On 2025-01-03 20:58, Denis Mukhin via B4 Relay wrote:
From: Denis Mukhin <dmuk...@ford.com>
Introduce a driver framework to abstract UART emulators in the hypervisor.
That allows for architecture-independent handling of virtual UARTs from Xen
console driver and simplifies enabling new architecture-dependent UART
emulators.
The framework is built under CONFIG_HAS_VUART, which is automatically enabled
once the user selects a specific UART emulator.
All domains w/ enabled vUART will have VIRTDEV_UART bit set in
d->arch.emulation_flags.
Current implementation supports maximum of one vUART per domain, excluding
emulators for hardware domains.
Use domain_has_vuart() in Xen console driver code to check whether the
domain can own the physical console focus.
Note, arm/vuart.c emulator is not hooked to virtdev-uart framework because the
emulator is limited to the hardware domains only and was not designed to own
the physical console input. Updated arm/vuart.c APIs to have 'hwdom_' prefix
instead of generic 'domain_' to avoid possible confusion.
It might be good to renmae xen/arch/arm/vuart.c to
xen/arch/arm/hwdom-vuart.c and then use just the shorter vuart prefix
instead virtdev_uart.
Signed-off-by: Denis Mukhin <dmuk...@ford.com>
diff --git a/xen/drivers/Kconfig b/xen/drivers/Kconfig
index
20050e9bb8b32bd16c2da76c2c3e0f68dab89394..355719c3af67683c153a4f7a35dad4944992846e
100644
--- a/xen/drivers/Kconfig
+++ b/xen/drivers/Kconfig
@@ -19,4 +19,9 @@ config HAS_VPCI_GUEST_SUPPORT
bool
select HAS_VPCI
+config HAS_VUART
+ bool "UART emulation framework"
+ help
+ This selects UART emulation framework.
This can be hidden, so just:
config HAS_VUART
bool
+
endmenu
+int virtdev_uart_init(struct domain *d, struct virtdev_uart_params *params)
+{
+ int rc;
+
+ ASSERT(__start_virtdev_uart);
+
+ rc = __start_virtdev_uart->init(d, params);
+ if ( rc )
+ return rc;
+
+#if !defined(__i386__) && !defined(__x86_64__)
+ d->arch.emulation_flags |= VIRTDEV_UART;
+#endif
+
+ return 0;
+}
+
+void virtdev_uart_exit(struct domain *d)
+{
+ ASSERT(__start_virtdev_uart);
+
+ __start_virtdev_uart->exit(d);
+
+#if !defined(__i386__) && !defined(__x86_64__)
+ d->arch.emulation_flags &= ~VIRTDEV_UART;
+#endif
I think this is only called at domain teardown, so you don't need to
clear flags.
+}
+
+int virtdev_uart_putchar(struct domain *d, char c)
+{
+ ASSERT(__start_virtdev_uart);
+ ASSERT(d->arch.emulation_flags & VIRTDEV_UART);
+
+ return __start_virtdev_uart->putchar(d, c);
+}
+
+void virtdev_uart_dump(struct domain *d)
+{
+ ASSERT(__start_virtdev_uart);
+
+ __start_virtdev_uart->dump(d);
+}
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
diff --git a/xen/include/xen/virtdev-uart.h b/xen/include/xen/virtdev-uart.h
new file mode 100644
index
0000000000000000000000000000000000000000..fbe48e513996404d793d011747b3f40c236a6a57
--- /dev/null
+++ b/xen/include/xen/virtdev-uart.h
@@ -0,0 +1,72 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+#ifndef XEN__VIRTDEV_UART_H
+#define XEN__VIRTDEV_UART_H
+
+#include <public/xen.h>
+#include <public/event_channel.h>
+#include <xen/types.h>
+
+struct virtdev_uart_params {
+ domid_t console_domid;
+ gfn_t gfn;
+ evtchn_port_t evtchn;
+};
+
+struct virtdev_uart {
+ int (*putchar)(struct domain *d, char c);
+ int (*init)(struct domain *d, struct virtdev_uart_params *params);
+ void (*exit)(struct domain *d);
+ void (*dump)(struct domain *d);
+};
+
+#define VIRTDEV_UART_REGISTER(x) \
+ static const struct virtdev_uart *x##_entry \
+ __used_section(".data.virtdev.uart") = \
+ &(const struct virtdev_uart){ \
+ .init = x ## _init, \
+ .exit = x ## _exit, \
+ .dump = x ## _dump, \
+ .putchar = x ## _putchar, \
+ }
Are multiple types of uarts for a given architecture expected? If only
a single implemention is needed per-architecture, using defines or
wrappers seems simpler to me.
Regards,
Jason
