[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [RFC PATCH 1/1] QEMU plugin interface extension
From: |
Peter Maydell |
Subject: |
Re: [RFC PATCH 1/1] QEMU plugin interface extension |
Date: |
Thu, 26 Aug 2021 15:25:22 +0100 |
On Thu, 26 Aug 2021 at 15:12, Florian Hauschild
<florian.hauschild@fs.ei.tum.de> wrote:
> Form my observation all three insn_cb would see x0 == 2. They are
> executed at the end of a tb execution.
The documentation for the insn_exec_cb says the cb is called
every time an instruction is executed. That won't always be at
the end of a TB, will it ?
> Please correct me if i am wrong:
> When the TB is executed, first the TB cb is executed, then the various
> instruction cb. If you would like to see x0 in between instructions (e.g
> mov and first add), QEMU need to be in single step mode.
> The plugin infrastructure does have some sort of infrastructure to tell
> the tcg if the registers are read or written to, but does apparently not
> use it. The register values seem to be written back before the various
> cbs are called.
Any new plugin API for "read/write registers" needs to work correctly
at any point where it is valid for it to be called, whether QEMU
is in single-step mode or not.
I guess we'll wait for Alex to get back from holiday and have a look at this...
thanks
-- PMM
Re: [RFC PATCH 0/1] QEMU TCG plugin interface extensions, Alexandre IOOSS, 2021/08/21