RE: [EXT] Re: [PATCH v6 2/7] coresight: tmc-etr: Add support to use reserved trace memory

From: Linu Cherian
Date: Fri Jan 12 2024 - 18:08:42 EST

Next message: Frederic Weisbecker: "Re: [PATCH v4 1/4] rcu: Reduce synchronize_rcu() latency"
Previous message: Matthew Wilcox: "Re: [RFC PATCH v1] mm/filemap: Allow arch to request folio size for exec memory"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Hi Suzuki,

> -----Original Message-----
> From: Suzuki K Poulose <suzuki.poulose@xxxxxxx>
> Sent: Friday, January 5, 2024 7:47 PM
> To: Linu Cherian <lcherian@xxxxxxxxxxx>; mike.leach@xxxxxxxxxx;
> james.clark@xxxxxxx; leo.yan@xxxxxxxxxx
> Cc: linux-arm-kernel@xxxxxxxxxxxxxxxxxxx; coresight@xxxxxxxxxxxxxxxx; linux-
> kernel@xxxxxxxxxxxxxxx; robh+dt@xxxxxxxxxx;
> krzysztof.kozlowski+dt@xxxxxxxxxx; conor+dt@xxxxxxxxxx;
> devicetree@xxxxxxxxxxxxxxx; Sunil Kovvuri Goutham
> <sgoutham@xxxxxxxxxxx>; George Cherian <gcherian@xxxxxxxxxxx>; Anil
> Kumar Reddy H <areddy3@xxxxxxxxxxx>
> Subject: [EXT] Re: [PATCH v6 2/7] coresight: tmc-etr: Add support to use
> reserved trace memory
>
> External Email
>
> ----------------------------------------------------------------------
> On 05/01/2024 05:58, Linu Cherian wrote:
> > Add support to use reserved memory for coresight ETR trace buffer.
> >
> > Introduce a new ETR buffer mode called ETR_MODE_RESRV, which becomes
> > available when ETR device tree node is supplied with a valid reserved
> > memory region.
> >
> > ETR_MODE_RESRV can be selected only by explicit user request.
> >
> > $ echo resrv
> >/sys/bus/coresight/devices/tmc_etr<N>/buf_mode_preferred
> >
> > Signed-off-by: Anil Kumar Reddy <areddy3@xxxxxxxxxxx>
> > Signed-off-by: Linu Cherian <lcherian@xxxxxxxxxxx>
> > ---
> > Changelog from v5:
> > * No changes
> >
> > .../hwtracing/coresight/coresight-tmc-core.c | 51 +++++++++++
> > .../hwtracing/coresight/coresight-tmc-etr.c | 87 ++++++++++++++++++-
> > drivers/hwtracing/coresight/coresight-tmc.h | 27 ++++++
> > 3 files changed, 164 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/hwtracing/coresight/coresight-tmc-core.c
> > b/drivers/hwtracing/coresight/coresight-tmc-core.c
> > index 7ec5365e2b64..224b969d7d90 100644
> > --- a/drivers/hwtracing/coresight/coresight-tmc-core.c
> > +++ b/drivers/hwtracing/coresight/coresight-tmc-core.c
> > @@ -22,6 +22,7 @@
> > #include <linux/spinlock.h>
> > #include <linux/pm_runtime.h>
> > #include <linux/of.h>
> > +#include <linux/of_address.h>
> > #include <linux/coresight.h>
> > #include <linux/amba/bus.h>
> >
> > @@ -370,6 +371,54 @@ static inline bool
> tmc_etr_has_non_secure_access(struct tmc_drvdata *drvdata)
> > return (auth & TMC_AUTH_NSID_MASK) == 0x3;
> > }
> >
> > +static struct device_node *tmc_get_region_byname(struct device_node
> *node,
> > + char *name)
> > +{
> > + int index;
> > +
> > + index = of_property_match_string(node, "memory-region-names",
> name);
> > + if (index < 0)
> > + return ERR_PTR(-ENODEV);
> > +
> > + return of_parse_phandle(node, "memory-region", index); }
> > +
> > +static void tmc_get_reserved_region(struct device *parent) {
> > + struct tmc_drvdata *drvdata = dev_get_drvdata(parent);
> > + struct device_node *node;
> > + struct resource res;
> > + int rc;
> > +
> > + node = tmc_get_region_byname(parent->of_node, "tracedata");
> > + if (IS_ERR_OR_NULL(node)) {
> > + dev_dbg(parent, "No reserved trace buffer specified\n");
> > + goto out;
>
> Given we don't return anything , the goto is pointless, you could return from
> here.
>

Ack.

