qemu-devel
[Top][All Lists]
Advanced

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

Re: [PATCH v1] mips/mips_malta: Allow more than 2G RAM


From: Philippe Mathieu-Daudé
Subject: Re: [PATCH v1] mips/mips_malta: Allow more than 2G RAM
Date: Sat, 14 Mar 2020 10:09:08 +0100
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101 Thunderbird/68.4.1

Hi Aleksandar,

(+Aurelien for Debian)
(+Peter regarding changing default)

On 3/14/20 4:25 AM, Aleksandar Markovic wrote:
On Thu, Mar 5, 2020 at 11:18 AM Philippe Mathieu-Daudé
<address@hidden> wrote:

Please post new patches as v2, and do not post them as reply to v1.

On 3/3/20 1:41 AM, Jiaxun Yang wrote:
When malta is coupled with MIPS64 cpu which have 64bit
address space, it is possible to have more than 2G RAM.

So we removed ram_size check and overwrite memory
layout for these targets.

Signed-off-by: Jiaxun Yang <address@hidden>
Suggested-by: Yunqiang Su <address@hidden>
--
v1: Do not overwrite cmdline when we don't have highmem.
---
   hw/mips/mips_malta.c | 29 +++++++++++++++++++++++------
   1 file changed, 23 insertions(+), 6 deletions(-)

diff --git a/hw/mips/mips_malta.c b/hw/mips/mips_malta.c
index 6e7ba9235d..4267958f35 100644
--- a/hw/mips/mips_malta.c
+++ b/hw/mips/mips_malta.c
@@ -98,7 +98,8 @@ typedef struct {
   } MaltaState;

   static struct _loaderparams {
-    int ram_size, ram_low_size;
+    unsigned int ram_low_size;
+    ram_addr_t ram_size;
       const char *kernel_filename;
       const char *kernel_cmdline;
       const char *initrd_filename;
@@ -1023,6 +1024,7 @@ static int64_t load_kernel(void)
   {
       int64_t kernel_entry, kernel_high, initrd_size;
       long kernel_size;
+    char mem_cmdline[128];
       ram_addr_t initrd_offset;
       int big_endian;
       uint32_t *prom_buf;
@@ -1099,20 +1101,33 @@ static int64_t load_kernel(void)
       prom_buf = g_malloc(prom_size);

       prom_set(prom_buf, prom_index++, "%s", loaderparams.kernel_filename);
+
+    memset(&mem_cmdline[0], 0, sizeof(mem_cmdline));
+    if (loaderparams.ram_size > 0x10000000) {

Please use 256 * MiB.

+        /*
+         * Always use cmdline to overwrite mem layout
+         * as kernel may reject large emesize.
+         */
+        sprintf(&mem_cmdline[0],
+                "mem=0x10000000@0x00000000 mem=0x%" PRIx64 "@0x90000000",
+                loaderparams.ram_size - 0x10000000);

Ditto.

Also please use g_strdup_printf("mem=0x%" PRIx64 "@...")/g_free.

+    }
+
       if (initrd_size > 0) {
           prom_set(prom_buf, prom_index++,
-                 "rd_start=0x%" PRIx64 " rd_size=%" PRId64 " %s",
-                 xlate_to_kseg0(NULL, initrd_offset),
+                 "%s rd_start=0x%" PRIx64 " rd_size=%" PRId64 " %s",
+                 &mem_cmdline[0], xlate_to_kseg0(NULL, initrd_offset),
                    initrd_size, loaderparams.kernel_cmdline);
       } else {
-        prom_set(prom_buf, prom_index++, "%s", loaderparams.kernel_cmdline);
+        prom_set(prom_buf, prom_index++, "%s %s", &mem_cmdline[0],
+                 loaderparams.kernel_cmdline);
       }

       prom_set(prom_buf, prom_index++, "memsize");
       prom_set(prom_buf, prom_index++, "%u", loaderparams.ram_low_size);

       prom_set(prom_buf, prom_index++, "ememsize");
-    prom_set(prom_buf, prom_index++, "%u", loaderparams.ram_size);
+    prom_set(prom_buf, prom_index++, "%lu", loaderparams.ram_size);

       prom_set(prom_buf, prom_index++, "modetty0");
       prom_set(prom_buf, prom_index++, "38400n8r");
@@ -1253,12 +1268,14 @@ void mips_malta_init(MachineState *machine)
       /* create CPU */
       mips_create_cpu(machine, s, &cbus_irq, &i8259_irq);

-    /* allocate RAM */
+#ifdef TARGET_MIPS32
+    /* MIPS32 won't accept more than 2GiB RAM due to limited address space */

Cc'ing Paul Burton due to commit 94c2b6aff43.

       if (ram_size > 2 * GiB) {
           error_report("Too much memory for this machine: %" PRId64 "MB,"
                        " maximum 2048MB", ram_size / MiB);

This is annoying, because the CoreLV/CoreFPGA core cards only have 4
DIMM slots for PC-100 SDRAM, and the Memory Controller of the GT–64120A
north bridge accept at most 256 MiB per SCS for address decoding, so we
have a maximum of 1 GiB on 32-bit boards.

In 64-bit emulation we use the a 20Kc core, provided by the Core20K core
card which doesn't use the GT–64120A but the Bonito64. So the model is
incorrect... The Bonito64 indeed allow more memory.

Maybe it is time to consider that for 64-bit targets, since we are not
modelling a Malta core board, we can name it the official 'virt' machine...


Philippe, I almost agree 100% with you wrt last three paragraphs.
(in any case, I know much less than you about details of GT-64120A
and Bonito64, but I trust you).

The only area I have a slightly different opinion is that I believe we
should never declare Malta as a virtual board, and try to be within the
boundaries of physical capabilities of real boards, even if in software
(QEMU) we could "enhance" some features.

If we want MIPS virtual board, than we should devise a full blown
virtual board, similar to what was already done for MIPS Android
emulator. I really don't want some real/virtual frankenstein in QEMU
upstream just because it is convenient for let's say a particular
test setup.

IIUC today all distributions supporting MIPS ports are building their MIPS packages on QEMU instances because it is faster than the native MIPS hardware they have.

Since one (or two?) years, some binaries (Linux kernel? QEMU?) are failing to link because the amount of guest memory is restricted to 2GB (probably advance of linker techniques, now linkers use more memory).

YunQiang, is this why you suggested this change?

See:
- https://www.mail-archive.com/address@hidden/msg10912.html
- https://alioth-lists.debian.net/pipermail/pkg-rust-maintainers/2019-January/004844.html

I believe most of the QEMU Malta board users don't care it is a Malta board, they only care it is a fast emulated MIPS machine.
Unfortunately it is the default board.

However 32-bit MIPS port is being dropped on Debian:
https://lists.debian.org/debian-mips/2019/07/msg00010.html

Maybe we can sync with the Malta users, ask them to switch to the Boston machines to build 64-bit packages, then later reduce the Malta board to 1GB. (The Boston board is more recent, but was not available at the time users started to use QEMU to build 64-bit packages).

Might it be easier starting introducing a malta-5.0 machine restricted to 1GB?


Aleksandar

           exit(1);
       }
+#endif

       /* register RAM at high address where it is undisturbed by IO */
       memory_region_add_subregion(system_memory, 0x80000000, machine->ram);








reply via email to

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