[PATCH v8 0/2] cxl: Export cxl1.1 device link status to sysfs

[PATCH v8 0/2] cxl: Export cxl1.1 device link status to sysfs

[Kobayashi,Daisuke]

Export cxl1.1 device link status register value to pci device sysfs.

CXL devices are extensions of PCIe. Therefore, from CXL2.0 onwards,
the link status can be output in the same way as traditional PCIe.
However, unlike devices from CXL2.0 onwards, CXL1.1 requires a
different method to obtain the link status from traditional PCIe.
This is because the link status of the CXL1.1 device is not mapped
in the configuration space (as per cxl3.0 specification 8.1).
Instead, the configuration space containing the link status is mapped
to the memory mapped register region (as per cxl3.0 specification 8.2,
Table 8-18). Therefore, the current lspci has a problem where it does
not display the link status of the CXL1.1 device. 
Solve these issues with sysfs attributes to export the status
registers hidden in the RCRB.

The procedure is as follows:
First, obtain the RCRB address within the cxl driver, then access 
the configuration space. Next, output the link status information from
the configuration space to sysfs. Ultimately, the expectation is that
this sysfs file will be consumed by PCI user tools to utilize link status
information.


Changes
v1[1] -> v2:
- Modified to perform rcrb access within the CXL driver.
- Added new attributes to the sysfs of the PCI device.
- Output the link status information to the sysfs of the PCI device.
- Retrieve information from sysfs as the source when displaying information in lspci.

v2[2] -> v3:
- Fix unnecessary initialization and wrong types (Bjohn).
- Create a helper function for getting a PCIe capability offset (Bjohn).
- Move platform-specific implementation to the lib directory in pciutils (Martin).

v3[3] -> v4:
- RCRB register values are read once and cached.
- Added a new attribute to the sysfs of the PCI device.
- Separate lspci implementation from this patch.

v4[4] -> v5:
- Use macros for bitwise operations
- Fix RCRB access to use cxl_memdev

v5[5] -> v6:
- Add and use masks for RCRB register values

v6[6] -> v7:
- Fix comments on white space inline

v7[7] -> v8:
- Change the cache value to offset
- Access memory map area in rcd_*_show() functions

[1]
https://lore.kernel.org/linux-cxl/20231220050738.178481-1-kobayashi.da-06@fujitsu.com/
[2]
https://lore.kernel.org/linux-cxl/20240227083313.87699-1-kobayashi.da-06@fujitsu.com/
[3]
https://lore.kernel.org/linux-cxl/20240312080559.14904-1-kobayashi.da-06@fujitsu.com/
[4]
https://lore.kernel.org/linux-cxl/20240409073528.13214-1-kobayashi.da-06@fujitsu.com/
[5]
https://lore.kernel.org/linux-cxl/20240412070715.16160-1-kobayashi.da-06@fujitsu.com/
[6]
https://lore.kernel.org/linux-cxl/20240424050102.26788-1-kobayashi.da-06@fujitsu.com/
[7]
https://lore.kernel.org/linux-cxl/20240510073710.98953-1-kobayashi.da-06@fujitsu.com/

Kobayashi,Daisuke (2):
  cxl/core/regs: Add rcd_regs initialization at __rcrb_to_component()
  cxl/pci: Add sysfs attribute for CXL 1.1 device link status

 drivers/cxl/core/core.h |  4 ++
 drivers/cxl/core/regs.c | 13 ++++++
 drivers/cxl/cxl.h       |  1 +
 drivers/cxl/pci.c       | 91 +++++++++++++++++++++++++++++++++++++++++
 4 files changed, 109 insertions(+)

-- 
2.44.0

[PATCH v8 1/2] cxl/core/regs: Add rcd_pcie_cap initialization at __rcrb_to_component()

[Kobayashi,Daisuke]

Add rcd_pcie_cap and its initialization at __rcrb_to_component() to cache
the offset of cxl1.1 device link status information. By caching it, avoid 
the walking memory map area to find the offset when output the register value.

