From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:46580) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1eOOYj-0004hR-Ia for qemu-devel@nongnu.org; Mon, 11 Dec 2017 08:55:54 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1eOOYd-0005fp-S9 for qemu-devel@nongnu.org; Mon, 11 Dec 2017 08:55:53 -0500 References: <20171211122146.162430-1-borntraeger@de.ibm.com> From: David Hildenbrand Message-ID: Date: Mon, 11 Dec 2017 14:55:44 +0100 MIME-Version: 1.0 In-Reply-To: <20171211122146.162430-1-borntraeger@de.ibm.com> Content-Type: text/plain; charset=utf-8 Content-Language: en-US Content-Transfer-Encoding: quoted-printable Subject: Re: [Qemu-devel] [qemu-s390x] [PATCH v3 1/1] s390-ccw-virtio: allow for systems larger that 7.999TB List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Christian Borntraeger , Cornelia Huck Cc: Alexander Graf , Thomas Huth , Halil Pasic , qemu-devel , qemu-s390x , Richard Henderson On 11.12.2017 13:21, Christian Borntraeger wrote: > KVM does not allow memory regions > KVM_MEM_MAX_NR_PAGES, basically > limiting the memory per slot to 8TB-4k. As memory slots on s390/kvm mus= t > be a multiple of 1MB we need start a new memory region if we cross > 8TB-1M. >=20 > With that (and optimistic overcommitment in the kernel) I was able to > start a 24TB guest on a 1TB system. >=20 > Signed-off-by: Christian Borntraeger > --- > hw/s390x/s390-virtio-ccw.c | 28 +++++++++++++++++++++++++--- > 1 file changed, 25 insertions(+), 3 deletions(-) >=20 > diff --git a/hw/s390x/s390-virtio-ccw.c b/hw/s390x/s390-virtio-ccw.c > index 8425534..073f6ed 100644 > --- a/hw/s390x/s390-virtio-ccw.c > +++ b/hw/s390x/s390-virtio-ccw.c > @@ -154,14 +154,36 @@ static void virtio_ccw_register_hcalls(void) > virtio_ccw_hcall_early_printk); > } > =20 > +/* > + * KVM does only support memory slots up to KVM_MEM_MAX_NR_PAGES pages > + * as the dirty bitmap must be managed by bitops that take an int as > + * position indicator. If we have a guest beyond that we will split of= f > + * new subregions. The split must happen on a segment boundary (1MB). > + */ > +#define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1) > +#define SEG_MSK (~0xfffffULL) > +#define KVM_SLOT_MAX_BYTES ((KVM_MEM_MAX_NR_PAGES * TARGET_PAGE_SIZE) = & SEG_MSK) Just wondering if we could get into trouble when calculating KVM_MEM_MAX_NR_PAGES * TARGET_PAGE_SIZE on a host with sizeof(long) =3D=3D 4 could it wrap? (e.g. crazy mingw stuff) > static void s390_memory_init(ram_addr_t mem_size) > { > MemoryRegion *sysmem =3D get_system_memory(); > - MemoryRegion *ram =3D g_new(MemoryRegion, 1); > + ram_addr_t chunk, offset =3D 0; > + unsigned int number =3D 0; > + gchar *name; > =20 > /* allocate RAM for core */ > - memory_region_allocate_system_memory(ram, NULL, "s390.ram", mem_si= ze); > - memory_region_add_subregion(sysmem, 0, ram); > + name =3D g_strdup_printf("s390.ram"); > + while (mem_size) { > + MemoryRegion *ram =3D g_new(MemoryRegion, 1); I'd add an empty line here. Reviewed-by: David Hildenbrand > + /* KVM does not allow memslots >=3D 8 TB */ > + chunk =3D MIN(mem_size, KVM_SLOT_MAX_BYTES); > + memory_region_allocate_system_memory(ram, NULL, name, chunk); > + memory_region_add_subregion(sysmem, offset, ram); > + mem_size -=3D chunk; > + offset +=3D chunk; > + g_free(name); > + name =3D g_strdup_printf("s390.ram.%u", ++number); > + } > + g_free(name); > =20 > /* Initialize storage key device */ > s390_skeys_init(); >=20 --=20 Thanks, David / dhildenb