Re: [PATCH 3/4] perf cs-etm: Track exception level

[Leo Yan <leo.yan@linaro.org>]

On Wed, May 24, 2023 at 02:19:57PM +0100, James Clark wrote:
> Currently we assume all trace belongs to the host machine so when
> the decoder should be looking at the guest kernel maps it can crash
> because it looks at the host ones instead.
> 
> Avoid one scenario (guest kernel running at EL1) by assigning the
> default guest machine to this trace. For userspace trace it's still not
> possible to determine guest vs host, but the PIDs should help in this
> case.
> 
> Signed-off-by: James Clark <james.clark@arm.com>
> ---
>  .../perf/util/cs-etm-decoder/cs-etm-decoder.c |  7 +-
>  tools/perf/util/cs-etm.c                      | 64 ++++++++++++++-----
>  tools/perf/util/cs-etm.h                      |  5 +-
>  3 files changed, 56 insertions(+), 20 deletions(-)
> 
> diff --git a/tools/perf/util/cs-etm-decoder/cs-etm-decoder.c b/tools/perf/util/cs-etm-decoder/cs-etm-decoder.c
> index 82a27ab90c8b..ac227cd03eb0 100644
> --- a/tools/perf/util/cs-etm-decoder/cs-etm-decoder.c
> +++ b/tools/perf/util/cs-etm-decoder/cs-etm-decoder.c
> @@ -573,12 +573,13 @@ cs_etm_decoder__set_tid(struct cs_etm_queue *etmq,
>  		break;
>  	}
>  
> +	if (cs_etm__etmq_set_tid_el(etmq, tid, trace_chan_id,
> +				 elem->context.exception_level))
> +		return OCSD_RESP_FATAL_SYS_ERR;
> +
>  	if (tid == -1)
>  		return OCSD_RESP_CONT;
>  
> -	if (cs_etm__etmq_set_tid(etmq, tid, trace_chan_id))
> -		return OCSD_RESP_FATAL_SYS_ERR;
> -
>  	/*
>  	 * A timestamp is generated after a PE_CONTEXT element so make sure
>  	 * to rely on that coming one.
> diff --git a/tools/perf/util/cs-etm.c b/tools/perf/util/cs-etm.c
> index a997fe79d458..b9ba19327f26 100644
> --- a/tools/perf/util/cs-etm.c
> +++ b/tools/perf/util/cs-etm.c
> @@ -14,7 +14,6 @@
>  #include <linux/types.h>
>  #include <linux/zalloc.h>
>  
> -#include <opencsd/ocsd_if_types.h>
>  #include <stdlib.h>
>  
>  #include "auxtrace.h"
> @@ -87,6 +86,8 @@ struct cs_etm_traceid_queue {
>  	union perf_event *event_buf;
>  	struct thread *thread;
>  	struct thread *prev_thread;
> +	ocsd_ex_level prev_el;
> +	ocsd_ex_level el;
>  	struct branch_stack *last_branch;
>  	struct branch_stack *last_branch_rb;
>  	struct cs_etm_packet *prev_packet;
> @@ -479,6 +480,7 @@ static int cs_etm__init_traceid_queue(struct cs_etm_queue *etmq,
>  
>  	queue = &etmq->etm->queues.queue_array[etmq->queue_nr];
>  	tidq->trace_chan_id = trace_chan_id;
> +	tidq->el = tidq->prev_el = ocsd_EL_unknown;
>  	tidq->thread = machine__findnew_thread(&etm->session->machines.host, -1,
>  					       queue->tid);
>  	tidq->prev_thread = machine__idle_thread(&etm->session->machines.host);
> @@ -618,6 +620,7 @@ static void cs_etm__packet_swap(struct cs_etm_auxtrace *etm,
>  		tmp = tidq->packet;
>  		tidq->packet = tidq->prev_packet;
>  		tidq->prev_packet = tmp;
> +		tidq->prev_el = tidq->el;
>  		thread__put(tidq->prev_thread);
>  		tidq->prev_thread = thread__get(tidq->thread);
>  	}
> @@ -879,11 +882,34 @@ static bool cs_etm__evsel_is_auxtrace(struct perf_session *session,
>  	return evsel->core.attr.type == aux->pmu_type;
>  }
>  
> -static u8 cs_etm__cpu_mode(struct cs_etm_queue *etmq, u64 address)
> +static struct machine *cs_etm__get_machine(struct cs_etm_auxtrace *etm,
> +					   ocsd_ex_level el)
>  {
> -	struct machine *machine;
> +	/*
> +	 * Not perfect, but assume anything in EL1 is the default guest, and
> +	 * everything else is the host. nHVE and pKVM may not work with this

