qemu-riscv
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-riscv] [Qemu-devel] [PATCH for 4.1] RISC-V: Ignore the S and U


From: Palmer Dabbelt
Subject: Re: [Qemu-riscv] [Qemu-devel] [PATCH for 4.1] RISC-V: Ignore the S and U extensions when formatting ISA strings
Date: Tue, 13 Aug 2019 15:54:46 -0700 (PDT)

On Wed, 07 Aug 2019 10:54:52 PDT (-0700), address@hidden wrote:
On Wed, Aug 7, 2019 at 8:00 AM Palmer Dabbelt <address@hidden> wrote:

The ISA strings we're providing from QEMU aren't actually legal RISC-V
ISA strings, as both the S and U extensions cannot exist as
single-letter extensions and must instead be multi-letter strings.
We're still using the ISA strings inside QEMU to track the availiable

s/availiable/available/g

extensions, so this patch just strips out the S and U extensions when
formatting ISA strings.

Atish and I were talking about this and we concluded that S and U
aren't extensions, but should be reported in the misa CSR.

Andrew agrees.



This boots Linux on top of 4.1-rc3, which no longer has the U extension
in /proc/cpuinfo.

Signed-off-by: Palmer Dabbelt <address@hidden>
---
This is another late one, but I'd like to target it for 4.1 as we're
providing illegal ISA strings and I don't want to bake that into a bunch
of other code.
---
 target/riscv/cpu.c | 17 ++++++++++++++++-
 1 file changed, 16 insertions(+), 1 deletion(-)

diff --git a/target/riscv/cpu.c b/target/riscv/cpu.c
index f8d07bd20ad7..4df14433d789 100644
--- a/target/riscv/cpu.c
+++ b/target/riscv/cpu.c
@@ -501,7 +501,22 @@ char *riscv_isa_string(RISCVCPU *cpu)
     char *p = isa_str + snprintf(isa_str, maxlen, "rv%d", TARGET_LONG_BITS);
     for (i = 0; i < sizeof(riscv_exts); i++) {
         if (cpu->env.misa & RV(riscv_exts[i])) {
-            *p++ = qemu_tolower(riscv_exts[i]);
+            char lower = qemu_tolower(riscv_exts[i]);
+            switch (lower) {
+            case 's':
+            case 'u':
+                /*
+                 * The 's' and 'u' extensions shouldn't be passed in the device
+                 * tree, but we still use them internally to track extension
+                 * sets.  Here we just explicitly remove them when formatting
+                 * an ISA string.

This should be updated to note mention 's' and 'u' as extensions, but
clarify that they are correctly include in the misa CSR.

I'll send a v2 that cleans up the wording on the comment and commit message.


Alistair

+                 */
+                break;
+
+            default:
+                *p++ = qemu_tolower(riscv_exts[i]);
+                break;
+            }
         }
     }
     *p = '\0';
--
2.21.0





reply via email to

[Prev in Thread] Current Thread [Next in Thread]