[PATCH] lib: sbi: Use SBI_SSE_EVENT_LOCAL_UNKNOWN_NMI with unknown interrupts

[PATCH] lib: sbi: Use SBI_SSE_EVENT_LOCAL_UNKNOWN_NMI with unknown interrupts

[Charlie Jenkins via B4 Relay]

From: Charlie Jenkins <thecharlesjenkins@gmail.com>

An mcause of 0 means "unknown cause" according to the riscv priviledged
spec. When this occurs, the interrupt should be forwarded to S-mode
software via SBI_SSE_EVENT_GLOBAL_UNKNOWN_NMI if SSE is enabled.

Signed-off-by: Charlie Jenkins <thecharlesjenkins@gmail.com>
---
This patch builds off of [1] to trigger the UNKNOWN_NMI SSE call. That
patch depends on the SBI spec patch [2].

[1] https://lore.kernel.org/opensbi/20251014055330.59506-1-weidong.wd@bytedance.com/
[2] https://github.com/riscv-non-isa/riscv-sbi-doc/pull/223
---
 include/sbi/riscv_encoding.h | 1 +
 lib/sbi/sbi_trap.c           | 4 ++++
 2 files changed, 5 insertions(+)

diff --git a/include/sbi/riscv_encoding.h b/include/sbi/riscv_encoding.h
index b5a4ce81..1c634a45 100644
--- a/include/sbi/riscv_encoding.h
+++ b/include/sbi/riscv_encoding.h
@@ -90,6 +90,7 @@
 
 #define MCAUSE_IRQ_MASK			(_UL(1) << (__riscv_xlen - 1))
 
+#define IRQ_UNKNOWN			0
 #define IRQ_S_SOFT			1
 #define IRQ_VS_SOFT			2
 #define IRQ_M_SOFT			3
diff --git a/lib/sbi/sbi_trap.c b/lib/sbi/sbi_trap.c
index f41db4d1..f6576bd9 100644
--- a/lib/sbi/sbi_trap.c
+++ b/lib/sbi/sbi_trap.c
@@ -234,6 +234,10 @@ int sbi_trap_redirect(struct sbi_trap_regs *regs,
 static int sbi_trap_nonaia_irq(unsigned long irq)
 {
 	switch (irq) {
+#if defined(CONFIG_SBI_ECALL_SSE)
+	case IRQ_UNKNOWN:
+		return sbi_sse_inject_event(SBI_SSE_EVENT_LOCAL_UNKNOWN_NMI);
+#endif
 	case IRQ_M_TIMER:
 		sbi_timer_process();
 		break;

---
base-commit: b27ecec76b8acfece9c28078d02cbc6bc762135c
change-id: 20260217-nmi-11064af8e897

- Charlie



-- 
opensbi mailing list
opensbi@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/opensbi

Re: [PATCH] lib: sbi: Use SBI_SSE_EVENT_LOCAL_UNKNOWN_NMI with unknown interrupts

[Himanshu Chauhan]

Hi,

On Wed, Feb 18, 2026 at 7:58 AM Charlie Jenkins via B4 Relay
<devnull+thecharlesjenkins.gmail.com@kernel.org> wrote:
>
> From: Charlie Jenkins <thecharlesjenkins@gmail.com>
>
> An mcause of 0 means "unknown cause" according to the riscv privileged
> spec. When this occurs, the interrupt should be forwarded to S-mode
> software via SBI_SSE_EVENT_GLOBAL_UNKNOWN_NMI if SSE is enabled.
>

This is an incorrect interpretation of the spec. The specification has
"reserved" the mcause exception code 0 of interrupts.
IMHO, a reserved interrupt, if occurs, shouldn't be forwarded to
supervisor software.

Regards
Himanshu

-- 
opensbi mailing list
opensbi@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/opensbi

Re: [PATCH] lib: sbi: Use SBI_SSE_EVENT_LOCAL_UNKNOWN_NMI with unknown interrupts

[Charlie Jenkins]

On Thu, Feb 19, 2026 at 12:23:23PM +0530, Himanshu Chauhan wrote:
> Hi,
> 
> On Wed, Feb 18, 2026 at 7:58 AM Charlie Jenkins via B4 Relay
> <devnull+thecharlesjenkins.gmail.com@kernel.org> wrote:
> >
> > From: Charlie Jenkins <thecharlesjenkins@gmail.com>
> >
> > An mcause of 0 means "unknown cause" according to the riscv privileged
> > spec. When this occurs, the interrupt should be forwarded to S-mode
> > software via SBI_SSE_EVENT_GLOBAL_UNKNOWN_NMI if SSE is enabled.
> >
> 
> This is an incorrect interpretation of the spec. The specification has
> "reserved" the mcause exception code 0 of interrupts.
> IMHO, a reserved interrupt, if occurs, shouldn't be forwarded to
> supervisor software.

Yeah that was a misinformed interpretation, thank you for your feedback.

- Charlie

> 
> Regards
> Himanshu
> 
> -- 
> opensbi mailing list
> opensbi@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/opensbi

-- 
opensbi mailing list
opensbi@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/opensbi