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

Re: [Minios-devel] [UNIKRAFT PATCHv5 36/46] plat/kvm: Parse memory info from device tree for Arm64


  • To: Simon Kuenzer <simon.kuenzer@xxxxxxxxx>, "minios-devel@xxxxxxxxxxxxxxxxxxxx" <minios-devel@xxxxxxxxxxxxxxxxxxxx>
  • From: "Wei Chen (Arm Technology China)" <Wei.Chen@xxxxxxx>
  • Date: Wed, 12 Sep 2018 03:23:25 +0000
  • Accept-language: en-US
  • Authentication-results: spf=none (sender IP is ) smtp.mailfrom=Wei.Chen@xxxxxxx;
  • Cc: "Kaly Xin \(Arm Technology China\)" <Kaly.Xin@xxxxxxx>, nd <nd@xxxxxxx>
  • Delivery-date: Wed, 12 Sep 2018 03:23:36 +0000
  • List-id: Mini-os development list <minios-devel.lists.xenproject.org>
  • Nodisclaimer: True
  • Spamdiagnosticmetadata: NSPM
  • Spamdiagnosticoutput: 1:99
  • Thread-index: AQHUMHk5EnD1xEdo40669OmenGaly6TrtXmAgAB3oZA=
  • Thread-topic: [Minios-devel] [UNIKRAFT PATCHv5 36/46] plat/kvm: Parse memory info from device tree for Arm64


