2017-09-18 20:40:00

by Roy Pledge

[permalink] [raw]
Subject: [v5 00/12] soc/fsl/qbman: Enable QBMan on ARM Platforms

This patch series enables DPAA1 QBMan devices for ARM and
ARM64 architectures. This allows the LS1043A and LS1046A to use
QBMan functionality which allows access to ethernet and cyptographic
devices for example.

Changes since v4:
- Introduce a common function for QBMan private memory initialization
- Fix sparse warnings making sure that __iomem and __be32 are respected
- Control different memremap() attributes using a #define

Changes since v3:
- Use memremap() instead of ioremap() for non iomem QBMan portal regions
- Ensured the __iomem attribute is respected when accessing iomem mapped regions
- Removed calls to flush/invalidate/prefetch for ARM/ARM64 since mapping is done as write combine

Changes since v2:
- Fixed some misspellings
- Added 'no-map' constraint to device tree bindings
- Described ordering contraint on regions in the device tree
- Removed confusing comment regarding non-shareable mappings
- Added warning if old reserved-memory technique is used on ARM

Changes since v1:
- Reworked private memory allocations to use shared-dma-pool on ARM platforms


Claudiu Manoil (2):
soc/fsl/qbman: Drop L1_CACHE_BYTES compile time check
soc/fsl/qbman: Add missing headers on ARM

Madalin Bucur (4):
soc/fsl/qbman: Drop set/clear_bits usage
soc/fsl/qbman: add QMAN_REV32
soc/fsl/qbman: different register offsets on ARM
soc/fsl/qbman: Enable FSL_LAYERSCAPE config on ARM

Roy Pledge (5):
soc/fsl/qbman: Add common routine for QBMan private allocations
soc/fsl/qbman: Use shared-dma-pool for BMan private memory allocations
soc/fsl/qbman: Use shared-dma-pool for QMan private memory allocations
dt-bindings: soc/fsl: Update reserved memory binding for QBMan
soc/fsl/qbman: Rework portal mapping calls for ARM/PPC

Valentin Rothberg (1):
soc/fsl/qbman: Fix ARM32 typo

Documentation/devicetree/bindings/soc/fsl/bman.txt | 12 +--
Documentation/devicetree/bindings/soc/fsl/qman.txt | 26 ++++--
drivers/soc/fsl/qbman/Kconfig | 2 +-
drivers/soc/fsl/qbman/Makefile | 2 +-
drivers/soc/fsl/qbman/bman.c | 42 ++++++++--
drivers/soc/fsl/qbman/bman_ccsr.c | 15 ++++
drivers/soc/fsl/qbman/bman_portal.c | 23 +++---
drivers/soc/fsl/qbman/bman_priv.h | 8 +-
drivers/soc/fsl/qbman/dpaa_sys.c | 78 ++++++++++++++++++
drivers/soc/fsl/qbman/dpaa_sys.h | 25 ++++--
drivers/soc/fsl/qbman/qman.c | 77 +++++++++++++-----
drivers/soc/fsl/qbman/qman_ccsr.c | 95 +++++++++++++++-------
drivers/soc/fsl/qbman/qman_portal.c | 23 +++---
drivers/soc/fsl/qbman/qman_priv.h | 11 +--
drivers/soc/fsl/qbman/qman_test.h | 2 -
15 files changed, 318 insertions(+), 123 deletions(-)
create mode 100644 drivers/soc/fsl/qbman/dpaa_sys.c

--
2.7.4


2017-09-18 20:40:07

by Roy Pledge

[permalink] [raw]
Subject: [v5 09/12] soc/fsl/qbman: add QMAN_REV32

From: Madalin Bucur <[email protected]>

Add revision 3.2 of the QBMan block. This is the version
for LS1043A and LS1046A SoCs.

