Re: [PATCH v3 6/9] spapr_pci: add spapr msi read method

[Peter Maydell]

On Tue, 30 Jun 2020 at 13:31, P J P <ppandit@redhat.com> wrote:
>
> From: Prasad J Pandit <pjp@fedoraproject.org>
>
> Add spapr msi mmio read method to avoid NULL pointer dereference
> issue.
>
> Reported-by: Lei Sun <slei.casper@gmail.com>
> Acked-by: David Gibson <david@gibson.dropbear.id.au>
> Reviewed-by: Li Qiang <liq3ea@gmail.com>
> Signed-off-by: Prasad J Pandit <pjp@fedoraproject.org>
> ---
>  hw/ppc/spapr_pci.c | 13 +++++++++++--
>  1 file changed, 11 insertions(+), 2 deletions(-)
>
> Update v3: Add Acked-by: ...
>   -> https://lists.gnu.org/archive/html/qemu-devel/2020-06/msg08054.html
>
> diff --git a/hw/ppc/spapr_pci.c b/hw/ppc/spapr_pci.c
> index 329002ac04..7033352834 100644
> --- a/hw/ppc/spapr_pci.c
> +++ b/hw/ppc/spapr_pci.c
> @@ -52,6 +52,7 @@
>  #include "sysemu/kvm.h"
>  #include "sysemu/hostmem.h"
>  #include "sysemu/numa.h"
> +#include "qemu/log.h"
>
>  /* Copied from the kernel arch/powerpc/platforms/pseries/msi.c */
>  #define RTAS_QUERY_FN           0
> @@ -738,6 +739,12 @@ static PCIINTxRoute spapr_route_intx_pin_to_irq(void 
> *opaque, int pin)
>      return route;
>  }
>
> +static uint64_t spapr_msi_read(void *opaque, hwaddr addr, unsigned size)
> +{
> +    qemu_log_mask(LOG_UNIMP, "%s not implemented\n", __func__);
> +    return 0;
> +}
> +
>  /*
>   * MSI/MSIX memory region implementation.
>   * The handler handles both MSI and MSIX.
> @@ -755,8 +762,10 @@ static void spapr_msi_write(void *opaque, hwaddr addr,
>  }
>
>  static const MemoryRegionOps spapr_msi_ops = {
> -    /* There is no .read as the read result is undefined by PCI spec */
> -    .read = NULL,
> +    /* .read result is undefined by PCI spec

QEMU multiline comments should have the '/*' on a line of its own.

> +     * define .read method to avoid assert failure in memory_region_init_io
> +     */

If this is undefined behaviour per the PCI spec then LOG_UNIMP
is the wrong thing -- this should either be LOG_GUEST_ERROR
(if the guest can do this or program the h/w to do this)
or assert() (if the only way this could happen would be a bug
in a QEMU model of a PCI device).

> +    .read = spapr_msi_read,
>      .write = spapr_msi_write,
>      .endianness = DEVICE_LITTLE_ENDIAN
>  };
> --
> 2.26.2

thanks
-- PMM