register: 0.964
files: 0.940
performance: 0.931
semantic: 0.928
ppc: 0.922
graphic: 0.905
architecture: 0.895
i386: 0.893
user-level: 0.893
x86: 0.889
peripherals: 0.887
mistranslation: 0.876
debug: 0.867
hypervisor: 0.851
permissions: 0.848
risc-v: 0.845
PID: 0.832
arm: 0.814
network: 0.810
KVM: 0.805
device: 0.799
socket: 0.793
assembly: 0.790
virtual: 0.763
VMM: 0.759
kernel: 0.757
TCG: 0.745
boot: 0.724
vnc: 0.694

cmsdk-apb-uart doesn't appear to clear interrupt flags

I have been writing a small operating system and using QEMU emulating the mps2-an385 board for some of my testing.

During development of the uart driver I observed some odd behaviour with the TX interrupt -- writing a '1' to bit 0 of the INTCLEAR register doesn't clear the TX interrupt flag, and the interrupt fires continuously.

It's possible that I have an error somewhere in my code, but after inspecting the QEMU source it does appear to be a QEMU bug. I applied the following patch and it solved my issue:

From 9875839c144fa60a3772f16ae44d32685f9328aa Mon Sep 17 00:00:00 2001
From: Patrick Oppenlander <email address hidden>
Date: Sat, 31 Mar 2018 15:10:28 +1100
Subject: [PATCH] hw/char/cmsdk-apb-uart: fix clearing of interrupt flags

---
 hw/char/cmsdk-apb-uart.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/hw/char/cmsdk-apb-uart.c b/hw/char/cmsdk-apb-uart.c
index 1ad1e14295..64991bd9d7 100644
--- a/hw/char/cmsdk-apb-uart.c
+++ b/hw/char/cmsdk-apb-uart.c
@@ -274,6 +274,7 @@ static void uart_write(void *opaque, hwaddr offset, uint64_t value,
          * is then reflected into the intstatus value by the update function).
          */
         s->state &= ~(value & (R_INTSTATUS_TXO_MASK | R_INTSTATUS_RXO_MASK));
+        s->intstatus &= ~(value & ~(R_INTSTATUS_TXO_MASK | R_INTSTATUS_RXO_MASK));
         cmsdk_apb_uart_update(s);
         break;
     case A_BAUDDIV:
-- 
2.16.2



Found in v2.12.0-rc1.

Thanks for the bug report; I've submitted this patch (which is similar to but not quite the same as your fix):
https://patchwork.ozlabs.org/patch/896715/

Hopefully this will get into 2.12, but we're quite close to release now so it will depend on whether we need to spin an extra release candidate for some other reason.


On Tue, Apr 10, 2018 at 11:45 PM, Peter Maydell
<email address hidden> wrote:
>
> Thanks for the bug report; I've submitted this patch (which is similar to but not quite the same as your fix):
> https://patchwork.ozlabs.org/patch/896715/
>
> Hopefully this will get into 2.12, but we're quite close to release now
> so it will depend on whether we need to spin an extra release candidate
> for some other reason.

Thanks for looking into it.

Patrick


https://git.qemu.org/?p=qemu.git;a=commitdiff;h=6670b494fdb23f74ecd9b