Re: [PATCH 42/46] cxl/hdm: Commit decoder state to hardware

From: Dan Williams <dan.j.williams@intel.com>
To: Jonathan Cameron <Jonathan.Cameron@huawei.com>,
	Dan Williams <dan.j.williams@intel.com>
Cc: <linux-cxl@vger.kernel.org>, <nvdimm@lists.linux.dev>,
	<linux-pci@vger.kernel.org>, <patches@lists.linux.dev>,
	<hch@lst.de>
Subject: Re: [PATCH 42/46] cxl/hdm: Commit decoder state to hardware
Date: Sun, 10 Jul 2022 20:02:20 -0700	[thread overview]
Message-ID: <62cb92bc27397_3535162945b@dwillia2-xfh.notmuch> (raw)
In-Reply-To: <20220630180541.0000259c@Huawei.com>

Jonathan Cameron wrote:
> On Thu, 23 Jun 2022 21:19:46 -0700
> Dan Williams <dan.j.williams@intel.com> wrote:
> 
> > After all the soft validation of the region has completed, convey the
> > region configuration to hardware while being careful to commit decoders
> > in specification mandated order. In addition to programming the endpoint
> > decoder base-addres, intereleave ways and granularity, the switch
> > decoder target lists are also established.
> > 
> > While the kernel can enforce spec-mandated commit order, it can not
> > enforce spec-mandated reset order. For example, the kernel can't stop
> > someone from removing an endpoint device that is occupying decoderN in a
> > switch decoder where decoderN+1 is also committed. To reset decoderN,
> > decoderN+1 must be torn down first. That "tear down the world"
> > implementation is saved for a follow-on patch.
> > 
> > Callback operations are provided for the 'commit' and 'reset'
> > operations. While those callbacks may prove useful for CXL accelerators
> > (Type-2 devices with memory) the primary motivation is to enable a
> > simple way for cxl_test to intercept those operations.
> > 
> > Signed-off-by: Dan Williams <dan.j.williams@intel.com>
> 
> Trivial comments only in this one.
> 
> Jonathan
> 
> > ---
> >  Documentation/ABI/testing/sysfs-bus-cxl |  16 ++
> >  drivers/cxl/core/hdm.c                  | 218 ++++++++++++++++++++++++
> >  drivers/cxl/core/port.c                 |   1 +
> >  drivers/cxl/core/region.c               | 189 ++++++++++++++++++--
> >  drivers/cxl/cxl.h                       |  11 ++
> >  tools/testing/cxl/test/cxl.c            |  46 +++++
> >  6 files changed, 471 insertions(+), 10 deletions(-)
> > 
> 
> > diff --git a/drivers/cxl/core/hdm.c b/drivers/cxl/core/hdm.c
> > index 2ee62dde8b23..72f98f1a782c 100644
> > --- a/drivers/cxl/core/hdm.c
> > +++ b/drivers/cxl/core/hdm.c
> > @@ -129,6 +129,8 @@ struct cxl_hdm *devm_cxl_setup_hdm(struct cxl_port *port)
> >  		return ERR_PTR(-ENXIO);
> >  	}
> >  
> > +	dev_set_drvdata(&port->dev, cxlhdm);
> 
> Trivial, but dev == &port->dev I think so you might as well use dev.

Sure.

> This feels like a bit of a hack as it just so happens nothing else is
> in the port drvdata.  Maybe it's better to add a pointer from
> port to cxlhdm?

It's only valid while the port is attached to the cxl_port driver which
sets it apart from other port data.