> > + }
> > +
> > + rc = of_address_to_resource(node, 0, &res);
> > + of_node_put(node);
> > + if (rc || res.start == 0 || resource_size(&res) == 0) {
> > + dev_err(parent, "Reserved trace buffer memory is
> invalid\n");
> > + goto out;
>
> Same here.

Ack.

>
> > + }
> > +
> > + drvdata->crash_tbuf.vaddr = memremap(res.start,
> > + resource_size(&res),
> > + MEMREMAP_WC);
> > + if (IS_ERR_OR_NULL(drvdata->crash_tbuf.vaddr)) {
> > + dev_err(parent, "Reserved trace buffer mapping failed\n");
> > + rc = PTR_ERR(drvdata->crash_tbuf.vaddr);
>
> This is pointless, as we never use this.
>

Ack.

> > + goto out;
>
> return
>
> > + }
> > +
> > + drvdata->crash_tbuf.paddr = res.start;
> > + drvdata->crash_tbuf.size = resource_size(&res);
> > +
> > +out:
> > + return;
> > +}
> > +
> > /* Detect and initialise the capabilities of a TMC ETR */
> > static int tmc_etr_setup_caps(struct device *parent, u32 devid, void
> *dev_caps)
> > {
> > @@ -482,6 +531,8 @@ static int tmc_probe(struct amba_device *adev,
> const struct amba_id *id)
> > drvdata->size = readl_relaxed(drvdata->base + TMC_RSZ) * 4;
> > }
> >
> > + tmc_get_reserved_region(dev);
> > +
> > desc.dev = dev;
> >
> > switch (drvdata->config_type) {
> > diff --git a/drivers/hwtracing/coresight/coresight-tmc-etr.c
> > b/drivers/hwtracing/coresight/coresight-tmc-etr.c
> > index af02ba5d5f15..7fd8c4afef74 100644
> > --- a/drivers/hwtracing/coresight/coresight-tmc-etr.c
> > +++ b/drivers/hwtracing/coresight/coresight-tmc-etr.c
> > @@ -30,6 +30,7 @@ struct etr_buf_hw {
> > bool has_iommu;
> > bool has_etr_sg;
> > bool has_catu;
> > + bool has_resrv;
> > };
> >
> > /*
> > @@ -694,6 +695,74 @@ static const struct etr_buf_operations
> etr_flat_buf_ops = {
> > .get_data = tmc_etr_get_data_flat_buf,
> > };
> >
> > +/*
> > + * tmc_etr_alloc_resrv_buf: Allocate a contiguous DMA buffer from
> reserved region.
> > + */
> > +static int tmc_etr_alloc_resrv_buf(struct tmc_drvdata *drvdata,
> > + struct etr_buf *etr_buf, int node,
> > + void **pages)
> > +{
> > + struct etr_flat_buf *resrv_buf;
> > + struct device *real_dev = drvdata->csdev->dev.parent;
> > +
> > + /* We cannot reuse existing pages for resrv buf */
> > + if (pages)
> > + return -EINVAL;
> > +
> > + resrv_buf = kzalloc(sizeof(*resrv_buf), GFP_KERNEL);
> > + if (!resrv_buf)
> > + return -ENOMEM;
>
> Shouldn't we also ensure the etr_buf->size < crash_tbuf.size ? Or we could
> simply allocate the full reserve buf, as we can't use it for anything else. In any
> case, the buffer should be bounded by the reserved space.

This is taken care in tmc_alloc_etr_buf.ie. etr_buf_size is updated to crash_tbuf.size with ETR_MODE_RESRV buffer mode. Anyways, will move the size settings to here as suggested below.
>
> > +
> > + resrv_buf->daddr = dma_map_resource(real_dev, drvdata-
> >crash_tbuf.paddr,
> > + etr_buf->size, DMA_FROM_DEVICE,
> 0);
> > + if (dma_mapping_error(real_dev, resrv_buf->daddr)) {
> > + dev_err(real_dev, "failed to map source buffer address\n");
> > + kfree(resrv_buf);
> > + return -ENOMEM;
> > + }
> > +
> > + resrv_buf->vaddr = drvdata->crash_tbuf.vaddr;
> > + resrv_buf->size = etr_buf->size;
> > + resrv_buf->dev = &drvdata->csdev->dev;
> > + etr_buf->hwaddr = resrv_buf->daddr;
> > + etr_buf->mode = ETR_MODE_RESRV;
> > + etr_buf->private = resrv_buf;
> > + return 0;
> > +}
> > +
> > +static void tmc_etr_free_resrv_buf(struct etr_buf *etr_buf) {
> > + struct etr_flat_buf *resrv_buf = etr_buf->private;
> > +
> > + if (resrv_buf && resrv_buf->daddr) {
> > + struct device *real_dev = resrv_buf->dev->parent;
> > +
> > + dma_unmap_resource(real_dev, resrv_buf->daddr,
> > + resrv_buf->size, DMA_FROM_DEVICE, 0);
> > + }
> > + kfree(resrv_buf);
> > +}
> > +
> > +static void tmc_etr_sync_resrv_buf(struct etr_buf *etr_buf, u64 rrp,
> > +u64 rwp) {
> > + /*
> > + * Adjust the buffer to point to the beginning of the trace data
> > + * and update the available trace data.
> > + */
> > + etr_buf->offset = rrp - etr_buf->hwaddr;
> > + if (etr_buf->full)
> > + etr_buf->len = etr_buf->size;
> > + else
> > + etr_buf->len = rwp - rrp;
>
> Don't we need to sync the buffer for CPU via dma_sync_single_for_cpu() ?
> Otherwise, we could be reading stale data ?
>

