[dpdk-dev,06/16] nfp: add nsp symbol resolution command

Message ID 1503591622-16232-7-git-send-email-alejandro.lucero@netronome.com (mailing list archive)
State Changes Requested, archived
Delegated to: Ferruh Yigit
Headers

Checks

Context Check Description
ci/Intel-compilation success Compilation OK
ci/checkpatch success coding style OK

Commit Message

Alejandro Lucero Aug. 24, 2017, 4:20 p.m. UTC
  Firmware has symbols helping to configure things like number of
PF ports, vNIC BARs addresses inside NFP memories, or ethernet
link state. Different firmware apps have different things to map
and likely different internal NFP addresses to use.

Host drivers can use the NSPU interface for getting symbol data
regarding different hardware configurations. Once the driver has
the information about a specific object, a mapping is required
configuring an NFP expansion bar creating a device PCI bar window.

Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
---
 drivers/net/nfp/nfp_nspu.c | 86 ++++++++++++++++++++++++++++++++++++++++++++++
 drivers/net/nfp/nfp_nspu.h |  3 ++
 2 files changed, 89 insertions(+)
  

Comments

Ferruh Yigit Aug. 28, 2017, 4:42 p.m. UTC | #1
On 8/24/2017 5:20 PM, Alejandro Lucero wrote:
> Firmware has symbols helping to configure things like number of
> PF ports, vNIC BARs addresses inside NFP memories, or ethernet
> link state. Different firmware apps have different things to map
> and likely different internal NFP addresses to use.
> 
> Host drivers can use the NSPU interface for getting symbol data
> regarding different hardware configurations. Once the driver has
> the information about a specific object, a mapping is required
> configuring an NFP expansion bar creating a device PCI bar window.
> 
> Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>

<...>

> +
> +	/* Adjusting address based on symbol location */
> +	if (domain >= 24 && domain << 28 && target == 7) {

gcc is giving following compiler warning [1]. Most probably intention is
the compare, but it is not clear, can you please check?

[1]
.../drivers/net/nfp/nfp_nspu.c: In function ‘nfp_nspu_set_bar_from_symbl’:
.../drivers/net/nfp/nfp_nspu.c:446:29: error: ‘<<’ in boolean context,
did you mean ‘<’ ? [-Werror=int-in-bool-context]
  if (domain >= 24 && domain << 28 && target == 7) {
                      ~~~~~~~^~~~~

> +		addr = 1ULL << 37 | addr | ((uint64_t)domain & 0x3) << 35;
> +	} else {
> +		addr = 1ULL << 39 | addr | ((uint64_t)domain & 0x3f) << 32;
> +		if (target == -7)
> +			target = 7;
> +	}
  
Alejandro Lucero Aug. 31, 2017, 9:35 a.m. UTC | #2
On Mon, Aug 28, 2017 at 5:42 PM, Ferruh Yigit <ferruh.yigit@intel.com>
wrote:

> On 8/24/2017 5:20 PM, Alejandro Lucero wrote:
> > Firmware has symbols helping to configure things like number of
> > PF ports, vNIC BARs addresses inside NFP memories, or ethernet
> > link state. Different firmware apps have different things to map
> > and likely different internal NFP addresses to use.
> >
> > Host drivers can use the NSPU interface for getting symbol data
> > regarding different hardware configurations. Once the driver has
> > the information about a specific object, a mapping is required
> > configuring an NFP expansion bar creating a device PCI bar window.
> >
> > Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
>
> <...>
>
> > +
> > +     /* Adjusting address based on symbol location */
> > +     if (domain >= 24 && domain << 28 && target == 7) {
>
> gcc is giving following compiler warning [1]. Most probably intention is
> the compare, but it is not clear, can you please check?
>

I have not seen this warning but I will add parenthesis for avoiding
ambiguity.

Thanks


>
> [1]
> .../drivers/net/nfp/nfp_nspu.c: In function ‘nfp_nspu_set_bar_from_symbl’:
> .../drivers/net/nfp/nfp_nspu.c:446:29: error: ‘<<’ in boolean context,
> did you mean ‘<’ ? [-Werror=int-in-bool-context]
>   if (domain >= 24 && domain << 28 && target == 7) {
>                       ~~~~~~~^~~~~
>
> > +             addr = 1ULL << 37 | addr | ((uint64_t)domain & 0x3) << 35;
> > +     } else {
> > +             addr = 1ULL << 39 | addr | ((uint64_t)domain & 0x3f) << 32;
> > +             if (target == -7)
> > +                     target = 7;
> > +     }
>
  

Patch

diff --git a/drivers/net/nfp/nfp_nspu.c b/drivers/net/nfp/nfp_nspu.c
index 57ee45f..4296be1 100644
--- a/drivers/net/nfp/nfp_nspu.c
+++ b/drivers/net/nfp/nfp_nspu.c
@@ -43,6 +43,7 @@ 
 /* NSP commands */
 #define NSP_CMD_RESET          1
 #define NSP_CMD_FW_LOAD        6
+#define NSP_CMD_GET_SYMBOL     14
 