Signed-off-by: Madalin Bucur <[email protected]>
Signed-off-by: Roy Pledge <[email protected]>
---
drivers/soc/fsl/qbman/qman_ccsr.c | 2 ++
drivers/soc/fsl/qbman/qman_priv.h | 1 +
2 files changed, 3 insertions(+)

diff --git a/drivers/soc/fsl/qbman/qman_ccsr.c b/drivers/soc/fsl/qbman/qman_ccsr.c
index 607355b9..79cba58 100644
--- a/drivers/soc/fsl/qbman/qman_ccsr.c
+++ b/drivers/soc/fsl/qbman/qman_ccsr.c
@@ -719,6 +719,8 @@ static int fsl_qman_probe(struct platform_device *pdev)
qman_ip_rev = QMAN_REV30;
else if (major == 3 && minor == 1)
qman_ip_rev = QMAN_REV31;
+ else if (major == 3 && minor == 2)
+ qman_ip_rev = QMAN_REV32;
else {
dev_err(dev, "Unknown QMan version\n");
return -ENODEV;
diff --git a/drivers/soc/fsl/qbman/qman_priv.h b/drivers/soc/fsl/qbman/qman_priv.h
index 9407d2e..75a8f90 100644
--- a/drivers/soc/fsl/qbman/qman_priv.h
+++ b/drivers/soc/fsl/qbman/qman_priv.h
@@ -183,6 +183,7 @@ struct qm_portal_config {
#define QMAN_REV20 0x0200
#define QMAN_REV30 0x0300
#define QMAN_REV31 0x0301
+#define QMAN_REV32 0x0302
extern u16 qman_ip_rev; /* 0 if uninitialised, otherwise QMAN_REVx */

#define QM_FQID_RANGE_START 1 /* FQID 0 reserved for internal use */
--
2.7.4

2017-09-18 20:40:23

by Roy Pledge

[permalink] [raw]
Subject: [v5 12/12] soc/fsl/qbman: Enable FSL_LAYERSCAPE config on ARM

From: Madalin Bucur <[email protected]>

Signed-off-by: Madalin Bucur <[email protected]>
Signed-off-by: Claudiu Manoil <[email protected]>
[Stuart: changed to use ARCH_LAYERSCAPE]
Signed-off-by: Stuart Yoder <[email protected]>
Signed-off-by: Roy Pledge <[email protected]>
---
drivers/soc/fsl/qbman/Kconfig | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/soc/fsl/qbman/Kconfig b/drivers/soc/fsl/qbman/Kconfig
index 757033c..fb4e6bf 100644
--- a/drivers/soc/fsl/qbman/Kconfig
+++ b/drivers/soc/fsl/qbman/Kconfig
@@ -1,6 +1,6 @@
menuconfig FSL_DPAA
bool "Freescale DPAA 1.x support"
- depends on FSL_SOC_BOOKE
+ depends on (FSL_SOC_BOOKE || ARCH_LAYERSCAPE)
select GENERIC_ALLOCATOR
help
The Freescale Data Path Acceleration Architecture (DPAA) is a set of
--
2.7.4

2017-09-18 20:40:05

by Roy Pledge

[permalink] [raw]
Subject: [v5 07/12] soc/fsl/qbman: Fix ARM32 typo

From: Valentin Rothberg <[email protected]>

The Kconfig symbol for 32bit ARM is 'ARM', not 'ARM32'.

Signed-off-by: Valentin Rothberg <[email protected]>
Signed-off-by: Claudiu Manoil <[email protected]>
Signed-off-by: Roy Pledge <[email protected]>
---
drivers/soc/fsl/qbman/dpaa_sys.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/soc/fsl/qbman/dpaa_sys.h b/drivers/soc/fsl/qbman/dpaa_sys.h
index 4b1a467..61cfdb3 100644
--- a/drivers/soc/fsl/qbman/dpaa_sys.h
+++ b/drivers/soc/fsl/qbman/dpaa_sys.h
@@ -53,7 +53,7 @@ static inline void dpaa_flush(void *p)
{
#ifdef CONFIG_PPC
flush_dcache_range((unsigned long)p, (unsigned long)p+64);
-#elif defined(CONFIG_ARM32)
+#elif defined(CONFIG_ARM)
__cpuc_flush_dcache_area(p, 64);
#elif defined(CONFIG_ARM64)
__flush_dcache_area(p, 64);
--
2.7.4

2017-09-18 20:40:47

by Roy Pledge

[permalink] [raw]
Subject: [v5 11/12] soc/fsl/qbman: Add missing headers on ARM

From: Claudiu Manoil <[email protected]>

Unlike PPC builds, ARM builds need following headers
explicitly:
+#include <linux/io.h> for ioread32be()
+#include <linux/delay.h> for udelay()

