qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] s390 - Parts of console output lost


From: Alexander Graf
Subject: Re: [Qemu-devel] s390 - Parts of console output lost
Date: Mon, 5 Apr 2010 20:04:14 +0200

On 05.04.2010, at 16:45, Bastian Blank wrote:

> Hi
> 
> 116348def2bb446d972bdc2f44bd77ff631f85de works on s390/kvm, but on startup
> parts of the console output is lost. The used kernel is a standard Debian
> distribution kernel.
> 
> | $ ./s390x-softmmu/qemu-system-s390x -nographic -enable-kvm -kernel 
> /boot/vmlinuz-2.6.32-4-s390x -initrd /boot/initrd.img-2.6.32-4-s390x -append 
> "break=premount"
> | [    0.583078] cio: Channel measurement facility initialized using format 
> basic (mode autodetected)
> | [    0.585475] TCP cubic registered
> 
> Everything up to the first line is lost on the virtio console. This
> could be also a kernel problem as this kernel includes two drivers
> responsible for hvc, virtio and iucv. The last line in the kernel log
> before the properly shown ones is:
> 
> | [    0.592667] hvc_iucv: The z/VM IUCV HVC device driver cannot be used 
> without z/VM

KVM for S390 does early printk using a special hypervisor callback. The patch 
to inject input from there into virtio-console was rejected upstream, because 
people believed it to be better to create a mux of an early and a real console 
device. While I agree that that approach would be superior I didn't get around 
to it yet and instead just carry this patch for personal development:


diff --git a/hw/s390-virtio.c b/hw/s390-virtio.c
index ad3386f..37d6c64 100644
--- a/hw/s390-virtio.c
+++ b/hw/s390-virtio.c
@@ -95,7 +99,10 @@ int s390_virtio_hypercall(CPUState *env)
             }
         } else {
             /* Early printk */
+
+            uint8_t *p = (uint8_t *)qemu_get_ram_ptr(mem);
+            printf("%s", p);
         }
         break;
     case KVM_S390_VIRTIO_RESET:
     {


Alex





reply via email to

[Prev in Thread] Current Thread [Next in Thread]