From both CPU and DMA perspective we are mapping them as non cacheable and hence considered coherent always.
Ie. For CPU we use memremap with MEMREMAP_WC and for DMA we use dma_map_resource to map the trace buffer.

> > +}
> > +
> > +static const struct etr_buf_operations etr_resrv_buf_ops = {
> > + .alloc = tmc_etr_alloc_resrv_buf,
> > + .free = tmc_etr_free_resrv_buf,
> > + .sync = tmc_etr_sync_resrv_buf,
> > + .get_data = tmc_etr_get_data_flat_buf, };
> > +
> > /*
> > * tmc_etr_alloc_sg_buf: Allocate an SG buf @etr_buf. Setup the
> parameters
> > * appropriately.
> > @@ -800,6 +869,7 @@ static const struct etr_buf_operations
> *etr_buf_ops[] = {
> > [ETR_MODE_FLAT] = &etr_flat_buf_ops,
> > [ETR_MODE_ETR_SG] = &etr_sg_buf_ops,
> > [ETR_MODE_CATU] = NULL,
> > + [ETR_MODE_RESRV] = &etr_resrv_buf_ops
> > };
> >
> > void tmc_etr_set_catu_ops(const struct etr_buf_operations *catu) @@
> > -825,6 +895,7 @@ static inline int tmc_etr_mode_alloc_buf(int mode,
> > case ETR_MODE_FLAT:
> > case ETR_MODE_ETR_SG:
> > case ETR_MODE_CATU:
> > + case ETR_MODE_RESRV:
> > if (etr_buf_ops[mode] && etr_buf_ops[mode]->alloc)
> > rc = etr_buf_ops[mode]->alloc(drvdata, etr_buf,
> > node, pages);
> > @@ -843,6 +914,7 @@ static void get_etr_buf_hw(struct device *dev,
> struct etr_buf_hw *buf_hw)
> > buf_hw->has_iommu = iommu_get_domain_for_dev(dev->parent);
> > buf_hw->has_etr_sg = tmc_etr_has_cap(drvdata, TMC_ETR_SG);
> > buf_hw->has_catu = !!tmc_etr_get_catu_device(drvdata);
> > + buf_hw->has_resrv = is_tmc_reserved_region_valid(dev->parent);
> > }
> >
> > static bool etr_can_use_flat_mode(struct etr_buf_hw *buf_hw, ssize_t
> > etr_buf_size) @@ -874,13 +946,19 @@ static struct etr_buf
> *tmc_alloc_etr_buf(struct tmc_drvdata *drvdata,
> > if (!etr_buf)
> > return ERR_PTR(-ENOMEM);
> >
> > - etr_buf->size = size;
> > + /* Overiride the buffer size here for reserved mode */
>
> minor nit: "Override"
>

Ack.

> > + etr_buf->size = (drvdata->etr_mode == ETR_MODE_RESRV) ?
> > + drvdata->crash_tbuf.size : size;
>
> Isn't it better to do this from the actual code where we are successful in
> allocating the buffer ? That way, it works for the AUTO mode too ?
>

Ack. Will do that.

