As g2h() is doing the convertion from guest address space to host address space, I think the fix should be applied here, something like:

diff --git a/include/user/guest-host.h b/include/user/guest-host.h
index 8f7ef7589627..cdc7f0e01573 100644
--- a/include/user/guest-host.h
+++ b/include/user/guest-host.h
@@ -39,12 +39,12 @@ static inline vaddr cpu_untagged_addr(CPUState *cs, vaddr x)
 }
 
 /* All direct uses of g2h and h2g need to go away for usermode softmmu.  */
-static inline void *g2h_untagged(vaddr x)
+static inline void *g2h_untagged(abi_ulong x)
 {
     return (void *)((uintptr_t)(x) + guest_base);
 }
 
-static inline void *g2h(CPUState *cs, vaddr x)
+static inline void *g2h(CPUState *cs, abi_ulong x)
 {
     return g2h_untagged(cpu_untagged_addr(cs, x));
 }

@rth7680 ?

JIRA: https://issues.redhat.com/browse/RHEL-127007

build : https://brewweb.engineering.redhat.com/brew/taskinfo?taskID=70218166

testby me

Signed-off-by: Cindy Lu [email protected]

@lulu6 @mstredhat @pabeni

There was a conflict with MR 1926, I needed to update the patch series, could you review again?

Laurent Vivier (1050a075) at 04 Feb 07:36

... and 906 more commits

JIRA: https://issues.redhat.com/browse/RHEL-58639

build: https://brewweb.engineering.redhat.com/brew/taskinfo?taskID=69718730

tested by me

Signed-off-by: Cindy Lu [email protected]

Laurent Vivier (5ff3488d) at 30 Jan 06:59

Ok, I'm removing the latest patch from the series.

ok, I understand. I've updated the latest commit of the series to switch off the features with RHEL-10.0.0 (pc_q35_rhel_machine_10_0_0_options()) machine type and previous (using hw_compat_rhel_10_2[]).

Laurent Vivier (84f5922b) at 29 Jan 18:53

Laurent Vivier (23d67f04) at 29 Jan 18:52

Laurent Vivier (ec0857d6) at 29 Jan 18:45

But these features don't exists in previous RHEL releases. So do they need to be added to rhel_compat field?

My understanding is they have been added to hw_compat_10_1[] to be set to "off" for QEMU-10.1 and previous releases because in QEMU-10.2 they have been set to "on" by default.

But the feature are added in rhel-10.2 machine and they will be set to "on" by default, and no previous machine type has the features. So there is no value change for previous RHEL release.

Laurent Vivier (25d424a3) at 29 Jan 12:41

@peterx Could you check virtio-net: Advertise UDP tunnel GSO support by default , I'm not sure if the changes must be backported to hw_compat_rhel_10_2 [] as they were introduced in QEMU 10.1.

Laurent Vivier (25d424a3) at 29 Jan 12:39

Laurent Vivier (25d424a3) at 29 Jan 12:38

... and 18932 more commits

According to Michael in upstream Re: [PATCH net] virtio_net: fix alignment for virtio_net_hdr_v1_hash:

Btw, it looks like there's an uAPI change that may break builds of the userspace:

No, I think this has not been out long enough to matter. QEMU imports linux headers extremely quickly but it can adapt.

JIRA: https://issues.redhat.com/browse/RHEL-143785

Signed-off-by: Laurent Vivier [email protected]

Laurent Vivier (84f5922b) at 27 Jan 14:21

... and 23491 more commits

Laurent Vivier (3e400f5b) at 27 Jan 12:05

... and 3 more commits