Signed-off-by: Claudiu Manoil <[email protected]>
Signed-off-by: Roy Pledge <[email protected]>
---
drivers/soc/fsl/qbman/dpaa_sys.h | 2 ++
1 file changed, 2 insertions(+)

diff --git a/drivers/soc/fsl/qbman/dpaa_sys.h b/drivers/soc/fsl/qbman/dpaa_sys.h
index 5a2c0af..9f37900 100644
--- a/drivers/soc/fsl/qbman/dpaa_sys.h
+++ b/drivers/soc/fsl/qbman/dpaa_sys.h
@@ -44,6 +44,8 @@
#include <linux/prefetch.h>
#include <linux/genalloc.h>
#include <asm/cacheflush.h>
+#include <linux/io.h>
+#include <linux/delay.h>

/* For 2-element tables related to cache-inhibited and cache-enabled mappings */
#define DPAA_PORTAL_CE 0
--
2.7.4

2017-09-18 20:40:53

by Roy Pledge

[permalink] [raw]
Subject: [v5 08/12] soc/fsl/qbman: Rework portal mapping calls for ARM/PPC

Rework portal mapping for PPC and ARM. The PPC devices require a
cacheable coherent mapping while ARM will work with a non-cachable/write
combine mapping. This also eliminates the need for manual cache
flushes on ARM. This also fixes the code so sparse checking is clean.

Signed-off-by: Roy Pledge <[email protected]>
---
drivers/soc/fsl/qbman/bman.c | 18 ++++++++++--------
drivers/soc/fsl/qbman/bman_portal.c | 23 ++++++++++-------------
drivers/soc/fsl/qbman/bman_priv.h | 8 +++-----
drivers/soc/fsl/qbman/dpaa_sys.h | 15 +++++++++++----
drivers/soc/fsl/qbman/qman.c | 31 +++++++++++++------------------
drivers/soc/fsl/qbman/qman_portal.c | 23 ++++++++++-------------
drivers/soc/fsl/qbman/qman_priv.h | 8 +++-----
7 files changed, 60 insertions(+), 66 deletions(-)

diff --git a/drivers/soc/fsl/qbman/bman.c b/drivers/soc/fsl/qbman/bman.c
index ff8998f..5dbb5cc 100644
--- a/drivers/soc/fsl/qbman/bman.c
+++ b/drivers/soc/fsl/qbman/bman.c
@@ -154,7 +154,8 @@ struct bm_mc {
};

struct bm_addr {
- void __iomem *ce; /* cache-enabled */
+ void *ce; /* cache-enabled */
+ __be32 *ce_be; /* Same as above but for direct access */
void __iomem *ci; /* cache-inhibited */
};

@@ -167,12 +168,12 @@ struct bm_portal {
/* Cache-inhibited register access. */
static inline u32 bm_in(struct bm_portal *p, u32 offset)
{
- return be32_to_cpu(__raw_readl(p->addr.ci + offset));
+ return ioread32be(p->addr.ci + offset);
}

static inline void bm_out(struct bm_portal *p, u32 offset, u32 val)
{
- __raw_writel(cpu_to_be32(val), p->addr.ci + offset);
+ iowrite32be(val, p->addr.ci + offset);
}

/* Cache Enabled Portal Access */
@@ -188,7 +189,7 @@ static inline void bm_cl_touch_ro(struct bm_portal *p, u32 offset)

static inline u32 bm_ce_in(struct bm_portal *p, u32 offset)
{
- return be32_to_cpu(__raw_readl(p->addr.ce + offset));
+ return be32_to_cpu(*(p->addr.ce_be + (offset/4)));
}

struct bman_portal {
@@ -408,7 +409,7 @@ static int bm_mc_init(struct bm_portal *portal)