> 
> > +
> >  	return cxlhdm;
> >  }
> >  EXPORT_SYMBOL_NS_GPL(devm_cxl_setup_hdm, CXL);
> > @@ -444,6 +446,213 @@ int cxl_dpa_alloc(struct cxl_endpoint_decoder *cxled, unsigned long long size)
> >  	return devm_add_action_or_reset(&port->dev, cxl_dpa_release, cxled);
> >  }
> >  
> 
> > +static int cxl_decoder_commit(struct cxl_decoder *cxld)
> > +{
> > +	struct cxl_port *port = to_cxl_port(cxld->dev.parent);
> > +	struct cxl_hdm *cxlhdm = dev_get_drvdata(&port->dev);
> > +	void __iomem *hdm = cxlhdm->regs.hdm_decoder;
> > +	int id = cxld->id, rc;
> > +	u64 base, size;
> > +	u32 ctrl;
> > +
> > +	if (cxld->flags & CXL_DECODER_F_ENABLE)
> > +		return 0;
> > +
> > +	if (port->commit_end + 1 != id) {
> > +		dev_dbg(&port->dev,
> > +			"%s: out of order commit, expected decoder%d.%d\n",
> > +			dev_name(&cxld->dev), port->id, port->commit_end + 1);
> > +		return -EBUSY;
> > +	}
> > +
> > +	down_read(&cxl_dpa_rwsem);
> > +	/* common decoder settings */
> > +	ctrl = readl(hdm + CXL_HDM_DECODER0_CTRL_OFFSET(cxld->id));
> > +	cxld_set_interleave(cxld, &ctrl);
> > +	cxld_set_type(cxld, &ctrl);
> > +	cxld_set_hpa(cxld, &base, &size);
> > +
> > +	writel(upper_32_bits(base), hdm + CXL_HDM_DECODER0_BASE_HIGH_OFFSET(id));
> > +	writel(lower_32_bits(base), hdm + CXL_HDM_DECODER0_BASE_LOW_OFFSET(id));
> > +	writel(upper_32_bits(size), hdm + CXL_HDM_DECODER0_SIZE_HIGH_OFFSET(id));
> > +	writel(lower_32_bits(size), hdm + CXL_HDM_DECODER0_SIZE_LOW_OFFSET(id));
> > +
> > +	if (is_switch_decoder(&cxld->dev)) {
> > +		struct cxl_switch_decoder *cxlsd =
> > +			to_cxl_switch_decoder(&cxld->dev);
> > +		void __iomem *tl_hi = hdm + CXL_HDM_DECODER0_TL_HIGH(id);
> > +		void __iomem *tl_lo = hdm + CXL_HDM_DECODER0_TL_LOW(id);
> > +		u64 targets;
> > +
> > +		rc = cxlsd_set_targets(cxlsd, &targets);
> > +		if (rc) {
> > +			dev_dbg(&port->dev, "%s: target configuration error\n",
> > +				dev_name(&cxld->dev));
> > +			goto err;
> > +		}
> > +
> > +		writel(upper_32_bits(targets), tl_hi);
> > +		writel(lower_32_bits(targets), tl_lo);
> > +	} else {
> > +		struct cxl_endpoint_decoder *cxled =
> > +			to_cxl_endpoint_decoder(&cxld->dev);
> > +		void __iomem *sk_hi = hdm + CXL_HDM_DECODER0_SKIP_HIGH(id);
> > +		void __iomem *sk_lo = hdm + CXL_HDM_DECODER0_SKIP_LOW(id);
> > +
> > +		writel(upper_32_bits(cxled->skip), sk_hi);
> > +		writel(lower_32_bits(cxled->skip), sk_lo);
> > +	}
> > +
> > +	writel(ctrl, hdm + CXL_HDM_DECODER0_CTRL_OFFSET(id));
> > +	up_read(&cxl_dpa_rwsem);
> > +
> > +	port->commit_end++;
> 
> Obviously doesn't matter as resetting on error, but
> feels like the increment of commit_end++ should only follow
> succesful commit / await_commit();

Then it would need a special ->reset() flavor to do everything but the
commit_end management. As long as cxl_region_rwsem is held over the
combination, nothing can sneak in and observe the intermediate state.

> > +	rc = cxld_await_commit(hdm, cxld->id);
> > +err:
> > +	if (rc) {
> > +		dev_dbg(&port->dev, "%s: error %d committing decoder\n",
> > +			dev_name(&cxld->dev), rc);
> > +		cxld->reset(cxld);
> > +		return rc;
> > +	}
> > +	cxld->flags |= CXL_DECODER_F_ENABLE;
> > +
> > +	return 0;
> > +}
> > +
> > +static int cxl_decoder_reset(struct cxl_decoder *cxld)
> > +{
> > +	struct cxl_port *port = to_cxl_port(cxld->dev.parent);
> > +	struct cxl_hdm *cxlhdm = dev_get_drvdata(&port->dev);
> > +	void __iomem *hdm = cxlhdm->regs.hdm_decoder;
> > +	int id = cxld->id;
> > +	u32 ctrl;
> > +
> > +	if ((cxld->flags & CXL_DECODER_F_ENABLE) ==  0)
> 
> extra space after ==

got it.

