On Fri, Jul 26, 2013 at 06:37:08PM +0200, Gerd Hoffmann wrote:
Qemu commit 9ee59f3 removed the bochs bios apm interface emulation at port 0x8900. That broke poweroff via APM. Fix it by powering off the machine using the acpi pm control register.
Old code is left in, so seabios wil try both poweroff methods. Cleaning that eventually up is left for another patch, after checking it isn't needed. Qemu never implemented "Standby" and "Suspend", only "Shutdown", so it looks like there might be non-qemu use cases (bochs probably).
Easiest way to test this is the syslinux poweroff module; modern linux distros usually have CONFIG_APM turned off.
Reported-by: Sebastian Herbszt herbszt@gmx.de Signed-off-by: Gerd Hoffmann kraxel@redhat.com
src/acpi.c | 5 +++++ src/acpi.h | 1 + src/apm.c | 2 ++ 3 files changed, 8 insertions(+)
diff --git a/src/acpi.c b/src/acpi.c index 3699898..44ba507 100644 --- a/src/acpi.c +++ b/src/acpi.c @@ -18,6 +18,8 @@
#include "acpi-dsdt.hex"
+u32 acpi_pm1a_cnt VARLOW = PORT_ACPI_PM_BASE + 0x04;
static void build_header(struct acpi_table_header *h, u32 sig, int len, u8 rev) { @@ -730,9 +732,12 @@ find_acpi_features(void) if (!fadt) return; u32 pm_tmr = le32_to_cpu(fadt->pm_tmr_blk);
u32 pm1a_cnt = le32_to_cpu(fadt->pm1a_cnt_blk); dprintf(4, "pm_tmr_blk=%x\n", pm_tmr); if (pm_tmr) pmtimer_setup(pm_tmr, 3579);
if (pm1a_cnt)
acpi_pm1a_cnt = pm1a_cnt;// Theoretically we should check the 'reset_reg_sup' flag, but Windows // doesn't and thus nobody seems to *set* it. If the table is large enough
diff --git a/src/acpi.h b/src/acpi.h index 5d1e104..f0d24d4 100644 --- a/src/acpi.h +++ b/src/acpi.h @@ -36,6 +36,7 @@ struct rsdp_descriptor { /* Root System Descriptor Pointer */ };
extern struct rsdp_descriptor *RsdpAddr; +extern u32 acpi_pm1a_cnt;
/* Table structure from Linux kernel (the ACPI tables are under the BSD license) */ diff --git a/src/apm.c b/src/apm.c index b2eac6d..b3b351c 100644 --- a/src/apm.c +++ b/src/apm.c @@ -12,6 +12,7 @@ #include "config.h" // CONFIG_* #include "biosvar.h" // GET_GLOBAL #include "paravirt.h" // runningOnQEMU +#include "acpi.h" // acpi_pm_ctl
static void out_str(const char *str_cs) @@ -109,6 +110,7 @@ void apm_shutdown(void) { irq_disable();
- outw(0x2000, acpi_pm1a_cnt);
This code is also run from 16bit protected mode and 32bit protected mode, so only GET_GLOBAL() access is permitted. Also, it can't default to PORT_ACPI_PM_BASE, because if a valid ACPI table isn't found we don't want to touch random ports.
Is the code to write always guaranteed to be 0x2000?
-Kevin