mc->cr = portal->addr.ce + BM_CL_CR;
mc->rr = portal->addr.ce + BM_CL_RR0;
- mc->rridx = (__raw_readb(&mc->cr->_ncw_verb) & BM_MCC_VERB_VBIT) ?
+ mc->rridx = (mc->cr->_ncw_verb & BM_MCC_VERB_VBIT) ?
0 : 1;
mc->vbit = mc->rridx ? BM_MCC_VERB_VBIT : 0;
#ifdef CONFIG_FSL_DPAA_CHECKING
@@ -466,7 +467,7 @@ static inline union bm_mc_result *bm_mc_result(struct bm_portal *portal)
* its command is submitted and completed. This includes the valid-bit,
* in case you were wondering...
*/
- if (!__raw_readb(&rr->verb)) {
+ if (!rr->verb) {
dpaa_invalidate_touch_ro(rr);
return NULL;
}
@@ -512,8 +513,9 @@ static int bman_create_portal(struct bman_portal *portal,
* config, everything that follows depends on it and "config" is more
* for (de)reference...
*/
- p->addr.ce = c->addr_virt[DPAA_PORTAL_CE];
- p->addr.ci = c->addr_virt[DPAA_PORTAL_CI];
+ p->addr.ce = c->addr_virt_ce;
+ p->addr.ce_be = c->addr_virt_ce;
+ p->addr.ci = c->addr_virt_ci;
if (bm_rcr_init(p, bm_rcr_pvb, bm_rcr_cce)) {
dev_err(c->dev, "RCR initialisation failed\n");
goto fail_rcr;
diff --git a/drivers/soc/fsl/qbman/bman_portal.c b/drivers/soc/fsl/qbman/bman_portal.c
index 39b39c8..2f71f7d 100644
--- a/drivers/soc/fsl/qbman/bman_portal.c
+++ b/drivers/soc/fsl/qbman/bman_portal.c
@@ -91,7 +91,6 @@ static int bman_portal_probe(struct platform_device *pdev)
struct device_node *node = dev->of_node;
struct bm_portal_config *pcfg;
struct resource *addr_phys[2];
- void __iomem *va;
int irq, cpu;

pcfg = devm_kmalloc(dev, sizeof(*pcfg), GFP_KERNEL);
@@ -123,23 +122,21 @@ static int bman_portal_probe(struct platform_device *pdev)
}
pcfg->irq = irq;

- va = ioremap_prot(addr_phys[0]->start, resource_size(addr_phys[0]), 0);
- if (!va) {
- dev_err(dev, "ioremap::CE failed\n");
+ pcfg->addr_virt_ce = memremap(addr_phys[0]->start,
+ resource_size(addr_phys[0]),
+ QBMAN_MEMREMAP_ATTR);
+ if (!pcfg->addr_virt_ce) {
+ dev_err(dev, "memremap::CE failed\n");
goto err_ioremap1;
}

- pcfg->addr_virt[DPAA_PORTAL_CE] = va;
-
- va = ioremap_prot(addr_phys[1]->start, resource_size(addr_phys[1]),
- _PAGE_GUARDED | _PAGE_NO_CACHE);
- if (!va) {
+ pcfg->addr_virt_ci = ioremap(addr_phys[1]->start,
+ resource_size(addr_phys[1]));
+ if (!pcfg->addr_virt_ci) {
dev_err(dev, "ioremap::CI failed\n");
goto err_ioremap2;
}

- pcfg->addr_virt[DPAA_PORTAL_CI] = va;
-
spin_lock(&bman_lock);
cpu = cpumask_next_zero(-1, &portal_cpus);
if (cpu >= nr_cpu_ids) {
@@ -164,9 +161,9 @@ static int bman_portal_probe(struct platform_device *pdev)
return 0;

err_portal_init:
- iounmap(pcfg->addr_virt[DPAA_PORTAL_CI]);
+ iounmap(pcfg->addr_virt_ci);
err_ioremap2:
- iounmap(pcfg->addr_virt[DPAA_PORTAL_CE]);
+ memunmap(pcfg->addr_virt_ce);
err_ioremap1:
return -ENXIO;
}
diff --git a/drivers/soc/fsl/qbman/bman_priv.h b/drivers/soc/fsl/qbman/bman_priv.h
index f6896a2..751ce90 100644
--- a/drivers/soc/fsl/qbman/bman_priv.h
+++ b/drivers/soc/fsl/qbman/bman_priv.h
@@ -46,11 +46,9 @@ extern u16 bman_ip_rev; /* 0 if uninitialised, otherwise BMAN_REVx */
extern struct gen_pool *bm_bpalloc;

