[Qemu-devel] [Bug 1814343] [NEW] Initrd not loaded on riscv32

From: Jonathan Behrens <1814343@bugs.launchpad.net>
To: qemu-devel@nongnu.org
Subject: [Qemu-devel] [Bug 1814343] [NEW] Initrd not loaded on riscv32
Date: Fri, 01 Feb 2019 22:06:34 -0000	[thread overview]
Message-ID: <154905879413.25667.2843143192142684785.malonedeb@chaenomeles.canonical.com> (raw)

Public bug reported:

I attempted to run qemu with a ram disk. However, when reading the
contents of the disk from within the VM I only get back zeros.

I was able to trace the issue to a mismatch of expectations on line 93
of hw/riscv/virt.c. Specifically, when running in 32-bit mode the value
of kernel_entry is sign extended to 64-bits, but load_image_targphys
expects the start address to not be sign extended.

Straw man patch (works for 32-bit but would probably break 64-bit VMs?):

diff --git a/hw/riscv/virt.c b/hw/riscv/virt.c
index e7f0716fb6..32216f993c 100644
--- a/hw/riscv/virt.c
+++ b/hw/riscv/virt.c
@@ -90,7 +90,7 @@ static hwaddr load_initrd(const char *filename, uint64_t mem_size,
      * halfway into RAM, and for boards with 256MB of RAM or more we put
      * the initrd at 128MB.
      */
-    *start = kernel_entry + MIN(mem_size / 2, 128 * MiB);
+    *start = (kernel_entry & 0xffffffff) + MIN(mem_size / 2, 128 * MiB);
 
     size = load_ramdisk(filename, *start, mem_size - *start);
     if (size == -1) {


Run command:

$ qemu/build/riscv32-softmmu/qemu-system-riscv32 -machine virt -kernel
mykernel.elf -nographic -initrd payload

Commit hash:

3a183e330dbd7dbcac3841737ac874979552cca2

** Affects: qemu
     Importance: Undecided
         Status: New

-- 
You received this bug notification because you are a member of qemu-
devel-ml, which is subscribed to QEMU.
https://bugs.launchpad.net/bugs/1814343

Title:
  Initrd not loaded on riscv32

Status in QEMU:
  New

Bug description:
  I attempted to run qemu with a ram disk. However, when reading the
  contents of the disk from within the VM I only get back zeros.

  I was able to trace the issue to a mismatch of expectations on line 93
  of hw/riscv/virt.c. Specifically, when running in 32-bit mode the
  value of kernel_entry is sign extended to 64-bits, but
  load_image_targphys expects the start address to not be sign extended.

  Straw man patch (works for 32-bit but would probably break 64-bit
  VMs?):

  diff --git a/hw/riscv/virt.c b/hw/riscv/virt.c
  index e7f0716fb6..32216f993c 100644
  --- a/hw/riscv/virt.c
  +++ b/hw/riscv/virt.c
  @@ -90,7 +90,7 @@ static hwaddr load_initrd(const char *filename, uint64_t mem_size,
        * halfway into RAM, and for boards with 256MB of RAM or more we put
        * the initrd at 128MB.
        */
  -    *start = kernel_entry + MIN(mem_size / 2, 128 * MiB);
  +    *start = (kernel_entry & 0xffffffff) + MIN(mem_size / 2, 128 * MiB);
   
       size = load_ramdisk(filename, *start, mem_size - *start);
       if (size == -1) {

  
  Run command:

  $ qemu/build/riscv32-softmmu/qemu-system-riscv32 -machine virt -kernel
  mykernel.elf -nographic -initrd payload

  Commit hash:

  3a183e330dbd7dbcac3841737ac874979552cca2

To manage notifications about this bug go to:
https://bugs.launchpad.net/qemu/+bug/1814343/+subscriptions