s/nHVE/nVHE ?

And it's good to say "if any virtulisation (e.g. pKVM) based on nVHE may
not work with this ....".

> +	 * assumption. And distinguishing between guest and host userspaces
> +	 * isn't currently supported either. Neither is multiple guest support.
> +	 * All this does is reduce the likeliness of decode errors where we look
> +	 * into the host kernel maps when it should have been the guest maps.
> +	 */
> +	switch (el) {
> +	case ocsd_EL1:
> +		return machines__find_guest(&etm->session->machines,
> +					    DEFAULT_GUEST_KERNEL_ID);

I share the same concern with Mike that this will break perf for any
Armv8 CPUs with nVHE (like Cortex-CA53, etc).

You could see CoreSight has trace data "elem->context.vmid", when a
guest kernel runs in EL1, can we use "elem->context.vmid" as a guest
kernel ID and finally retrieve the corresponding machine pointer?

Thanks,
Leo

> +	case ocsd_EL3:
> +	case ocsd_EL2:
> +	case ocsd_EL0:
> +	case ocsd_EL_unknown:
> +	default:
> +		return &etm->session->machines.host;
> +	}
> +}
>  
> -	machine = &etmq->etm->session->machines.host;
> +static u8 cs_etm__cpu_mode(struct cs_etm_queue *etmq, u64 address,
> +			   ocsd_ex_level el)
> +{
> +	struct machine *machine = cs_etm__get_machine(etmq->etm, el);
>  
>  	if (address >= machine__kernel_start(machine)) {
>  		if (machine__is_host(machine))
> @@ -893,10 +919,14 @@ static u8 cs_etm__cpu_mode(struct cs_etm_queue *etmq, u64 address)
>  	} else {
>  		if (machine__is_host(machine))
>  			return PERF_RECORD_MISC_USER;
> -		else if (perf_guest)
> +		else {
> +			/*
> +			 * Can't really happen at the moment because
> +			 * cs_etm__get_machine() will always return
> +			 * machines.host for any non EL1 trace.
> +			 */
>  			return PERF_RECORD_MISC_GUEST_USER;
> -		else
> -			return PERF_RECORD_MISC_HYPERVISOR;
> +		}
>  	}
>  }
>  
> @@ -913,11 +943,12 @@ static u32 cs_etm__mem_access(struct cs_etm_queue *etmq, u8 trace_chan_id,
>  	if (!etmq)
>  		return 0;
>  
> -	cpumode = cs_etm__cpu_mode(etmq, address);
>  	tidq = cs_etm__etmq_get_traceid_queue(etmq, trace_chan_id);
>  	if (!tidq)
>  		return 0;
>  
> +	cpumode = cs_etm__cpu_mode(etmq, address, tidq->el);
> +
>  	if (!thread__find_map(tidq->thread, cpumode, address, &al))
>  		return 0;
>  
> @@ -1296,9 +1327,10 @@ cs_etm__get_trace(struct cs_etm_queue *etmq)
>  }
>  
>  static void cs_etm__set_thread(struct cs_etm_auxtrace *etm,
> -			       struct cs_etm_traceid_queue *tidq, pid_t tid)
> +			       struct cs_etm_traceid_queue *tidq, pid_t tid,
> +			       ocsd_ex_level el)
>  {
> -	struct machine *machine = &etm->session->machines.host;
