
> > > > > Bisect points to this patch (see below). Reverting it fixes the problem.
> > >
> > > Sorry Guenter for missing your email :(
> > >
> > > >    Hi Philippe!
> > > >
> > > > Today I noticed that our sh4 test from tests/avocado/tuxrun_baselines.py is 
> > > > failing (which is not run by default, that's why nobody noticed), and bisection 
> > > > took me to the same result that Guenter already reported.
> > >
> > > "not run by default" because flaky.
> > >
> > > Having a quick look, the problem seems hw/ide/core.c uses non-QOM
> > > shortcuts. In particular ide_bus_init_output_irq() expects a preset
> > > IRQ. Not something trivial to fix.
> >
> > I wonder whether the other spots that use ide_bus_init_output_irq() or similar 
> > constructs are broken now, too. Bernhard apparently already fixed (reverted) 
> > one in commit 143f3fd3d8b51d6526c8ea80e8a2a085f6f01cdf.
> >
> > But what about fc432ba0f58343c8912b80e9056315bb9bd8df92 ?
>
> There is an indirection going on in pmac_ide_irq() which triggers real_*_irq. 
> These get wired via sysbus API after realize() while ide_bus_init_output_irq() 
> wires to pmac_ide_irq(). So fc432ba0f58343c8912b80e9056315bb9bd8df92 seems safe 
> to me (haven't tested it though).

Not sure if that answers the question, but booting from ide works
for both mac99 and g3beige emulations.

Guenter