[PATCH v5 01/14] coresight: trace-id: Add API to dynamically assign Trace ID values
Mike Leach
mike.leach at linaro.org
Tue Nov 22 01:28:21 PST 2022
Hi Suzuki
On Tue, 8 Nov 2022 at 14:03, Suzuki K Poulose <suzuki.poulose at arm.com> wrote:
>
> Hi Mike
>
> On 01/11/2022 16:30, Mike Leach wrote:
> > The existing mechanism to assign Trace ID values to sources is limited
> > and does not scale for larger multicore / multi trace source systems.
> >
> > The API introduces functions that reserve IDs based on availabilty
> > represented by a coresight_trace_id_map structure. This records the
> > used and free IDs in a bitmap.
> >
> > CPU bound sources such as ETMs use the coresight_trace_id_get_cpu_id
> > coresight_trace_id_put_cpu_id pair of functions. The API will record
> > the ID associated with the CPU. This ensures that the same ID will be
> > re-used while perf events are active on the CPU. The put_cpu_id function
> > will pend release of the ID until all perf cs_etm sessions are complete.
> >
> > For backward compatibility the functions will attempt to use the same
> > CPU IDs as the legacy system would have used if these are still available.
> >
> > Non-cpu sources, such as the STM can use coresight_trace_id_get_system_id /
> > coresight_trace_id_put_system_id.
> >
> > Signed-off-by: Mike Leach <mike.leach at linaro.org>
>
> Thanks for the rework, patch looks good to me. Most of the comments are
> style related or nits.
>
> The important one is about the allocation of trace id for system
> components, where a legacy_cpu_id(cpu) could be allocated to a
> system component. It would be good to skip this (and we can do that
> with minimal changes, see below).
>
Agreed - good idea!
> There are many lines crossing the 80 characters width, please intend
> them if possible. Not strict about all of them, but some of the function
> definitions could be split into new lines.
>
checkpatch.pl now allows <=100 - I believe the rationale for this
increase was to reduce the number of unnecessary newlines.
Where I have gone over 80 I tend to do so to avoid unnecessarily
splitting lines which otherwise read perfectly well without the split
So my emacs is now adjusted to highlight > 100 in whitespace mode.
In the past when I have split a function definition that could fit
into 80 chars I have been told to unsplit it - so to split it when it
exceeds the old limit, but not the new is pretty counter-intuitive.
>
> > ---
> > drivers/hwtracing/coresight/Makefile | 2 +-
> > drivers/hwtracing/coresight/coresight-core.c | 4 +
> > .../hwtracing/coresight/coresight-trace-id.c | 225 ++++++++++++++++++
> > .../hwtracing/coresight/coresight-trace-id.h | 154 ++++++++++++
> > include/linux/coresight-pmu.h | 10 +
> > 5 files changed, 394 insertions(+), 1 deletion(-)
> > create mode 100644 drivers/hwtracing/coresight/coresight-trace-id.c
> > create mode 100644 drivers/hwtracing/coresight/coresight-trace-id.h
> >
> > diff --git a/drivers/hwtracing/coresight/Makefile b/drivers/hwtracing/coresight/Makefile
> > index b6c4a48140ec..329a0c704b87 100644
> > --- a/drivers/hwtracing/coresight/Makefile
> > +++ b/drivers/hwtracing/coresight/Makefile
> > @@ -6,7 +6,7 @@ obj-$(CONFIG_CORESIGHT) += coresight.o
> > coresight-y := coresight-core.o coresight-etm-perf.o coresight-platform.o \