> 
> > +		return 0;
> > +
> 
> ...
> 
> 
> >  		
> > diff --git a/drivers/cxl/core/port.c b/drivers/cxl/core/port.c
> > index 7034300e72b2..eee1615d2319 100644
> > --- a/drivers/cxl/core/port.c
> > +++ b/drivers/cxl/core/port.c
> > @@ -630,6 +630,7 @@ static struct cxl_port *cxl_port_alloc(struct device *uport,
> >  	port->component_reg_phys = component_reg_phys;
> >  	ida_init(&port->decoder_ida);
> >  	port->dpa_end = -1;
> > +	port->commit_end = -1;
> >  	xa_init(&port->dports);
> >  	xa_init(&port->endpoints);
> >  	xa_init(&port->regions);
> > diff --git a/drivers/cxl/core/region.c b/drivers/cxl/core/region.c
> > index 071b8cafe2bb..b90160c4f975 100644
> > --- a/drivers/cxl/core/region.c
> > +++ b/drivers/cxl/core/region.c
> > @@ -112,6 +112,168 @@ static ssize_t uuid_store(struct device *dev, struct device_attribute *attr,
> >  }
> >  static DEVICE_ATTR_RW(uuid);
> 
> ...
> 
> 
> > +static int cxl_region_decode_reset(struct cxl_region *cxlr, int count)
> > +{
> > +	struct cxl_region_params *p = &cxlr->params;
> > +	int i;
> > +
> > +	for (i = count - 1; i >= 0; i--) {
> > +		struct cxl_endpoint_decoder *cxled = p->targets[i];
> > +		struct cxl_memdev *cxlmd = cxled_to_memdev(cxled);
> > +		struct cxl_port *iter = cxled_to_port(cxled);
> > +		struct cxl_ep *ep;
> > +		int rc;
> > +
> > +		while (!is_cxl_root(to_cxl_port(iter->dev.parent)))
> > +			iter = to_cxl_port(iter->dev.parent);
> > +
> > +		for (ep = cxl_ep_load(iter, cxlmd); iter;
> > +		     iter = ep->next, ep = cxl_ep_load(iter, cxlmd)) {
> > +			struct cxl_region_ref *cxl_rr;
> > +			struct cxl_decoder *cxld;
> > +
> > +			cxl_rr = cxl_rr_load(iter, cxlr);
> > +			cxld = cxl_rr->decoder;
> > +			rc = cxld->reset(cxld);
> > +			if (rc)
> > +				return rc;
> > +		}
> > +
> > +		rc = cxled->cxld.reset(&cxled->cxld);
> > +		if (rc)
> > +			return rc;
> > +	}
> > +
> > +	return 0;
> > +}
> > +
> > +static int cxl_region_decode_commit(struct cxl_region *cxlr)
> > +{
> > +	struct cxl_region_params *p = &cxlr->params;
> > +	int i, rc;
> > +
> > +	for (i = 0; i < p->nr_targets; i++) {
> > +		struct cxl_endpoint_decoder *cxled = p->targets[i];
> > +		struct cxl_memdev *cxlmd = cxled_to_memdev(cxled);
> > +		struct cxl_region_ref *cxl_rr;
> > +		struct cxl_decoder *cxld;
> > +		struct cxl_port *iter;
> > +		struct cxl_ep *ep;
> > +
> > +		/* commit bottom up */
> > +		for (iter = cxled_to_port(cxled); !is_cxl_root(iter);
> > +		     iter = to_cxl_port(iter->dev.parent)) {
> > +			cxl_rr = cxl_rr_load(iter, cxlr);
> > +			cxld = cxl_rr->decoder;
> > +			rc = cxld->commit(cxld);
> > +			if (rc)
> > +				break;
> > +		}
> > +
> > +		if (is_cxl_root(iter))
> > +			continue;
> > +
> > +		/* teardown top down */
> 
> Comment on why we are tearing down.  I guess because previous
> somehow didn't end up at the root?

Correct, one of those commits in the loop above failed causing it to
break out. Added a comment.