struct bm_portal_config {
- /*
- * Corenet portal addresses;
- * [0]==cache-enabled, [1]==cache-inhibited.
- */
- void __iomem *addr_virt[2];
+ /* Portal addresses */
+ void *addr_virt_ce;
+ void __iomem *addr_virt_ci;
/* Allow these to be joined in lists */
struct list_head list;
struct device *dev;
diff --git a/drivers/soc/fsl/qbman/dpaa_sys.h b/drivers/soc/fsl/qbman/dpaa_sys.h
index 61cfdb3..5a2c0af 100644
--- a/drivers/soc/fsl/qbman/dpaa_sys.h
+++ b/drivers/soc/fsl/qbman/dpaa_sys.h
@@ -51,12 +51,12 @@

static inline void dpaa_flush(void *p)
{
+ /*
+ * Only PPC needs to flush the cache currently - on ARM the mapping
+ * is non cacheable
+ */
#ifdef CONFIG_PPC
flush_dcache_range((unsigned long)p, (unsigned long)p+64);
-#elif defined(CONFIG_ARM)
- __cpuc_flush_dcache_area(p, 64);
-#elif defined(CONFIG_ARM64)
- __flush_dcache_area(p, 64);
#endif
}

@@ -102,4 +102,11 @@ static inline u8 dpaa_cyc_diff(u8 ringsize, u8 first, u8 last)
int qbman_init_private_mem(struct device *dev, int idx, dma_addr_t *addr,
size_t *size);

+/* memremap() attributes for different platforms */
+#ifdef CONFIG_PPC
+#define QBMAN_MEMREMAP_ATTR MEMREMAP_WB
+#else
+#define QBMAN_MEMREMAP_ATTR MEMREMAP_WC
+#endif
+
#endif /* __DPAA_SYS_H */
diff --git a/drivers/soc/fsl/qbman/qman.c b/drivers/soc/fsl/qbman/qman.c
index 25419e1..8934c27 100644
--- a/drivers/soc/fsl/qbman/qman.c
+++ b/drivers/soc/fsl/qbman/qman.c
@@ -300,7 +300,8 @@ struct qm_mc {
};

struct qm_addr {
- void __iomem *ce; /* cache-enabled */
+ void *ce; /* cache-enabled */
+ __be32 *ce_be; /* same value as above but for direct access */
void __iomem *ci; /* cache-inhibited */
};

@@ -321,12 +322,12 @@ struct qm_portal {
/* Cache-inhibited register access. */
static inline u32 qm_in(struct qm_portal *p, u32 offset)
{
- return be32_to_cpu(__raw_readl(p->addr.ci + offset));
+ return ioread32be(p->addr.ci + offset);
}

static inline void qm_out(struct qm_portal *p, u32 offset, u32 val)
{
- __raw_writel(cpu_to_be32(val), p->addr.ci + offset);
+ iowrite32be(val, p->addr.ci + offset);
}

/* Cache Enabled Portal Access */
@@ -342,7 +343,7 @@ static inline void qm_cl_touch_ro(struct qm_portal *p, u32 offset)

static inline u32 qm_ce_in(struct qm_portal *p, u32 offset)
{
- return be32_to_cpu(__raw_readl(p->addr.ce + offset));
+ return be32_to_cpu(*(p->addr.ce_be + (offset/4)));
}

/* --- EQCR API --- */
@@ -646,11 +647,7 @@ static inline void qm_dqrr_pvb_update(struct qm_portal *portal)
*/
dpaa_invalidate_touch_ro(res);
#endif
- /*
- * when accessing 'verb', use __raw_readb() to ensure that compiler
- * inlining doesn't try to optimise out "excess reads".
- */
- if ((__raw_readb(&res->verb) & QM_DQRR_VERB_VBIT) == dqrr->vbit) {
+ if ((res->verb & QM_DQRR_VERB_VBIT) == dqrr->vbit) {
dqrr->pi = (dqrr->pi + 1) & (QM_DQRR_SIZE - 1);
if (!dqrr->pi)
dqrr->vbit ^= QM_DQRR_VERB_VBIT;
@@ -777,11 +774,8 @@ static inline void qm_mr_pvb_update(struct qm_portal *portal)
union qm_mr_entry *res = qm_cl(mr->ring, mr->pi);

