Re: [PATCH v1] xen/riscv: add initialization support for virtual SBI UART (vSBI UART)

[Oleksii Kurochko]

On 5/15/25 12:08 PM, Jan Beulich wrote:

On 12.05.2025 17:55, Oleksii Kurochko wrote:

--- a/xen/arch/riscv/Makefile
+++ b/xen/arch/riscv/Makefile
@@ -1,5 +1,6 @@
  obj-y += aplic.o
  obj-y += cpufeature.o
+obj-y += dom0less-build.o

Arm uses

obj-$(CONFIG_DOM0LESS_BOOT) += dom0less-build.init.o

Why the two differences?

Missed that. I have to understand why Arm uses *.init.o, but likely I should do 
the same for RISC-V.

--- /dev/null
+++ b/xen/arch/riscv/dom0less-build.c
@@ -0,0 +1,30 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+
+#include <xen/bug.h>
+#include <xen/device_tree.h>
+#include <xen/errno.h>
+#include <xen/fdt-kernel.h>
+#include <xen/init.h>
+#include <xen/sched.h>
+
+#include <asm/vsbi-uart.h>
+
+int __init init_vuart(struct domain *d, struct kernel_info *kinfo,
+                      const struct dt_device_node *node)
+{
+    int rc = -EINVAL;
+
+    kinfo->arch.vsbi_uart = dt_property_read_bool(node, "vsbi_uart");
+
+    if ( kinfo->arch.vsbi_uart )
+    {
+        rc = domain_vsbi_uart_init(d, NULL);
+        if ( rc < 0 )
+            return rc;
+    }
+
+    if ( rc )
+        panic("%s: what a domain should use as an UART?\n", __func__);

Is this a reason to panic()? Isn't it possible for domains to be fine
without any UART?

Good point.
I think that it is possible to leave without UART. At this stage,
of development I need UART for debug, so automatically the panic() was added.
But I agree, it should be dropped.

+    domain_vsbi_uart_deinit(d);
+
+    return rc;
+}
+
+void domain_vsbi_uart_deinit(struct domain *d)
+{
+    struct vsbi_uart *vsbi_uart = &d->arch.vsbi_uart;
+
+    if ( vsbi_uart->backend_in_domain )
+        printk("%s: backed in a domain isn't supported\n", __func__);

Is this relevant in a de-init function?

Probably not, it was just added to not forget to update 
domain_vsbi_uart_deinit().

Thanks.

~ Oleksii