[libvirt] [PATCH] Changed LPC slot from 1 to 31

[Roman Bogorodskiy]

  Ivan Mishonov wrote:

> ---
>  src/bhyve/bhyve_command.c | 2 +-
>  src/bhyve/bhyve_device.c  | 4 ++--
>  2 files changed, 3 insertions(+), 3 deletions(-)

Hi Ivan,

Thanks for your contribution!

Some general comments:

 - For quite some time already commits should have the Signed-off line
   with your name (you can use '-s' flag for git commit to get that
   automatically injected)
 - The good practice is to provide some more details in the commit
   message why you're doing this change. Also, commit titles are usually
   prefixed with a subsystem name this commit touches, for example:

    bhyve: Change LPC slot from 1 to 31

    Some guests (Windows, or whatever guests these are) require
    LPC on slot 31 to ... function properly (or whatever reason that
    is). So bind LPC to slot 31 instead of 1 which is used currently.

Code specific comments below.

> diff --git a/src/bhyve/bhyve_command.c b/src/bhyve/bhyve_command.c
> index 802997bd2d..e595b3d6c1 100644
> --- a/src/bhyve/bhyve_command.c
> +++ b/src/bhyve/bhyve_command.c
> @@ -329,7 +329,7 @@ static int
>  bhyveBuildLPCArgStr(const virDomainDef *def ATTRIBUTE_UNUSED,
>                      virCommandPtr cmd)
>  {
> -    virCommandAddArgList(cmd, "-s", "1,lpc", NULL);
> +    virCommandAddArgList(cmd, "-s", "31,lpc", NULL);
>      return 0;
>  }
>  
> diff --git a/src/bhyve/bhyve_device.c b/src/bhyve/bhyve_device.c
> index 03aa6c93bd..2ed3013552 100644
> --- a/src/bhyve/bhyve_device.c
> +++ b/src/bhyve/bhyve_device.c
> @@ -48,9 +48,9 @@ bhyveCollectPCIAddress(virDomainDefPtr def ATTRIBUTE_UNUSED,
>      if (addr->domain == 0 && addr->bus == 0) {
>          if (addr->slot == 0) {
>              return 0;
> -        } else if (addr->slot == 1) {
> +        } else if (addr->slot == 31) {
>              virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
> -                           _("PCI bus 0 slot 1 is reserved for the implicit "
> +                           _("PCI bus 0 slot 31 is reserved for the implicit "
>                               "LPC PCI-ISA bridge"));
>              return -1;
>          }


This patch makes unittest fail. If you run 'gmake check' you'll notice
one failing test:

...
PASS: xml2vmxtest
PASS: vmwarevertest
FAIL: bhyvexml2argvtest
PASS: bhyvexml2xmltest
PASS: bhyveargv2xmltest
...


To run this specific test and get some more verbose output you could
use:

VIR_TEST_DEBUG=3 ./tests/bhyvexml2argvtest

This is because many tests expect lpc to use slot 1. You might manually
update test data, or might regenerate it using the
VIR_TEST_REGENERATE_OUTPUT environment variable (more details here:
https://libvirt.org/hacking.html), but in this case it's necessary to
carefully verify the result to make sure unneeded changes don't squash
in.

One more thing: I think this patch misses the following chunk:

diff --git a/src/bhyve/bhyve_device.c b/src/bhyve/bhyve_device.c
index 2ed3013552..97688f5e70 100644
--- a/src/bhyve/bhyve_device.c
+++ b/src/bhyve/bhyve_device.c
@@ -97,7 +97,7 @@ bhyveAssignDevicePCISlots(virDomainDefPtr def,

     /* explicitly reserve slot 1 for LPC-ISA bridge */
     memset(&lpc_addr, 0, sizeof(lpc_addr));
-    lpc_addr.slot = 0x1;
+    lpc_addr.slot = 0x1f;

     if (virDomainPCIAddressReserveAddr(addrs, &lpc_addr,
                                        VIR_PCI_CONNECT_TYPE_PCI_DEVICE, 0) < 0) {

Because now it is slot 31 that needs to be reserved. That will likely
cause some more tests to fail because it changes the way address
auto-assignment works.

Feel free to reach me if you have any questions.

Thanks,
Roman
> -- 
> 2.18.0
> 
> --
> libvir-list mailing list
> libvir-list at redhat.com
> https://www.redhat.com/mailman/listinfo/libvir-list

Roman Bogorodskiy
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 455 bytes
Desc: not available
URL: <http://listman.redhat.com/archives/libvir-list/attachments/20180711/f8f3d0bd/attachment-0001.sig>