 #define NSP_BUFFER_CFG_SIZE_MASK	(0xff)
 
@@ -368,3 +369,88 @@ 
 
 	return ret;
 }
+
+/* Firmware symbol descriptor size */
+#define NFP_SYM_DESC_LEN 40
+
+#define SYMBOL_DATA(b, off)     (*(int64_t *)((b) + (off)))
+#define SYMBOL_UDATA(b, off)     (*(uint64_t *)((b) + (off)))
+
+/* Firmware symbols contain information about how to access what they
+ * represent. It can be as simple as an numeric variable declared at a
+ * specific NFP memory, but it can also be more complex structures and
+ * related to specific hardware functionalities or components. Target,
+ * domain and address allow to create the BAR window for accessing such
+ * hw object and size defines the length to map.
+ *
+ * A vNIC is a network interface implemented inside the NFP and using a
+ * subset of device PCI BARs. Specific firmware symbols allow to map those
+ * vNIC bars by host drivers like the NFP PMD.
+ *
+ * Accessing what the symbol represents implies to map the access through
+ * a PCI BAR window. NFP expansion BARs are used in this regard through
+ * the NSPU interface.
+ */
+int
+nfp_nspu_set_bar_from_symbl(nspu_desc_t *desc, const char *symbl,
+			    uint32_t expbar, uint64_t *pcie_offset,
+			    ssize_t *size)
+{
+	int64_t type;
+	int64_t target;
+	int64_t domain;
+	uint64_t addr;
+	char *sym_buf;
+	int ret = 0;
+
+	sym_buf = malloc(desc->buf_size);
+	strncpy(sym_buf, symbl, strlen(symbl));
+	ret = nspu_command(desc, NSP_CMD_GET_SYMBOL, 1, 1, sym_buf,
+			   NFP_SYM_DESC_LEN, strlen(symbl));
+	if (ret) {
+		RTE_LOG(DEBUG, PMD, "symbol resolution (%s) failed\n", symbl);
+		goto clean;
+	}
+
+	/* Reading symbol information */
+	type = SYMBOL_DATA(sym_buf, 0);
+	target = SYMBOL_DATA(sym_buf, 8);
+	domain =  SYMBOL_DATA(sym_buf, 16);
+	addr = SYMBOL_UDATA(sym_buf, 24);
+	*size = (ssize_t)SYMBOL_UDATA(sym_buf, 32);
+
+	if (type != 1) {
+		RTE_LOG(INFO, PMD, "wrong symbol type\n");
+		ret = -EINVAL;
+		goto clean;
+	}
+	if (!(target == 7 || target == -7)) {
+		RTE_LOG(INFO, PMD, "wrong symbol target\n");
+		ret = -EINVAL;
+		goto clean;
+	}
+	if (domain == 8 || domain == 9) {
+		RTE_LOG(INFO, PMD, "wrong symbol domain\n");
+		ret = -EINVAL;
+		goto clean;
+	}
+
+	/* Adjusting address based on symbol location */
+	if (domain >= 24 && domain << 28 && target == 7) {
+		addr = 1ULL << 37 | addr | ((uint64_t)domain & 0x3) << 35;
+	} else {
+		addr = 1ULL << 39 | addr | ((uint64_t)domain & 0x3f) << 32;
+		if (target == -7)
+			target = 7;
+	}
+
+	/* Configuring NFP expansion bar for mapping specific PCI BAR window */
+	nfp_nspu_mem_bar_cfg(desc, expbar, target, addr, pcie_offset);
+
+	/* This is the PCI BAR offset to use by the host */
+	*pcie_offset |= ((expbar & 0x7) << (desc->barsz - 3));
+
+clean:
+	free(sym_buf);
+	return ret;
+}
diff --git a/drivers/net/nfp/nfp_nspu.h b/drivers/net/nfp/nfp_nspu.h
index 6e1c25f..7734b4f 100644
--- a/drivers/net/nfp/nfp_nspu.h
+++ b/drivers/net/nfp/nfp_nspu.h
@@ -74,3 +74,6 @@  int nfp_nspu_init(nspu_desc_t *desc, int nfp, int pcie_bar, size_t pcie_barsz,
 int nfp_nsp_get_abi_version(nspu_desc_t *desc, int *major, int *minor);
 int nfp_fw_reset(nspu_desc_t *nspu_desc);
 int nfp_fw_upload(nspu_desc_t *nspu_desc);
+int nfp_nspu_set_bar_from_symbl(nspu_desc_t *desc, const char *symbl,
+				uint32_t expbar, uint64_t *pcie_offset,
+				ssize_t *size);