> >
> > /* If there is user directive for buffer mode, try that first */
> > if (drvdata->etr_mode != ETR_MODE_AUTO)
> > rc = tmc_etr_mode_alloc_buf(drvdata->etr_mode, drvdata,
> > etr_buf, node, pages);
> >
> > + /* Fallback mechanism is not valid for reserved mode */
> > + if (rc && (drvdata->etr_mode == ETR_MODE_RESRV))
> > + goto done;
> > +
> > /*
> > * If we have to use an existing list of pages, we cannot reliably
> > * use a contiguous DMA memory (even if we have an IOMMU).
> > Otherwise, @@ -902,6 +980,7 @@ static struct etr_buf
> *tmc_alloc_etr_buf(struct tmc_drvdata *drvdata,
> > if (rc && buf_hw.has_catu)
> > rc = tmc_etr_mode_alloc_buf(ETR_MODE_CATU, drvdata,
> > etr_buf, node, pages);
> > +done:
> > if (rc) {
> > kfree(etr_buf);
> > return ERR_PTR(rc);
> > @@ -1829,6 +1908,7 @@ static const char *const buf_modes_str[] = {
> > [ETR_MODE_FLAT] = "flat",
> > [ETR_MODE_ETR_SG] = "tmc-sg",
> > [ETR_MODE_CATU] = "catu",
> > + [ETR_MODE_RESRV] = "resrv",
> > [ETR_MODE_AUTO] = "auto",
> > };
> >
> > @@ -1847,6 +1927,9 @@ static ssize_t buf_modes_available_show(struct
> device *dev,
> > if (buf_hw.has_catu)
> > size += sysfs_emit_at(buf, size, "%s ",
> > buf_modes_str[ETR_MODE_CATU]);
> >
> > + if (buf_hw.has_resrv)
> > + size += sysfs_emit_at(buf, size, "%s ",
> > +buf_modes_str[ETR_MODE_RESRV]);
> > +
> > size += sysfs_emit_at(buf, size, "\n");
> > return size;
> > }
> > @@ -1874,6 +1957,8 @@ static ssize_t buf_mode_preferred_store(struct
> device *dev,
> > drvdata->etr_mode = ETR_MODE_ETR_SG;
> > else if (sysfs_streq(buf, buf_modes_str[ETR_MODE_CATU]) &&
> buf_hw.has_catu)
> > drvdata->etr_mode = ETR_MODE_CATU;
> > + else if (sysfs_streq(buf, buf_modes_str[ETR_MODE_RESRV]) &&
> buf_hw.has_resrv)
> > + drvdata->etr_mode = ETR_MODE_RESRV;
> > else if (sysfs_streq(buf, buf_modes_str[ETR_MODE_AUTO]))
> > drvdata->etr_mode = ETR_MODE_AUTO;
> > else
> > diff --git a/drivers/hwtracing/coresight/coresight-tmc.h
> > b/drivers/hwtracing/coresight/coresight-tmc.h
> > index 8dcb426ac3e7..10dba9f7d76a 100644
> > --- a/drivers/hwtracing/coresight/coresight-tmc.h
> > +++ b/drivers/hwtracing/coresight/coresight-tmc.h
> > @@ -135,6 +135,7 @@ enum etr_mode {
> > ETR_MODE_FLAT, /* Uses contiguous flat buffer */
> > ETR_MODE_ETR_SG, /* Uses in-built TMC ETR SG mechanism */
> > ETR_MODE_CATU, /* Use SG mechanism in CATU */
> > + ETR_MODE_RESRV, /* Use reserved region contiguous
> buffer */
> > ETR_MODE_AUTO, /* Use the default mechanism */
> > };
> >
> > @@ -164,6 +165,17 @@ struct etr_buf {
> > void *private;
> > };
> >
> > +/**
> > + * @paddr : Start address of reserved memory region.
> > + * @vaddr : Corresponding CPU virtual address.
> > + * @size : Size of reserved memory region.
> > + */
> > +struct tmc_resrv_buf {
> > + phys_addr_t paddr;
> > + void *vaddr;
> > + size_t size;
> > +};
> > +
> > /**
> > * struct tmc_drvdata - specifics associated to an TMC component
> > * @base: memory mapped base address for this component.
> > @@ -188,6 +200,10 @@ struct etr_buf {
> > * @idr_mutex: Access serialisation for idr.
> > * @sysfs_buf: SYSFS buffer for ETR.
> > * @perf_buf: PERF buffer for ETR.
> > + * @crash_tbuf: Used by ETR as hardware trace buffer and for trace data
> > + * retention (after crash) only when ETR_MODE_RESRV buffer
> > + * mode is enabled. Used by ETF for trace data retention
> > + * (after crash) by default.
>
> minor nit: Please align the comment like the other fields.
>

Ack.

Linu Cherian.

Next message: Frederic Weisbecker: "Re: [PATCH v4 1/4] rcu: Reduce synchronize_rcu() latency"
Previous message: Matthew Wilcox: "Re: [RFC PATCH v1] mm/filemap: Allow arch to request folio size for exec memory"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]