> 
> > +		for (ep = cxl_ep_load(iter, cxlmd); ep && iter;
> > +		     iter = ep->next, ep = cxl_ep_load(iter, cxlmd)) {
> > +			cxl_rr = cxl_rr_load(iter, cxlr);
> > +			cxld = cxl_rr->decoder;
> > +			cxld->reset(cxld);
> > +		}
> > +
> > +		cxled->cxld.reset(&cxled->cxld);
> > +		if (i == 0)
> > +			return rc;
> > +		break;
> > +	}
> > +
> > +	if (i >= p->nr_targets)
> > +		return 0;
> > +
> > +	/* undo the targets that were successfully committed */
> > +	cxl_region_decode_reset(cxlr, i);
> > +	return rc;
> > +}
> > +
> > +static ssize_t commit_store(struct device *dev, struct device_attribute *attr,
> > +			    const char *buf, size_t len)
> > +{
> > +	struct cxl_region *cxlr = to_cxl_region(dev);
> > +	struct cxl_region_params *p = &cxlr->params;
> > +	bool commit;
> > +	ssize_t rc;
> > +
> > +	rc = kstrtobool(buf, &commit);
> > +	if (rc)
> > +		return rc;
> > +
> > +	rc = down_write_killable(&cxl_region_rwsem);
> > +	if (rc)
> > +		return rc;
> > +
> > +	/* Already in the requested state? */
> > +	if (commit && p->state >= CXL_CONFIG_COMMIT)
> > +		goto out;
> > +	if (!commit && p->state < CXL_CONFIG_COMMIT)
> > +		goto out;
> > +
> > +	/* Not ready to commit? */
> > +	if (commit && p->state < CXL_CONFIG_ACTIVE) {
> > +		rc = -ENXIO;
> > +		goto out;
> > +	}
> > +
> > +	if (commit)
> > +		rc = cxl_region_decode_commit(cxlr);
> > +	else {
> > +		p->state = CXL_CONFIG_RESET_PENDING;
> > +		up_write(&cxl_region_rwsem);
> > +		device_release_driver(&cxlr->dev);
> > +		down_write(&cxl_region_rwsem);
> > +
> > +		if (p->state == CXL_CONFIG_RESET_PENDING)
> 
> What path results in that changing in last few lines?
> Perhaps a comment if there is something we need to protect against?

The lock needs to be dropped before calling device_release_driver(),
after reacquiring the lock need to revalidate that the reset is still
pending. Added a comment.

> 
> 
> > +			rc = cxl_region_decode_reset(cxlr, p->interleave_ways);
> > +	}
> > +
> > +	if (rc)
> > +		goto out;
> > +
> > +	if (commit)
> > +		p->state = CXL_CONFIG_COMMIT;
> > +	else if (p->state == CXL_CONFIG_RESET_PENDING)
> > +		p->state = CXL_CONFIG_ACTIVE;
> > +
> > +out:
> > +	up_write(&cxl_region_rwsem);
> > +
> > +	if (rc)
> > +		return rc;
> > +	return len;
> > +}
> 
> 
> ...
> 
> 
> > diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
> > index a93d7c4efd1a..fc14f6805f2c 100644
> > --- a/drivers/cxl/cxl.h
> > +++ b/drivers/cxl/cxl.h
> > @@ -54,6 +54,7 @@
> >  #define   CXL_HDM_DECODER0_CTRL_LOCK BIT(8)
> >  #define   CXL_HDM_DECODER0_CTRL_COMMIT BIT(9)
> >  #define   CXL_HDM_DECODER0_CTRL_COMMITTED BIT(10)
> > +#define   CXL_HDM_DECODER0_CTRL_COMMIT_ERROR BIT(11)
> >  #define   CXL_HDM_DECODER0_CTRL_TYPE BIT(12)
> >  #define CXL_HDM_DECODER0_TL_LOW(i) (0x20 * (i) + 0x24)
> >  #define CXL_HDM_DECODER0_TL_HIGH(i) (0x20 * (i) + 0x28)
> > @@ -257,6 +258,8 @@ enum cxl_decoder_type {
> >   * @target_type: accelerator vs expander (type2 vs type3) selector
> >   * @region: currently assigned region for this decoder
> >   * @flags: memory type capabilities and locking
> > + * @commit: device/decoder-type specific callback to commit settings to hw
> > + * @commit: device/decoder-type specific callback to reset hw settings
> 
> @reset

Yup.

> 
> >  */
> >  struct cxl_decoder {
> >  	struct device dev;
> > @@ -267,6 +270,8 @@ struct cxl_decoder {
> >  	enum cxl_decoder_type target_type;
> >  	struct cxl_region *region;
> >  	unsigned long flags;
> > +	int (*commit)(struct cxl_decoder *cxld);
> > +	int (*reset)(struct cxl_decoder *cxld);
> >  };
> >  
> 
> 
> > diff --git a/tools/testing/cxl/test/cxl.c b/tools/testing/cxl/test/cxl.c
> > index 51d517fa62ee..94653201631c 100644
> > --- a/tools/testing/cxl/test/cxl.c
> > +++ b/tools/testing/cxl/test/cxl.c
> > @@ -429,6 +429,50 @@ static int map_targets(struct device *dev, void *data)
> >  	return 0;
> >  }
> >  
> 
> ...
> 
> > +static int mock_decoder_reset(struct cxl_decoder *cxld)
> > +{
> > +	struct cxl_port *port = to_cxl_port(cxld->dev.parent);
> > +	int id = cxld->id;
> > +
> > +	if ((cxld->flags & CXL_DECODER_F_ENABLE) ==  0)
> 
> bonus space after ==

copy-pasta plus missed clang-format. Fixed.