> > coresight-sysfs.o coresight-syscfg.o coresight-config.o \
> > coresight-cfg-preload.o coresight-cfg-afdo.o \
> > - coresight-syscfg-configfs.o
> > + coresight-syscfg-configfs.o coresight-trace-id.o
> > obj-$(CONFIG_CORESIGHT_LINK_AND_SINK_TMC) += coresight-tmc.o
> > coresight-tmc-y := coresight-tmc-core.o coresight-tmc-etf.o \
> > coresight-tmc-etr.o
> > diff --git a/drivers/hwtracing/coresight/coresight-core.c b/drivers/hwtracing/coresight/coresight-core.c
> > index f3068175ca9d..554a18039e10 100644
> > --- a/drivers/hwtracing/coresight/coresight-core.c
> > +++ b/drivers/hwtracing/coresight/coresight-core.c
> > @@ -22,6 +22,7 @@
> > #include "coresight-etm-perf.h"
> > #include "coresight-priv.h"
> > #include "coresight-syscfg.h"
> > +#include "coresight-trace-id.h"
> >
> > static DEFINE_MUTEX(coresight_mutex);
> > static DEFINE_PER_CPU(struct coresight_device *, csdev_sink);
> > @@ -1804,6 +1805,9 @@ static int __init coresight_init(void)
> > if (ret)
> > goto exit_bus_unregister;
> >
> > + /* initialise the trace ID allocator */
> > + coresight_trace_id_init();
> > +
> > /* initialise the coresight syscfg API */
> > ret = cscfg_init();
> > if (!ret)
> > diff --git a/drivers/hwtracing/coresight/coresight-trace-id.c b/drivers/hwtracing/coresight/coresight-trace-id.c
> > new file mode 100644
> > index 000000000000..8e05a244c9d6
> > --- /dev/null
> > +++ b/drivers/hwtracing/coresight/coresight-trace-id.c
> > @@ -0,0 +1,225 @@
> > +// SPDX-License-Identifier: GPL-2.0
> > +/*
> > + * Copyright (c) 2022, Linaro Limited, All rights reserved.
> > + * Author: Mike Leach <mike.leach at linaro.org>
> > + */
> > +#include <linux/coresight-pmu.h>
> > +#include <linux/kernel.h>
> > +#include <linux/spinlock.h>
> > +#include <linux/types.h>
> > +
> > +#include "coresight-trace-id.h"
> > +
> > +/* default trace ID map. Used for systems that do not require per sink mappings */
> > +static struct coresight_trace_id_map id_map_default;
> > +
> > +/* maintain a record of the current mapping of cpu IDs and pending releases per cpu */
> > +static DEFINE_PER_CPU(atomic_t, cpu_id);
>
> We can statically initialise the cpu_id as below and get rid of the
> coresight_trace_id_init() altogether.
>
> static DEFINE_PER_CPU(atomic_t, cpu_id) = ATOMIC_INIT(0);
>
> > +static DECLARE_BITMAP(cpu_id_release_pending, NR_CPUS);
>
> If you do respin this, please consider switching to a cpumask_t.
> Apologies, should have said that this last time. Then, we could
> do :
>
> for_each_cpu(cpu, &cpui_id_release_pending) {
> ..
> }
>
agreed and implemented with the cpumask_... fns where required.
> > +
> > +/* perf session active counter */
> > +static atomic_t perf_cs_etm_session_active = ATOMIC_INIT(0);
> > +
> > +/* lock to protect id_map and cpu data */
> > +static DEFINE_SPINLOCK(id_map_lock);
> > +
> > +/*
> > + * allocate new ID and set in use
> > + * if @preferred_id is a valid id then try to use that value if available.
> > + */
> > +static int coresight_trace_id_alloc_new_id(struct coresight_trace_id_map *id_map,
> > + int preferred_id)
> > +{
> > + int id;
> > +
> > + /* for backwards compatibility reasons, cpu Ids may have a preferred value */
> > + if (IS_VALID_ID(preferred_id) && !test_bit(preferred_id, id_map->used_ids))
>
> minor nit: line width exceeds 80.
