MIPS Malta fails booting due to IDE error

[Philippe Mathieu-Daudé]

As of commit 3e407488349:

$ avocado --show=console run -t machine:malta tests/acceptance/boot_linux_console.py
console: [ 0.000000] Linux version 4.5.0-2-4kc-malta (<email address hidden>) (gcc version 5.3.1 20160519 (Debian 5.3.1-20) ) #1 Debian 4.5.5-1 (2016-05-29)
console: [ 0.000000] earlycon: Early serial console at I/O port 0x3f8 (options '38400n8')
console: [ 0.000000] bootconsole [uart0] enabled
console: [ 0.000000] CPU0 revision is: 00019300 (MIPS 24Kc)
console: [ 0.000000] FPU revision is: 00739300
console: [ 0.000000] MIPS: machine is mti,malta
[...]
console: ata2.00: ATAPI: QEMU DVD-ROM, 2.5+, max UDMA/100
console: ata2.00: Drive reports diagnostics failure. This may indicate a drive
console: ata2.00: fault or invalid emulation. Contact drive vendor for information.
console: ata2.00: configured for UDMA/33
console: scsi 1:0:0:0: CD-ROM QEMU QEMU DVD-ROM 2.5+ PQ: 0 ANSI: 5
console: Freeing unused kernel memory: 412K (80979000 - 809e0000)
console: do_page_fault(): sending SIGSEGV to mount for invalid write access to 0018a000
console: epc = 775cca54 in libc-2.27.so[775b3000+177000]
console: ra = 77754618 in ld-2.27.so[77743000+24000]
console: do_page_fault(): sending SIGSEGV to klogd for invalid write access to 0018a000
console: epc = 770f0a54 in libc-2.27.so[770d7000+177000]
console: ra = 77278618 in ld-2.27.so[77267000+24000]
console: do_page_fault(): sending SIGSEGV to S20urandom for invalid write access to 0018a000
console: epc = 77d0ca54 in libc-2.27.so[77cf3000+177000]
console: ra = 77e94618 in ld-2.27.so[77e83000+24000]
console: do_page_fault(): sending SIGSEGV to mkdir for invalid write access to 0018a000
console: epc = 776b8a54 in libc-2.27.so[7769f000+177000]
console: ra = 77840618 in ld-2.27.so[7782f000+24000]
console: do_page_fault(): sending SIGSEGV to sh for invalid write access to 0018a000
console: epc = 77364a54 in libc-2.27.so[7734b000+177000]
console: ra = 774ec618 in ld-2.27.so[774db000+24000]
console: do_page_fault(): sending SIGSEGV to sh for invalid write access to 0018a000
console: epc = 77bd4a54 in libc-2.27.so[77bbb000+177000]
console: ra = 77d5c618 in ld-2.27.so[77d4b000+24000]
console: do_page_fault(): sending SIGSEGV to awk for invalid write access to 0018a000
console: epc = 76f44a54 in libc-2.27.so[76f2b000+177000]
console: ra = 770cc618 in ld-2.27.so[770bb000+24000]
console: do_page_fault(): sending SIGSEGV to cat for invalid write access to 0018a000
console: epc = 770cca54 in libc-2.27.so[770b3000+177000]
console: ra = 77254618 in ld-2.27.so[77243000+24000]
$ echo $?
8

55adb3c45620c31f29978f209e2a44a08d34e2da is the first bad commit
commit 55adb3c45620c31f29978f209e2a44a08d34e2da
Author: John Snow <email address hidden>
Date: Fri Jul 24 01:23:00 2020 -0400

    ide: cancel pending callbacks on SRST

    The SRST implementation did not keep up with the rest of IDE; it is
    possible to perform a weak reset on an IDE device to remove the BSY/DRQ
    bits, and then issue writes to the control/device registers which can
    cause chaos with the state machine.

    Fix that by actually performing a real reset.

[John Snow (jnsnow)]

Yup. Mark Cave-Ayland pointed this out to me. I have a patch ready for it:

diff --git a/hw/ide/core.c b/hw/ide/core.c
index 693b352d5e..98cea7ad45 100644
--- a/hw/ide/core.c
+++ b/hw/ide/core.c
@@ -2254,10 +2254,8 @@ static void ide_perform_srst(IDEState *s)
      /* Cancel PIO callback, reset registers/signature, etc */
      ide_reset(s);

- if (s->drive_kind == IDE_CD) {
- /* ATAPI drives do not set READY or SEEK */
- s->status = 0x00;
- }
+ /* perform diagnostic */
+ cmd_exec_dev_diagnostic(s, WIN_DIAGNOSE);
  }

  static void ide_bus_perform_srst(void *opaque)
@@ -2282,9 +2280,7 @@ void ide_ctrl_write(void *opaque, uint32_t addr, uint32_t val)

      /* Device0 and Device1 each have their own control register,
       * but QEMU models it as just one register in the controller. */
- if ((bus->cmd & IDE_CTRL_RESET) &&
- !(val & IDE_CTRL_RESET)) {
- /* SRST triggers on falling edge */
+ if (!(bus->cmd & IDE_CTRL_RESET) && (val & IDE_CTRL_RESET)) {
          for (i = 0; i < 2; i++) {
              s = &bus->ifs[i];
              s->status |= BUSY_STAT;

[Philippe Mathieu-Daudé (philmd)]

Fixed by commits 4ac4e7281a2..1a9925e3390.

[Thomas Huth (th-huth)]

Released with QEMU v5.2.0.