DPAA_ASSERT(mr->pmode == qm_mr_pvb);
- /*
- * when accessing 'verb', use __raw_readb() to ensure that compiler
- * inlining doesn't try to optimise out "excess reads".
- */
- if ((__raw_readb(&res->verb) & QM_MR_VERB_VBIT) == mr->vbit) {
+
+ if ((res->verb & QM_MR_VERB_VBIT) == mr->vbit) {
mr->pi = (mr->pi + 1) & (QM_MR_SIZE - 1);
if (!mr->pi)
mr->vbit ^= QM_MR_VERB_VBIT;
@@ -822,7 +816,7 @@ static inline int qm_mc_init(struct qm_portal *portal)

mc->cr = portal->addr.ce + QM_CL_CR;
mc->rr = portal->addr.ce + QM_CL_RR0;
- mc->rridx = (__raw_readb(&mc->cr->_ncw_verb) & QM_MCC_VERB_VBIT)
+ mc->rridx = (mc->cr->_ncw_verb & QM_MCC_VERB_VBIT)
? 0 : 1;
mc->vbit = mc->rridx ? QM_MCC_VERB_VBIT : 0;
#ifdef CONFIG_FSL_DPAA_CHECKING
@@ -880,7 +874,7 @@ static inline union qm_mc_result *qm_mc_result(struct qm_portal *portal)
* its command is submitted and completed. This includes the valid-bit,
* in case you were wondering...
*/
- if (!__raw_readb(&rr->verb)) {
+ if (!rr->verb) {
dpaa_invalidate_touch_ro(rr);
return NULL;
}
@@ -1123,8 +1117,9 @@ static int qman_create_portal(struct qman_portal *portal,
* config, everything that follows depends on it and "config" is more
* for (de)reference
*/
- p->addr.ce = c->addr_virt[DPAA_PORTAL_CE];
- p->addr.ci = c->addr_virt[DPAA_PORTAL_CI];
+ p->addr.ce = c->addr_virt_ce;
+ p->addr.ce_be = c->addr_virt_ce;
+ p->addr.ci = c->addr_virt_ci;
/*
* If CI-stashing is used, the current defaults use a threshold of 3,
* and stash with high-than-DQRR priority.
diff --git a/drivers/soc/fsl/qbman/qman_portal.c b/drivers/soc/fsl/qbman/qman_portal.c
index cbacdf4..a120002 100644
--- a/drivers/soc/fsl/qbman/qman_portal.c
+++ b/drivers/soc/fsl/qbman/qman_portal.c
@@ -224,7 +224,6 @@ static int qman_portal_probe(struct platform_device *pdev)
struct device_node *node = dev->of_node;
struct qm_portal_config *pcfg;
struct resource *addr_phys[2];
- void __iomem *va;
int irq, cpu, err;
u32 val;

@@ -262,23 +261,21 @@ static int qman_portal_probe(struct platform_device *pdev)
}
pcfg->irq = irq;

- va = ioremap_prot(addr_phys[0]->start, resource_size(addr_phys[0]), 0);
- if (!va) {
- dev_err(dev, "ioremap::CE failed\n");
+ pcfg->addr_virt_ce = memremap(addr_phys[0]->start,
+ resource_size(addr_phys[0]),
+ QBMAN_MEMREMAP_ATTR);
+ if (!pcfg->addr_virt_ce) {
+ dev_err(dev, "memremap::CE failed\n");
goto err_ioremap1;
}

- pcfg->addr_virt[DPAA_PORTAL_CE] = va;
-
- va = ioremap_prot(addr_phys[1]->start, resource_size(addr_phys[1]),
- _PAGE_GUARDED | _PAGE_NO_CACHE);
- if (!va) {
+ pcfg->addr_virt_ci = ioremap(addr_phys[1]->start,
+ resource_size(addr_phys[1]));
+ if (!pcfg->addr_virt_ci) {
dev_err(dev, "ioremap::CI failed\n");
goto err_ioremap2;
}