>
> if (IS_VALID_ID(preferred_id) &&
> !test_bit(preferred_id, id_map->used_ids)) {
> > + id = preferred_id;
> } else {
>
> > + else {
> > + /* skip reserved bit 0, look from bit 1 to CORESIGHT_TRACE_ID_RES_TOP */
> > + id = find_next_zero_bit(id_map->used_ids, 1, CORESIGHT_TRACE_ID_RES_TOP);
> > + if (id >= CORESIGHT_TRACE_ID_RES_TOP)
> > + return -EINVAL;
> > + }
> > +
> > + /* mark as used */
> > + set_bit(id, id_map->used_ids);
> > + return id;
> > +}
> > +
> > +static void coresight_trace_id_free(int id, struct coresight_trace_id_map *id_map)
> > +{
> > + if (WARN(!IS_VALID_ID(id), "%s: Invalid Trace ID %d\n", __func__, id))
> > + return;
>
> minor nit: __func__ is not be need as the WARN would already give you
> the file+line number of the warning and also a stack trace. This could
> simply be:
>
> if (!WARN_ON(!IS_VALID_ID(id))
> clear_bit(id, id_map->used_ids);
>
Done
>
> > + if (WARN(!test_bit(id, id_map->used_ids),
> > + "%s: Freeing unused ID %d\n", __func__, id))
> > + return;
> > + clear_bit(id, id_map->used_ids);
> > +}
> > +
> > +static void coresight_trace_id_set_pend_rel(int id, struct coresight_trace_id_map *id_map)
>
> minor nit: exceeds 80
>
> > +{
> > + if (WARN(!IS_VALID_ID(id), "%s: Invalid Trace ID %d\n", __func__, id))
> > + return;
> > + set_bit(id, id_map->pend_rel_ids);
> > +}
> > +
> > +/* release all pending IDs for all current maps & clear CPU associations */
>
> Minor nit: This function operates on the "default id map" and as such
> it may be a good idea to call that out in the comment or even in the
> function name to make it explicit. Every other function accepts an
> id_map.
>
> > +static void coresight_trace_id_release_all_pending(void)
> > +{
> > + struct coresight_trace_id_map *id_map = &id_map_default;
> > + unsigned long flags;
> > + int cpu, bit;
> > +
> > + spin_lock_irqsave(&id_map_lock, flags);
> > + for_each_set_bit(bit, id_map->pend_rel_ids, CORESIGHT_TRACE_ID_RES_TOP) {
> > + clear_bit(bit, id_map->used_ids);
> > + clear_bit(bit, id_map->pend_rel_ids);
> > + }
> > + for_each_set_bit(cpu, cpu_id_release_pending, NR_CPUS) {
> > + atomic_set(&per_cpu(cpu_id, cpu), 0);
> > + clear_bit(cpu, cpu_id_release_pending);
> > + }
> > + spin_unlock_irqrestore(&id_map_lock, flags);
> > +}
> > +
> > +static int coresight_trace_id_map_get_cpu_id(int cpu, struct coresight_trace_id_map *id_map)
>
> minor nit: Please split the line
splitting after .. int cpu still leaves the end of the line > 80 chars!
>
> > +{
> > + unsigned long flags;
> > + int id;
> > +
> > + spin_lock_irqsave(&id_map_lock, flags);
>
> > +
> > + /* check for existing allocation for this CPU */
> > + id = atomic_read(&per_cpu(cpu_id, cpu));
> > + if (id)
>
> Please could we use coresight_trace_id_read_cpu_id() ?
> i.e Could we keep the code abstracted as much as possible, than open
> coding at different places ?
>
> i.e.,
>
> id = coresight_trace_id_read_cpu_id()
>
> we could move the definition around. That way, it is easier to fix
> potential changes in the future.
>
agreed - used an internal fn for the read then used it at these
locations + api fn at the end of the file
> > + goto get_cpu_id_clr_pend;
> > +
> > + /*
> > + * Find a new ID.
> > + *
> > + * Use legacy values where possible in the dynamic trace ID allocator to
> > + * allow older tools to continue working if they are not upgraded at the same
> > + * time as the kernel drivers.
> > + *
> > + * If the generated legacy ID is invalid, or not available then the next
> > + * available dynamic ID will be used.
> > + */
>
> Thanks, this will help the legacy perf tools. One question here though.
> In order to make sure we get the legacy allocated to CPUs, do we need
> to tweak the "get_system_id" to "prefer" "odd" traceIDs ? e.g., I see
> that the TPDA requests a trace-id at "probe" time and keeps that
> forever. If the TPDA gets to the trace-id first and gets the trace-id
> of CPU0, we may break the legacy id scheme here. We could live with it,
> but having this sort of a scheme may help to avoid the situation.
