From joel at jms.id.au  Thu Aug  4 15:45:29 2022
From: joel at jms.id.au (Joel Stanley)
Date: Thu, 4 Aug 2022 05:45:29 +0000
Subject: [Pdbg] [PATCH 05/10] libpdbg: implement fused-core and
 lpar-per-thread thread status
In-Reply-To: <20220531091457.2208488-6-npiggin@gmail.com>
References: <20220531091457.2208488-1-npiggin@gmail.com> <20220531091457.2208488-6-npiggin@gmail.com>
Message-ID: <CACPK8XdazeymtQt+n_K_GDtyhe2=1Dws2sVEoagW3db1wfVxdQ@mail.gmail.com>

On Tue, 31 May 2022 at 09:15, Nicholas Piggin <npiggin at gmail.com> wrote:
>
> Like smt_state, these are really a per-core state but for now they are
> put in the thread_state. A future change will use fused-core, and the
> lpar mode may be generally useful to know in some situations.
>
> Signed-off-by: Nicholas Piggin <npiggin at gmail.com>
> ---
>  libpdbg/libpdbg.h | 2 ++
>  libpdbg/p10chip.c | 3 +++
>  libpdbg/p8chip.c  | 3 +++
>  libpdbg/p9chip.c  | 3 +++
>  4 files changed, 11 insertions(+)
>
> diff --git a/libpdbg/libpdbg.h b/libpdbg/libpdbg.h
> index 05e54f76..252b1b16 100644
> --- a/libpdbg/libpdbg.h
> +++ b/libpdbg/libpdbg.h
> @@ -1043,6 +1043,8 @@ enum pdbg_smt_state {
>  struct thread_state {
>         bool active;
>         bool quiesced;
> +       bool lpar_per_thread;
> +       bool fused_core_mode;
>         enum pdbg_sleep_state sleep_state;
>         enum pdbg_smt_state smt_state;
>  };
> diff --git a/libpdbg/p10chip.c b/libpdbg/p10chip.c
> index 70481f1f..632a1a9c 100644
> --- a/libpdbg/p10chip.c
> +++ b/libpdbg/p10chip.c
> @@ -107,6 +107,9 @@ struct thread_state p10_thread_state(struct thread *thread)
>         else
>                 thread_state.sleep_state = PDBG_THREAD_STATE_RUN;
>
> +       thread_state.lpar_per_thread = !(value & PPC_BIT(62));
> +       thread_state.fused_core_mode = !!(value & PPC_BIT(63));
> +
>         return thread_state;
>  }
>
> diff --git a/libpdbg/p8chip.c b/libpdbg/p8chip.c
> index 5b2a90a9..0592ca08 100644
> --- a/libpdbg/p8chip.c
> +++ b/libpdbg/p8chip.c
> @@ -247,6 +247,9 @@ static struct thread_state p8_thread_state(struct thread *thread)
>         /* Clear debug mode */
>         pib_write(&thread->target, RAS_MODE_REG, mode_reg);
>
> +       thread_state.lppar_per_thread = false;
> +       thread_state.fused_core_mode = false;

There were a few typos here that I fixed when applying:

-       thread_state.lppar_per_thread = false;
-       thread_state.fused_core_mode = false;
+       thread_status.lpar_per_thread = false;
+       thread_status.fused_core_mode = false;



> +
>         return thread_status;
>  }
>
> diff --git a/libpdbg/p9chip.c b/libpdbg/p9chip.c
> index 0040402c..036be521 100644
> --- a/libpdbg/p9chip.c
> +++ b/libpdbg/p9chip.c
> @@ -128,6 +128,9 @@ struct thread_state p9_thread_state(struct thread *thread)
>         else
>                 thread_state.sleep_state = PDBG_THREAD_STATE_RUN;
>
> +       thread_state.lpar_per_thread = !(value & PPC_BIT(62));
> +       thread_state.fused_core_mode = !!(value & PPC_BIT(63));
> +
>         return thread_state;
>  }
>
> --
> 2.35.1
>
> --
> Pdbg mailing list
> Pdbg at lists.ozlabs.org
> https://lists.ozlabs.org/listinfo/pdbg

From joel at jms.id.au  Thu Aug  4 15:46:36 2022
From: joel at jms.id.au (Joel Stanley)
Date: Thu, 4 Aug 2022 05:46:36 +0000
Subject: [Pdbg] [PATCH 00/10] pdbg fixes and improvements related to
 gdbserver
In-Reply-To: <20220531091457.2208488-1-npiggin@gmail.com>
References: <20220531091457.2208488-1-npiggin@gmail.com>
Message-ID: <CACPK8XdRrbPCgwqEBw22MK2aDAgQpxeQeEXZnaGcf4kNBa9_iw@mail.gmail.com>

On Tue, 31 May 2022 at 09:15, Nicholas Piggin <npiggin at gmail.com> wrote:
>
> Here's a bunch of things that came up when trying to work out what's
> wrong with using attn for breakpoints in gdbserver.
>
> After fixing some fused-core issues, turning off attn_handler service
> on the BMC seems to allow breakpoints to work reliably. Although
> executing attn can still cause the host skiboot PSI subsystem to
> print the "Masking unhandled LSI global_err" error. Not sure why the
> host gets these, I thought they only went to the BMC... That's
> another thing to work out and fix up but it doesn't seem to be
> harmful.
>
> Thanks,
> Nick
>
> Nicholas Piggin (10):
>   sbefifo: correct typo in thread target name
>   gdbserver: Fix attn enable HID change icache flushing sequence
>   gdbserver: use TRAP for step stop reason
>   p9chip: implement SMT state thread state
>   libpdbg: implement fused-core and lpar-per-thread thread status

>   gdbserver: Add a note about the attn_handler service on the BMC

I applied this set of patches. The next patch with its libpdbg change
required a bit more review, which I'll do later.

>   libpdbg: move core state into struct core
>   thread: print core state with threadstatus
>   libpdbg: add fused-core thread id
>   gdbserver: Fix POWER10 fused-core SPATTN register handling

>
>  README.md         | 15 +++++++--
>  libpdbg/chip.h    |  3 ++
>  libpdbg/hwunit.h  |  3 ++
>  libpdbg/libpdbg.h | 19 +++++++++++
>  libpdbg/p10chip.c | 77 +++++++++++++++++++++++++++++++++------------
>  libpdbg/p8chip.c  | 39 ++++++++++++++++++-----
>  libpdbg/p9chip.c  | 58 ++++++++++++++++++++++++++++++++--
>  libpdbg/sbefifo.c | 51 +++++++++++++++++++++++++++++-
>  libpdbg/thread.c  |  9 ++++++
>  src/pdbgproxy.c   | 80 +++++++++++++++++++++++++++++++++++++----------
>  src/thread.c      | 46 +++++++++++++++++++++++++--
>  11 files changed, 349 insertions(+), 51 deletions(-)
>
> --
> 2.35.1
>
> --
> Pdbg mailing list
> Pdbg at lists.ozlabs.org
> https://lists.ozlabs.org/listinfo/pdbg