> +	struct machine *machine = cs_etm__get_machine(etm, el);
>  
>  	if (tid != -1) {
>  		thread__zput(tidq->thread);
> @@ -1308,10 +1340,12 @@ static void cs_etm__set_thread(struct cs_etm_auxtrace *etm,
>  	/* Couldn't find a known thread */
>  	if (!tidq->thread)
>  		tidq->thread = machine__idle_thread(machine);
> +
> +	tidq->el = el;
>  }
>  
> -int cs_etm__etmq_set_tid(struct cs_etm_queue *etmq,
> -			 pid_t tid, u8 trace_chan_id)
> +int cs_etm__etmq_set_tid_el(struct cs_etm_queue *etmq, pid_t tid, u8 trace_chan_id,
> +			 ocsd_ex_level el)
>  {
>  	struct cs_etm_traceid_queue *tidq;
>  
> @@ -1319,7 +1353,7 @@ int cs_etm__etmq_set_tid(struct cs_etm_queue *etmq,
>  	if (!tidq)
>  		return -EINVAL;
>  
> -	cs_etm__set_thread(etmq->etm, tidq, tid);
> +	cs_etm__set_thread(etmq->etm, tidq, tid, el);
>  	return 0;
>  }
>  
> @@ -1389,7 +1423,7 @@ static int cs_etm__synth_instruction_sample(struct cs_etm_queue *etmq,
>  	struct perf_sample sample = {.ip = 0,};
>  
>  	event->sample.header.type = PERF_RECORD_SAMPLE;
> -	event->sample.header.misc = cs_etm__cpu_mode(etmq, addr);
> +	event->sample.header.misc = cs_etm__cpu_mode(etmq, addr, tidq->el);
>  	event->sample.header.size = sizeof(struct perf_event_header);
>  
>  	/* Set time field based on etm auxtrace config. */
> @@ -1448,7 +1482,7 @@ static int cs_etm__synth_branch_sample(struct cs_etm_queue *etmq,
>  	ip = cs_etm__last_executed_instr(tidq->prev_packet);
>  
>  	event->sample.header.type = PERF_RECORD_SAMPLE;
> -	event->sample.header.misc = cs_etm__cpu_mode(etmq, ip);
> +	event->sample.header.misc = cs_etm__cpu_mode(etmq, ip, tidq->prev_el);
>  	event->sample.header.size = sizeof(struct perf_event_header);
>  
>  	/* Set time field based on etm auxtrace config. */
> diff --git a/tools/perf/util/cs-etm.h b/tools/perf/util/cs-etm.h
> index 70cac0375b34..88e9b25a8a9f 100644
> --- a/tools/perf/util/cs-etm.h
> +++ b/tools/perf/util/cs-etm.h
> @@ -232,10 +232,11 @@ int cs_etm__process_auxtrace_info(union perf_event *event,
>  struct perf_event_attr *cs_etm_get_default_config(struct perf_pmu *pmu);
>  
>  #ifdef HAVE_CSTRACE_SUPPORT
> +#include <opencsd/ocsd_if_types.h>
>  int cs_etm__get_cpu(u8 trace_chan_id, int *cpu);
>  int cs_etm__get_pid_fmt(u8 trace_chan_id, u64 *pid_fmt);
> -int cs_etm__etmq_set_tid(struct cs_etm_queue *etmq,
> -			 pid_t tid, u8 trace_chan_id);
> +int cs_etm__etmq_set_tid_el(struct cs_etm_queue *etmq, pid_t tid, u8 trace_chan_id,
> +			 ocsd_ex_level el);
>  bool cs_etm__etmq_is_timeless(struct cs_etm_queue *etmq);
>  void cs_etm__etmq_set_traceid_queue_timestamp(struct cs_etm_queue *etmq,
>  					      u8 trace_chan_id);
> -- 
> 2.34.1
>