>
> e.g., we could use
>
>
> coresight_trace_id_alloc_preferred(id_map,
> CORESIGHT_LEGACY_CPU_TRACE_ID(cpu));
>
> where:
>
> coresight_trace_id_alloc_mask(id_map, mask, match)
> {
> for_each_clear_bit(id, id_map->used_ids) {
> if ((id & mask) == match)
> allocate_id and return id;
> }
> }
>
> coresight_trace_id_alloc_new_id(id_map)
> {
> return coresight_trace_id_alloc_mask(id_map, 0, 0);
> }
>
> coresight_trace_id_alloc_preferred(id_map, preferred)
> {
> if (prefered is free) {
> allocate preferred and mark as used;
> } else {
> coresight_trace_id_alloc_new_id(id_map);
> }
> }
>
> /* Prefer Odd traceid values for system components) */
> coresight_trace_id_alloc_mask(id_map, 0x1, 0x1);
>
> Thoughts ?
>
fixed up to prefer odd IDs for system components.
> > + id = coresight_trace_id_alloc_new_id(id_map, CORESIGHT_LEGACY_CPU_TRACE_ID(cpu));
> > + if (!IS_VALID_ID(id))
> > + goto get_cpu_id_out_unlock;
> > +
> > + /* allocate the new id to the cpu */
> > + atomic_set(&per_cpu(cpu_id, cpu), id);
> > +
> > +get_cpu_id_clr_pend:
> > + /* we are (re)using this ID - so ensure it is not marked for release */
> > + clear_bit(cpu, cpu_id_release_pending);
> > + clear_bit(id, id_map->pend_rel_ids);
> > +
> > +get_cpu_id_out_unlock:
> > + spin_unlock_irqrestore(&id_map_lock, flags);
> > +
> > + return id;
> > +}
> > +
> > +static void coresight_trace_id_map_put_cpu_id(int cpu, struct coresight_trace_id_map *id_map)
> > +{
> > + unsigned long flags;
> > + int id;
> > +
> > + /* check for existing allocation for this CPU */
> > + id = atomic_read(&per_cpu(cpu_id, cpu));
>
> same as above, coresight_trace_id_read_cpu_id();
>
> > + if (!id)
> > + return;
> > +
> > + spin_lock_irqsave(&id_map_lock, flags);
> > +
> > + if (atomic_read(&perf_cs_etm_session_active)) {
> > + /* set release at pending if perf still active */
> > + coresight_trace_id_set_pend_rel(id, id_map);
> > + set_bit(cpu, cpu_id_release_pending);
> > + } else {
> > + /* otherwise clear id */
> > + coresight_trace_id_free(id, id_map);
> > + atomic_set(&per_cpu(cpu_id, cpu), 0);
> > + }
> > +
> > + spin_unlock_irqrestore(&id_map_lock, flags);
> > +}
> > +
> > +static int coresight_trace_id_map_get_system_id(struct coresight_trace_id_map *id_map)
> > +{
> > + unsigned long flags;
> > + int id;
> > +
> > + spin_lock_irqsave(&id_map_lock, flags);
> > + id = coresight_trace_id_alloc_new_id(id_map, 0);
> > + spin_unlock_irqrestore(&id_map_lock, flags);
> > +
> > + return id;
> > +}
> > +
> > +static void coresight_trace_id_map_put_system_id(struct coresight_trace_id_map *id_map, int id)
> > +{
> > + unsigned long flags;
> > +
> > + spin_lock_irqsave(&id_map_lock, flags);
> > + coresight_trace_id_free(id, id_map);
> > + spin_unlock_irqrestore(&id_map_lock, flags);
> > +}
> > +
> > +/* API functions */
> > +
> > +int coresight_trace_id_get_cpu_id(int cpu)
> > +{
> > + return coresight_trace_id_map_get_cpu_id(cpu, &id_map_default);
> > +}
> > +EXPORT_SYMBOL_GPL(coresight_trace_id_get_cpu_id);
> > +
> > +void coresight_trace_id_put_cpu_id(int cpu)
> > +{
> > + coresight_trace_id_map_put_cpu_id(cpu, &id_map_default);
> > +}
> > +EXPORT_SYMBOL_GPL(coresight_trace_id_put_cpu_id);
> > +
> > +int coresight_trace_id_read_cpu_id(int cpu)
> > +{
> > + return atomic_read(&per_cpu(cpu_id, cpu));
> > +}
> > +EXPORT_SYMBOL_GPL(coresight_trace_id_read_cpu_id);
> > +
> > +int coresight_trace_id_get_system_id(void)
> > +{