Signed-off-by: "Kobayashi,Daisuke" <kobayashi.da-06@fujitsu.com>

---
 drivers/cxl/core/core.h |  4 ++++
 drivers/cxl/core/regs.c | 13 +++++++++++++
 drivers/cxl/cxl.h       |  1 +
 3 files changed, 18 insertions(+)

diff --git a/drivers/cxl/core/core.h b/drivers/cxl/core/core.h
index 3b64fb1b9ed0..42e3483b4a14 100644
--- a/drivers/cxl/core/core.h
+++ b/drivers/cxl/core/core.h
@@ -75,6 +75,10 @@ resource_size_t __rcrb_to_component(struct device *dev,
 				    enum cxl_rcrb which);
 u16 cxl_rcrb_to_aer(struct device *dev, resource_size_t rcrb);
 
+#define PCI_RCRB_CAP_LIST_ID_MASK	GENMASK(7, 0)
+#define PCI_RCRB_CAP_HDR_ID_MASK	GENMASK(7, 0)
+#define PCI_RCRB_CAP_HDR_NEXT_MASK	GENMASK(15, 8)
+
 extern struct rw_semaphore cxl_dpa_rwsem;
 extern struct rw_semaphore cxl_region_rwsem;
 
diff --git a/drivers/cxl/core/regs.c b/drivers/cxl/core/regs.c
index 372786f80955..3e64b49f74ed 100644
--- a/drivers/cxl/core/regs.c
+++ b/drivers/cxl/core/regs.c
@@ -514,6 +514,8 @@ resource_size_t __rcrb_to_component(struct device *dev, struct cxl_rcrb_info *ri
 	u32 bar0, bar1;
 	u16 cmd;
 	u32 id;
+	u16 offset;
+	u32 cap_hdr;
 
 	if (which == CXL_RCRB_UPSTREAM)
 		rcrb += SZ_4K;
@@ -537,6 +539,17 @@ resource_size_t __rcrb_to_component(struct device *dev, struct cxl_rcrb_info *ri
 	cmd = readw(addr + PCI_COMMAND);
 	bar0 = readl(addr + PCI_BASE_ADDRESS_0);
 	bar1 = readl(addr + PCI_BASE_ADDRESS_1);
+	offset = FIELD_GET(PCI_RCRB_CAP_LIST_ID_MASK, readw(addr + PCI_CAPABILITY_LIST));
+	cap_hdr = readl(addr + offset);
+	while ((FIELD_GET(PCI_RCRB_CAP_HDR_ID_MASK, cap_hdr)) != PCI_CAP_ID_EXP) {
+		offset = FIELD_GET(PCI_RCRB_CAP_HDR_NEXT_MASK, cap_hdr);
+		if (offset == 0 || offset > SZ_4K)
+			break;
+		cap_hdr = readl(addr + offset);
+	}
+	if (offset)
+		ri->rcd_pcie_cap = offset;
+
 	iounmap(addr);
 	release_mem_region(rcrb, SZ_4K);
 
diff --git a/drivers/cxl/cxl.h b/drivers/cxl/cxl.h
index 003feebab79b..40396e2c4bae 100644
--- a/drivers/cxl/cxl.h
+++ b/drivers/cxl/cxl.h
@@ -646,6 +646,7 @@ cxl_find_dport_by_dev(struct cxl_port *port, const struct device *dport_dev)
 
 struct cxl_rcrb_info {
 	resource_size_t base;
+	u16 rcd_pcie_cap;
 	u16 aer_cap;
 };
 
-- 
2.44.0

[PATCH v8 2/2] cxl/pci: Add sysfs attribute for CXL 1.1 device link status

[Kobayashi,Daisuke]

Add sysfs attribute for CXL 1.1 device link status to the cxl pci device.

In CXL1.1, the link status of the device is included in the RCRB mapped to
the memory mapped register area. Critically, that arrangement makes the
link status and control registers invisible to existing PCI user tooling.

Export those registers via sysfs with the expectation that PCI user
tooling will alternatively look for these sysfs files when attempting to
access to these CXL 1.1 endpoints registers.

Signed-off-by: "Kobayashi,Daisuke" <kobayashi.da-06@fujitsu.com>
---
 drivers/cxl/pci.c | 91 +++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 91 insertions(+)

diff --git a/drivers/cxl/pci.c b/drivers/cxl/pci.c
index 2ff361e756d6..e157959dffc2 100644
--- a/drivers/cxl/pci.c
+++ b/drivers/cxl/pci.c
@@ -786,6 +786,96 @@ static int cxl_event_config(struct pci_host_bridge *host_bridge,
 	return 0;
 }
 