> -----Original Message-----
> From: Simon Kuenzer <simon.kuenzer@xxxxxxxxx>
> Sent: 2018年9月12日 4:09
> To: Wei Chen (Arm Technology China) <Wei.Chen@xxxxxxx>; minios-
> devel@xxxxxxxxxxxxxxxxxxxx
> Cc: Kaly Xin (Arm Technology China) <Kaly.Xin@xxxxxxx>; nd <nd@xxxxxxx>
> Subject: Re: [Minios-devel] [UNIKRAFT PATCHv5 36/46] plat/kvm: Parse memory
> info from device tree for Arm64
> 
> 
> 
> On 10.08.2018 09:08, Wei Chen wrote:
> > From: Wei Chen <Wei.Chen@xxxxxxx>
> >
> > QEMU/KVM will store the memory informations like memory
> > region, memory base address and memory size to device
> > tree. We parse these informations for memory allocater and
> > new stack setting. QEMU/KVM will always provide us at least
> > one bank of memory and unikraft will use the first bank for
> > the time-being. So currently we only supported one bank.
> >
> > Signed-off-by: Wei Chen <Wei.Chen@xxxxxxx>
> > ---
> >   plat/kvm/arm/setup.c | 76 ++++++++++++++++++++++++++++++++++++++++++++
> >   1 file changed, 76 insertions(+)
> >
> > diff --git a/plat/kvm/arm/setup.c b/plat/kvm/arm/setup.c
> > index d59956a..a74c85f 100644
> > --- a/plat/kvm/arm/setup.c
> > +++ b/plat/kvm/arm/setup.c
> > @@ -21,7 +21,13 @@
> >   #include <libfdt.h>
> >   #include <kvm/console.h>
> >   #include <uk/assert.h>
> > +#include <mm.h>
> > +#include <uk/arch/limits.h>
> >
> > +void *_libkvmplat_pagetable;
> > +void *_libkvmplat_heap_start;
> > +void *_libkvmplat_stack_top;
> > +void *_libkvmplat_mem_end;
> >   void *_libkvmplat_dtb;
> >
> >   #define MAX_CMDLINE_SIZE 1024
> > @@ -38,6 +44,68 @@ static void _init_dtb(void *dtb_pointer)
> >     uk_printd(DLVL_INFO, "Found device tree on: %p\n", dtb_pointer);
> >   }
> >
> > +static void _init_dtb_mem(void)
> > +{
> > +   extern char _text[];
> > +   extern char _end[];
> > +   int fdt_mem, prop_len = 0, prop_min_len;
> > +   int naddr, nsize;
> > +   const uint64_t *regs;
> > +   uint64_t mem_base, mem_size, max_addr;
> > +
> > +   /* search for assigned VM memory in DTB */
> > +   if (fdt_num_mem_rsv(_libkvmplat_dtb) != 0)
> > +           uk_printd(DLVL_WARN, "Reserved memory is not supported\n");
> > +
> > +   fdt_mem = fdt_node_offset_by_prop_value(_libkvmplat_dtb, -1,
> > +                                           "device_type",
> > +                                           "memory", sizeof("memory"));
> > +   if (fdt_mem < 0) {
> > +           uk_printd(DLVL_WARN, "No memory found in DTB\n");
> > +           return;
> > +   }
> > +
> > +   naddr = fdt_address_cells(_libkvmplat_dtb, fdt_mem);
> > +   if (naddr < 0 || naddr >= FDT_MAX_NCELLS)
> > +           UK_CRASH("Could not find proper address cells!\n");
> > +
> > +   nsize = fdt_size_cells(_libkvmplat_dtb, fdt_mem);
> > +   if (nsize < 0 || nsize >= FDT_MAX_NCELLS)
> > +           UK_CRASH("Could not find proper size cells!\n");
> > +
> > +   /*
> > +    * QEMU will always provide us at least one bank of memory.
> > +    * unikraft will use the first bank for the time-being.
> > +    */
> > +   regs = fdt_getprop(_libkvmplat_dtb, fdt_mem, "reg", &prop_len);
> > +
> > +   /*
> > +    * The property must contain at least the start address
> > +    * and size, each of which is 8-bytes.
> > +    */
> > +   prop_min_len = (int)sizeof(fdt32_t) * (naddr + nsize);
> > +   if (regs == NULL || prop_len < prop_min_len)
> > +           UK_CRASH("Bad 'reg' property: %p %d\n", regs, prop_len);
> > +
> > +   /* If we have more than one memory bank, give a warning messasge */
> > +   if (prop_len > prop_min_len)
> > +           uk_printd(DLVL_WARN,
> > +                   "Currently, we support only one memory bank!\n");
> > +
> > +   mem_base = fdt64_to_cpu(regs[0]);
> > +   mem_size = fdt64_to_cpu(regs[1]);
> > +   if (mem_base > (uint64_t)&_text)
> > +           UK_CRASH("Fatal: Image outside of RAM\n");
> > +
> > +   max_addr = mem_base + mem_size;
> > +   _libkvmplat_pagetable =(void *) ALIGN_UP((size_t)&_end, __PAGE_SIZE);
> > +   _libkvmplat_heap_start = _libkvmplat_pagetable + PAGE_TABLE_SIZE;
> > +   _libkvmplat_mem_end = (void *) max_addr;
> > +
> > +   /* AArch64 require stack be 16-bytes alignment by default */
> > +   _libkvmplat_stack_top = (void *) ALIGN_UP(max_addr, __STACK_ALIGN_SIZE);
> 
> Shouldn't this be ALIGN_DOWN()? We may leave our RAM region, right?

Yes, @_@ I think I must be muddled while I was writing this code.

> > +}
> > +
> >   static void _dtb_get_cmdline(char *cmdline, size_t maxlen)
> >   {
> >     int fdtchosen, len;
> > @@ -72,5 +140,13 @@ void _libkvmplat_start(void *dtb_pointer)
> >     uk_printd(DLVL_INFO, "Entering from KVM (arm64)...\n");
> >
> >     /* Get command line from DTB */
> > +
> >     _dtb_get_cmdline(cmdline, sizeof(cmdline));
> > +
> > +   /* Initialize memory from DTB */
> > +   _init_dtb_mem();
> > +
> > +   uk_printd(DLVL_INFO, "pagetable start: %p\n", _libkvmplat_pagetable);
> > +   uk_printd(DLVL_INFO, "     heap start: %p\n", _libkvmplat_heap_start);
> > +   uk_printd(DLVL_INFO, "      stack top: %p\n", _libkvmplat_stack_top);
> >   }
> >
_______________________________________________
Minios-devel mailing list
Minios-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/minios-devel

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.