> > + return coresight_trace_id_map_get_system_id(&id_map_default);
> > +}
> > +EXPORT_SYMBOL_GPL(coresight_trace_id_get_system_id);
> > +
> > +void coresight_trace_id_put_system_id(int id)
> > +{
> > + coresight_trace_id_map_put_system_id(&id_map_default, id);
> > +}
> > +EXPORT_SYMBOL_GPL(coresight_trace_id_put_system_id);
> > +
> > +void coresight_trace_id_perf_start(void)
> > +{
> > + atomic_inc(&perf_cs_etm_session_active);
> > +}
> > +EXPORT_SYMBOL_GPL(coresight_trace_id_perf_start);
> > +
> > +void coresight_trace_id_perf_stop(void)
> > +{
> > + if (!atomic_dec_return(&perf_cs_etm_session_active))
> > + coresight_trace_id_release_all_pending();
> > +}
> > +EXPORT_SYMBOL_GPL(coresight_trace_id_perf_stop);
> > +
>
>
> > +void coresight_trace_id_init(void)
> > +{
> > + int cpu;
> > +
> > + /* initialise the atomic trace ID values */
> > + for_each_possible_cpu(cpu)
> > + atomic_set(&per_cpu(cpu_id, cpu), 0);
> > +}
> > +EXPORT_SYMBOL_GPL(coresight_trace_id_init);
>
> As mentioned above, we could remove this.
>
done
> > diff --git a/drivers/hwtracing/coresight/coresight-trace-id.h b/drivers/hwtracing/coresight/coresight-trace-id.h
> > new file mode 100644
> > index 000000000000..1d27977346b3
> > --- /dev/null
> > +++ b/drivers/hwtracing/coresight/coresight-trace-id.h
> > @@ -0,0 +1,154 @@
> > +/* SPDX-License-Identifier: GPL-2.0 */
> > +/*
> > + * Copyright(C) 2022 Linaro Limited. All rights reserved.
> > + * Author: Mike Leach <mike.leach at linaro.org>
> > + */
> > +
> > +#ifndef _CORESIGHT_TRACE_ID_H
> > +#define _CORESIGHT_TRACE_ID_H
> > +
> > +/*
> > + * Coresight trace ID allocation API
> > + *
> > + * With multi cpu systems, and more additional trace sources a scalable
> > + * trace ID reservation system is required.
> > + *
> > + * The system will allocate Ids on a demand basis, and allow them to be
> > + * released when done.
> > + *
> > + * In order to ensure that a consistent cpu / ID matching is maintained
> > + * throughout a perf cs_etm event session - a session in progress flag will
> > + * be maintained, and released IDs not cleared until the perf session is
> > + * complete. This allows the same CPU to be re-allocated its prior ID.
> > + *
> > + *
> > + * Trace ID maps will be created and initialised to prevent architecturally
> > + * reserved IDs from being allocated.
> > + *
> > + * API permits multiple maps to be maintained - for large systems where
> > + * different sets of cpus trace into different independent sinks.
> > + */
> > +
> > +#include <linux/bitops.h>
> > +#include <linux/types.h>
> > +
> > +
> > +/* architecturally we have 128 IDs some of which are reserved */
> > +#define CORESIGHT_TRACE_IDS_MAX 128
> > +
> > +/* ID 0 is reserved */
> > +#define CORESIGHT_TRACE_ID_RES_0 0
> > +
> > +/* ID 0x70 onwards are reserved */
> > +#define CORESIGHT_TRACE_ID_RES_TOP 0x70
>
> super minor nit: Could we make this CORESIGHT_TRACE_ID_MAX ?
>
>
actually prefer it as it as as this value is not in fact the MAX ID -
either usable (0x6f is the highest allocatable) or theoretical (0x7f -
which is not permitted anyway)
The value also acts as the number of bits to scan for when searching
the bitmaps for usable ids, in the C code.
>
> Rest looks fine to me.
>
> Suzuki
also changed IS_VALID_ID to something more relevant to Coresight.
Thanks for the review
Mike
--
Mike Leach
Principal Engineer, ARM Ltd.
Manchester Design Centre. UK
More information about the linux-arm-kernel
mailing list