- pcfg->addr_virt[DPAA_PORTAL_CI] = va;
-
pcfg->pools = qm_get_pools_sdqcr();

spin_lock(&qman_lock);
@@ -310,9 +307,9 @@ static int qman_portal_probe(struct platform_device *pdev)
return 0;

err_portal_init:
- iounmap(pcfg->addr_virt[DPAA_PORTAL_CI]);
+ iounmap(pcfg->addr_virt_ci);
err_ioremap2:
- iounmap(pcfg->addr_virt[DPAA_PORTAL_CE]);
+ memunmap(pcfg->addr_virt_ce);
err_ioremap1:
return -ENXIO;
}
diff --git a/drivers/soc/fsl/qbman/qman_priv.h b/drivers/soc/fsl/qbman/qman_priv.h
index b1e2cbf..9407d2e 100644
--- a/drivers/soc/fsl/qbman/qman_priv.h
+++ b/drivers/soc/fsl/qbman/qman_priv.h
@@ -153,11 +153,9 @@ static inline void qman_cgrs_xor(struct qman_cgrs *dest,
void qman_init_cgr_all(void);

struct qm_portal_config {
- /*
- * Corenet portal addresses;
- * [0]==cache-enabled, [1]==cache-inhibited.
- */
- void __iomem *addr_virt[2];
+ /* Portal addresses */
+ void *addr_virt_ce;
+ void __iomem *addr_virt_ci;
struct device *dev;
struct iommu_domain *iommu_domain;
/* Allow these to be joined in lists */
--
2.7.4

2017-09-18 20:40:50

by Roy Pledge

[permalink] [raw]
Subject: [v5 10/12] soc/fsl/qbman: different register offsets on ARM

From: Madalin Bucur <[email protected]>

Signed-off-by: Madalin Bucur <[email protected]>
Signed-off-by: Claudiu Manoil <[email protected]>
Signed-off-by: Roy Pledge <[email protected]>
---
drivers/soc/fsl/qbman/bman.c | 22 ++++++++++++++++++++++
drivers/soc/fsl/qbman/qman.c | 38 ++++++++++++++++++++++++++++++++++++++
2 files changed, 60 insertions(+)

diff --git a/drivers/soc/fsl/qbman/bman.c b/drivers/soc/fsl/qbman/bman.c
index 5dbb5cc..2e6e682 100644
--- a/drivers/soc/fsl/qbman/bman.c
+++ b/drivers/soc/fsl/qbman/bman.c
@@ -35,6 +35,27 @@

/* Portal register assists */

+#if defined(CONFIG_ARM) || defined(CONFIG_ARM64)
+/* Cache-inhibited register offsets */
+#define BM_REG_RCR_PI_CINH 0x3000
+#define BM_REG_RCR_CI_CINH 0x3100
+#define BM_REG_RCR_ITR 0x3200
+#define BM_REG_CFG 0x3300
+#define BM_REG_SCN(n) (0x3400 + ((n) << 6))
+#define BM_REG_ISR 0x3e00
+#define BM_REG_IER 0x3e40
+#define BM_REG_ISDR 0x3e80
+#define BM_REG_IIR 0x3ec0
+
+/* Cache-enabled register offsets */
+#define BM_CL_CR 0x0000
+#define BM_CL_RR0 0x0100
+#define BM_CL_RR1 0x0140
+#define BM_CL_RCR 0x1000
+#define BM_CL_RCR_PI_CENA 0x3000
+#define BM_CL_RCR_CI_CENA 0x3100
+
+#else
/* Cache-inhibited register offsets */
#define BM_REG_RCR_PI_CINH 0x0000
#define BM_REG_RCR_CI_CINH 0x0004
@@ -53,6 +74,7 @@
#define BM_CL_RCR 0x1000
#define BM_CL_RCR_PI_CENA 0x3000
#define BM_CL_RCR_CI_CENA 0x3100
+#endif

/*
* Portal modes.
diff --git a/drivers/soc/fsl/qbman/qman.c b/drivers/soc/fsl/qbman/qman.c
index 8934c27..7cb7bad 100644
--- a/drivers/soc/fsl/qbman/qman.c
+++ b/drivers/soc/fsl/qbman/qman.c
@@ -41,6 +41,43 @@