+static u32 get_rcd_pcie_caps(struct device *dev, u16 offset)
+{
+	struct cxl_dev_state *cxlds = dev_get_drvdata(dev);
+	struct cxl_memdev *cxlmd = cxlds->cxlmd;
+	struct device *endpoint_parent;
+	struct cxl_dport *dport;
+	struct cxl_port *port;
+	resource_size_t rcrb;
+	void __iomem *addr;
+	u32 ret;
+
+	port = cxl_mem_find_port(cxlmd, &dport);
+	if (!port)
+		return 0;
+
+	endpoint_parent = port->uport_dev;
+	if (!endpoint_parent)
+		return 0;
+
+	guard(device)(endpoint_parent);
+	if (!endpoint_parent->driver)
+		return 0;
+
+	rcrb = dport->rcrb.base;
+	if (!request_mem_region(rcrb, SZ_4K, "CXL RCRB"))
+		return 0;
+	addr = ioremap(rcrb, SZ_4K);
+	if (!addr) {
+		dev_err(dev, "Failed to map region %pr\n", addr);
+		release_mem_region(rcrb, SZ_4K);
+		return 0;
+	}
+
+	ret = readl(addr + dport->rcrb.rcd_pcie_cap + offset);
+	release_mem_region(rcrb, SZ_4K);
+	return ret;
+}
+
+static ssize_t rcd_link_cap_show(struct device *dev,
+				   struct device_attribute *attr, char *buf)
+{
+	u32 linkcap = get_rcd_pcie_caps(dev, PCI_EXP_LNKCAP);
+
+	return sysfs_emit(buf, "%x\n", linkcap);
+}
+static DEVICE_ATTR_RO(rcd_link_cap);
+
+static ssize_t rcd_link_ctrl_show(struct device *dev,
+				   struct device_attribute *attr, char *buf)
+{
+	u16 linkctl = get_rcd_pcie_caps(dev, PCI_EXP_LNKCTL);
+
+	return sysfs_emit(buf, "%x\n", linkctl);
+}
+static DEVICE_ATTR_RO(rcd_link_ctrl);
+
+static ssize_t rcd_link_status_show(struct device *dev,
+				   struct device_attribute *attr, char *buf)
+{
+	u16 linksta = get_rcd_pcie_caps(dev, PCI_EXP_LNKSTA);
+
+	return sysfs_emit(buf, "%x\n", linksta);
+}
+static DEVICE_ATTR_RO(rcd_link_status);
+
+static struct attribute *cxl_rcd_attrs[] = {
+		&dev_attr_rcd_link_cap.attr,
+		&dev_attr_rcd_link_ctrl.attr,
+		&dev_attr_rcd_link_status.attr,
+		NULL
+};
+
+static umode_t cxl_rcd_visible(struct kobject *kobj,
+					  struct attribute *a, int n)
+{
+	struct device *dev = kobj_to_dev(kobj);
+	struct pci_dev *pdev = to_pci_dev(dev);
+
+	if (is_cxl_restricted(pdev))
+		return a->mode;
+
+	return 0;
+}
+
+static struct attribute_group cxl_rcd_group = {
+		.attrs = cxl_rcd_attrs,
+		.is_visible = cxl_rcd_visible,
+};
+__ATTRIBUTE_GROUPS(cxl_rcd);
+
 static int cxl_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 {
 	struct pci_host_bridge *host_bridge = pci_find_host_bridge(pdev->bus);
@@ -969,6 +1059,7 @@ static struct pci_driver cxl_pci_driver = {
 	.id_table		= cxl_mem_pci_tbl,
 	.probe			= cxl_pci_probe,
 	.err_handler		= &cxl_error_handlers,
+	.dev_groups		= cxl_rcd_groups,
 	.driver	= {
 		.probe_type	= PROBE_PREFER_ASYNCHRONOUS,
 	},
-- 
2.44.0

Re: [PATCH v8 2/2] cxl/pci: Add sysfs attribute for CXL 1.1 device link status

[Jonathan Cameron]

On Thu,  6 Jun 2024 16:48:14 +0900
"Kobayashi,Daisuke" <kobayashi.da-06@fujitsu.com> wrote:

> Add sysfs attribute for CXL 1.1 device link status to the cxl pci device.
> 
> In CXL1.1, the link status of the device is included in the RCRB mapped to
> the memory mapped register area. Critically, that arrangement makes the
> link status and control registers invisible to existing PCI user tooling.
> 
> Export those registers via sysfs with the expectation that PCI user
> tooling will alternatively look for these sysfs files when attempting to
> access to these CXL 1.1 endpoints registers.
> 
> Signed-off-by: "Kobayashi,Daisuke" <kobayashi.da-06@fujitsu.com>
> ---
>  drivers/cxl/pci.c | 91 +++++++++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 91 insertions(+)
> 
> diff --git a/drivers/cxl/pci.c b/drivers/cxl/pci.c
> index 2ff361e756d6..e157959dffc2 100644
> --- a/drivers/cxl/pci.c
> +++ b/drivers/cxl/pci.c
> @@ -786,6 +786,96 @@ static int cxl_event_config(struct pci_host_bridge *host_bridge,
>  	return 0;
>  }
>  
> +static u32 get_rcd_pcie_caps(struct device *dev, u16 offset)
> +{
> +	struct cxl_dev_state *cxlds = dev_get_drvdata(dev);
> +	struct cxl_memdev *cxlmd = cxlds->cxlmd;
> +	struct device *endpoint_parent;
> +	struct cxl_dport *dport;
> +	struct cxl_port *port;
> +	resource_size_t rcrb;
> +	void __iomem *addr;
> +	u32 ret;
> +
> +	port = cxl_mem_find_port(cxlmd, &dport);
> +	if (!port)
> +		return 0;
> +
> +	endpoint_parent = port->uport_dev;
> +	if (!endpoint_parent)
> +		return 0;
> +
> +	guard(device)(endpoint_parent);
> +	if (!endpoint_parent->driver)
> +		return 0;
> +
> +	rcrb = dport->rcrb.base;
> +	if (!request_mem_region(rcrb, SZ_4K, "CXL RCRB"))

I was assuming we would find the address of the capability only once
and map the relevant cap in for all the time the driver is
bound, as is done for aer_cap.

I'd be worried that this might race with other accesses to rcrb
in the future.

I suspect best option is to keep the capability always mapped
in similar fashion to cxl_dport_map_rch_aer() does it for
the aer registers.



> +		return 0;
> +	addr = ioremap(rcrb, SZ_4K);
> +	if (!addr) {
> +		dev_err(dev, "Failed to map region %pr\n", addr);
> +		release_mem_region(rcrb, SZ_4K);
> +		return 0;
> +	}
> +
> +	ret = readl(addr + dport->rcrb.rcd_pcie_cap + offset);
> +	release_mem_region(rcrb, SZ_4K);
> +	return ret;
> +}
> +
> +static ssize_t rcd_link_cap_show(struct device *dev,
> +				   struct device_attribute *attr, char *buf)
> +{
> +	u32 linkcap = get_rcd_pcie_caps(dev, PCI_EXP_LNKCAP);
> +
> +	return sysfs_emit(buf, "%x\n", linkcap);
> +}
> +static DEVICE_ATTR_RO(rcd_link_cap);
> +
> +static ssize_t rcd_link_ctrl_show(struct device *dev,
> +				   struct device_attribute *attr, char *buf)
> +{
> +	u16 linkctl = get_rcd_pcie_caps(dev, PCI_EXP_LNKCTL);
> +
> +	return sysfs_emit(buf, "%x\n", linkctl);
> +}
> +static DEVICE_ATTR_RO(rcd_link_ctrl);
> +
> +static ssize_t rcd_link_status_show(struct device *dev,
> +				   struct device_attribute *attr, char *buf)
> +{
> +	u16 linksta = get_rcd_pcie_caps(dev, PCI_EXP_LNKSTA);
> +
> +	return sysfs_emit(buf, "%x\n", linksta);

Local variable doesn't add to readability as the PCI_EXP_LNKSTA
makes it pretty obvious what is being queried.

	return sysfs_emit(buf, "%x\n",
			  get_rcd_pcie_caps(dev, PCI_EXP_LNKSTA));

> +}
> +static DEVICE_ATTR_RO(rcd_link_status);
> +
> +static struct attribute *cxl_rcd_attrs[] = {
> +		&dev_attr_rcd_link_cap.attr,
> +		&dev_attr_rcd_link_ctrl.attr,

Indent one less tab.

> +		&dev_attr_rcd_link_status.attr,
> +		NULL
> +};

RE: [PATCH v8 2/2] cxl/pci: Add sysfs attribute for CXL 1.1 device link status

[Daisuke Kobayashi (Fujitsu)]

