Re: [qemu-s390x] [Qemu-devel] [PATCH v4 02/11] machine: make MemoryHotplugState accessible via the machine

[Paolo Bonzini]

On 04/05/2018 21:26, Eduardo Habkost wrote:
> On Mon, Apr 23, 2018 at 06:51:17PM +0200, David Hildenbrand wrote:
> [...]
>> +    /* always allocate the device memory information */
>> +    machine->device_memory = g_malloc(sizeof(*machine->device_memory));
> [...]
>> -    /* initialize hotplug memory address space */
>> +    /* always allocate the device memory information */
>> +    machine->device_memory = g_malloc(sizeof(*machine->device_memory));
> 
> This makes QEMU crash because machine->device_memory->base is initialized with
> garbage:
> 
> #1  0x00007fffef30a8f8 in abort () at /lib64/libc.so.6
> #2  0x00007fffef302026 in __assert_fail_base () at /lib64/libc.so.6
> #3  0x00007fffef3020d2 in  () at /lib64/libc.so.6
> #4  0x0000555555833483 in int128_get64 (a=<optimized out>) at 
> .../qemu-build/include/qemu/int128.h:22
> #5  0x0000555555837c2e in memory_region_size (a=<optimized out>) at 
> .../qemu-build/memory.c:1735
> #6  0x0000555555837c2e in memory_region_size (mr=<optimized out>) at 
> .../qemu-build/memory.c:1739
> #7  0x00005555558a2b14 in pc_memory_init (address@hidden, address@hidden, 
> address@hidden, address@hidden)
>     at .../qemu-build/hw/i386/pc.c:1440
> #8  0x00005555558a5a73 in pc_init1 (machine=0x555556850050, 
> pci_type=0x555555cb6fd0 "i440FX", host_type=0x555555c43e41 "i440FX-pcihost") 
> at .../qemu-build/hw/i386/pc_piix.c:179
> #9  0x00005555559abbda in machine_run_board_init (machine=0x555556850050) at 
> .../qemu-build/hw/core/machine.c:829
> #10 0x00005555557dc515 in main (argc=<optimized out>, argv=<optimized out>, 
> envp=<optimized out>) at .../qemu-build/vl.c:4563
> 
> 
> I will squash the following fixup:
> 
> From 6216fdb28476ed21c4ced4672003c9c7cb0e04d2 Mon Sep 17 00:00:00 2001
> From: David Hildenbrand <address@hidden>
> Date: Fri, 4 May 2018 15:54:46 +0200
> Subject: [PATCH] memory-device: fix device_memory creation on pc and spapr
> 
> We have to inititalize the struct to 0. Otherwise, without "maxmem",
> the content is undefined, which might result in random asserts
> striking when e.g. reading out the size of the contained memory region.
> 
> Signed-off-by: David Hildenbrand <address@hidden>
> ---
>  hw/i386/pc.c   | 2 +-
>  hw/ppc/spapr.c | 2 +-
>  2 files changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/hw/i386/pc.c b/hw/i386/pc.c
> index ffcd7b85d9..868893d0a1 100644
> --- a/hw/i386/pc.c
> +++ b/hw/i386/pc.c
> @@ -1372,7 +1372,7 @@ void pc_memory_init(PCMachineState *pcms,
>      }
>  
>      /* always allocate the device memory information */
> -    machine->device_memory = g_malloc(sizeof(*machine->device_memory));
> +    machine->device_memory = g_malloc0(sizeof(*machine->device_memory));
>  
>      /* initialize device memory address space */
>      if (pcmc->has_reserved_memory &&
> diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> index ef05075232..a1abcba6ad 100644
> --- a/hw/ppc/spapr.c
> +++ b/hw/ppc/spapr.c
> @@ -2637,7 +2637,7 @@ static void spapr_machine_init(MachineState *machine)
>      memory_region_add_subregion(sysmem, 0, ram);
>  
>      /* always allocate the device memory information */
> -    machine->device_memory = g_malloc(sizeof(*machine->device_memory));
> +    machine->device_memory = g_malloc0(sizeof(*machine->device_memory));

g_new0 since you are at it? :)

Paolo

>  
>      /* initialize hotplug memory address space */
>      if (machine->ram_size < machine->maxram_size) {
>