Jonathan Cameron wrote:
> On Thu,  6 Jun 2024 16:48:14 +0900
> "Kobayashi,Daisuke" <kobayashi.da-06@fujitsu.com> wrote:
> 
> > Add sysfs attribute for CXL 1.1 device link status to the cxl pci device.
> >
> > In CXL1.1, the link status of the device is included in the RCRB
> > mapped to the memory mapped register area. Critically, that
> > arrangement makes the link status and control registers invisible to existing
> PCI user tooling.
> >
> > Export those registers via sysfs with the expectation that PCI user
> > tooling will alternatively look for these sysfs files when attempting
> > to access to these CXL 1.1 endpoints registers.
> >
> > Signed-off-by: "Kobayashi,Daisuke" <kobayashi.da-06@fujitsu.com>
> > ---
> >  drivers/cxl/pci.c | 91
> > +++++++++++++++++++++++++++++++++++++++++++++++
> >  1 file changed, 91 insertions(+)
> >
> > diff --git a/drivers/cxl/pci.c b/drivers/cxl/pci.c index
> > 2ff361e756d6..e157959dffc2 100644
> > --- a/drivers/cxl/pci.c
> > +++ b/drivers/cxl/pci.c
> > @@ -786,6 +786,96 @@ static int cxl_event_config(struct pci_host_bridge
> *host_bridge,
> >  	return 0;
> >  }
> >
> > +static u32 get_rcd_pcie_caps(struct device *dev, u16 offset) {
> > +	struct cxl_dev_state *cxlds = dev_get_drvdata(dev);
> > +	struct cxl_memdev *cxlmd = cxlds->cxlmd;
> > +	struct device *endpoint_parent;
> > +	struct cxl_dport *dport;
> > +	struct cxl_port *port;
> > +	resource_size_t rcrb;
> > +	void __iomem *addr;
> > +	u32 ret;
> > +
> > +	port = cxl_mem_find_port(cxlmd, &dport);
> > +	if (!port)
> > +		return 0;
> > +
> > +	endpoint_parent = port->uport_dev;
> > +	if (!endpoint_parent)
> > +		return 0;
> > +
> > +	guard(device)(endpoint_parent);
> > +	if (!endpoint_parent->driver)
> > +		return 0;
> > +
> > +	rcrb = dport->rcrb.base;
> > +	if (!request_mem_region(rcrb, SZ_4K, "CXL RCRB"))
> 
> I was assuming we would find the address of the capability only once and map
> the relevant cap in for all the time the driver is bound, as is done for aer_cap.
> 
> I'd be worried that this might race with other accesses to rcrb in the future.
> 
> I suspect best option is to keep the capability always mapped in similar fashion
> to cxl_dport_map_rch_aer() does it for the aer registers.
> 

Thanks for your comment.

It seems like we had different assumptions.
I reviewed the implementation of aer in cxl driver again.
I will post a patch later that changes the memory mapping method to be similar to 
what cxl_dport_map_rch_aer() does.

I'd appreciate it if you could review the patch once it's posted.

> 
> 
> > +		return 0;
> > +	addr = ioremap(rcrb, SZ_4K);
> > +	if (!addr) {
> > +		dev_err(dev, "Failed to map region %pr\n", addr);
> > +		release_mem_region(rcrb, SZ_4K);
> > +		return 0;
> > +	}
> > +
> > +	ret = readl(addr + dport->rcrb.rcd_pcie_cap + offset);
> > +	release_mem_region(rcrb, SZ_4K);
> > +	return ret;
> > +}
> > +
> > +static ssize_t rcd_link_cap_show(struct device *dev,
> > +				   struct device_attribute *attr, char *buf) {
> > +	u32 linkcap = get_rcd_pcie_caps(dev, PCI_EXP_LNKCAP);
> > +
> > +	return sysfs_emit(buf, "%x\n", linkcap); } static
> > +DEVICE_ATTR_RO(rcd_link_cap);
> > +
> > +static ssize_t rcd_link_ctrl_show(struct device *dev,
> > +				   struct device_attribute *attr, char *buf) {
> > +	u16 linkctl = get_rcd_pcie_caps(dev, PCI_EXP_LNKCTL);
> > +
> > +	return sysfs_emit(buf, "%x\n", linkctl); } static
> > +DEVICE_ATTR_RO(rcd_link_ctrl);
> > +
> > +static ssize_t rcd_link_status_show(struct device *dev,
> > +				   struct device_attribute *attr, char *buf) {
> > +	u16 linksta = get_rcd_pcie_caps(dev, PCI_EXP_LNKSTA);
> > +
> > +	return sysfs_emit(buf, "%x\n", linksta);
> 
> Local variable doesn't add to readability as the PCI_EXP_LNKSTA makes it
> pretty obvious what is being queried.
> 

I want to specify the format of the output.
I will specify the format of the output by casting the value to the desired format.

> 	return sysfs_emit(buf, "%x\n",
> 			  get_rcd_pcie_caps(dev, PCI_EXP_LNKSTA));
> 
> > +}
> > +static DEVICE_ATTR_RO(rcd_link_status);
> > +
> > +static struct attribute *cxl_rcd_attrs[] = {
> > +		&dev_attr_rcd_link_cap.attr,
> > +		&dev_attr_rcd_link_ctrl.attr,
> 
> Indent one less tab.
> 
I will fix it.
> > +		&dev_attr_rcd_link_status.attr,
> > +		NULL
> > +};