/* Portal register assists */

+#if defined(CONFIG_ARM) || defined(CONFIG_ARM64)
+/* Cache-inhibited register offsets */
+#define QM_REG_EQCR_PI_CINH 0x3000
+#define QM_REG_EQCR_CI_CINH 0x3040
+#define QM_REG_EQCR_ITR 0x3080
+#define QM_REG_DQRR_PI_CINH 0x3100
+#define QM_REG_DQRR_CI_CINH 0x3140
+#define QM_REG_DQRR_ITR 0x3180
+#define QM_REG_DQRR_DCAP 0x31C0
+#define QM_REG_DQRR_SDQCR 0x3200
+#define QM_REG_DQRR_VDQCR 0x3240
+#define QM_REG_DQRR_PDQCR 0x3280
+#define QM_REG_MR_PI_CINH 0x3300
+#define QM_REG_MR_CI_CINH 0x3340
+#define QM_REG_MR_ITR 0x3380
+#define QM_REG_CFG 0x3500
+#define QM_REG_ISR 0x3600
+#define QM_REG_IER 0x3640
+#define QM_REG_ISDR 0x3680
+#define QM_REG_IIR 0x36C0
+#define QM_REG_ITPR 0x3740
+
+/* Cache-enabled register offsets */
+#define QM_CL_EQCR 0x0000
+#define QM_CL_DQRR 0x1000
+#define QM_CL_MR 0x2000
+#define QM_CL_EQCR_PI_CENA 0x3000
+#define QM_CL_EQCR_CI_CENA 0x3040
+#define QM_CL_DQRR_PI_CENA 0x3100
+#define QM_CL_DQRR_CI_CENA 0x3140
+#define QM_CL_MR_PI_CENA 0x3300
+#define QM_CL_MR_CI_CENA 0x3340
+#define QM_CL_CR 0x3800
+#define QM_CL_RR0 0x3900
+#define QM_CL_RR1 0x3940
+
+#else
/* Cache-inhibited register offsets */
#define QM_REG_EQCR_PI_CINH 0x0000
#define QM_REG_EQCR_CI_CINH 0x0004
@@ -75,6 +112,7 @@
#define QM_CL_CR 0x3800
#define QM_CL_RR0 0x3900
#define QM_CL_RR1 0x3940
+#endif

/*
* BTW, the drivers (and h/w programming model) already obtain the required
--
2.7.4

2017-09-18 20:41:55

by Roy Pledge

[permalink] [raw]
Subject: [v5 06/12] soc/fsl/qbman: Drop L1_CACHE_BYTES compile time check

From: Claudiu Manoil <[email protected]>

Not relevant and arch dependent. Overkill for PPC.

Signed-off-by: Claudiu Manoil <[email protected]>
Signed-off-by: Roy Pledge <[email protected]>
---
drivers/soc/fsl/qbman/dpaa_sys.h | 4 ----
1 file changed, 4 deletions(-)

diff --git a/drivers/soc/fsl/qbman/dpaa_sys.h b/drivers/soc/fsl/qbman/dpaa_sys.h
index 676af82..4b1a467 100644
--- a/drivers/soc/fsl/qbman/dpaa_sys.h
+++ b/drivers/soc/fsl/qbman/dpaa_sys.h
@@ -49,10 +49,6 @@
#define DPAA_PORTAL_CE 0
#define DPAA_PORTAL_CI 1

-#if (L1_CACHE_BYTES != 32) && (L1_CACHE_BYTES != 64)
-#error "Unsupported Cacheline Size"
-#endif
-
static inline void dpaa_flush(void *p)
{
#ifdef CONFIG_PPC
--
2.7.4

2017-09-18 20:42:16

by Roy Pledge

[permalink] [raw]
Subject: [v5 05/12] soc/fsl/qbman: Drop set/clear_bits usage

From: Madalin Bucur <[email protected]>

Replace PPC specific set/clear_bits API with standard
bit twiddling so driver is portalable outside PPC.

Signed-off-by: Madalin Bucur <[email protected]>
Signed-off-by: Claudiu Manoil <[email protected]>
Signed-off-by: Roy Pledge <[email protected]>
---
drivers/soc/fsl/qbman/bman.c | 2 +-
drivers/soc/fsl/qbman/qman.c | 8 ++++----
2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/drivers/soc/fsl/qbman/bman.c b/drivers/soc/fsl/qbman/bman.c
index 604e45c..ff8998f 100644
--- a/drivers/soc/fsl/qbman/bman.c
+++ b/drivers/soc/fsl/qbman/bman.c
@@ -616,7 +616,7 @@ int bman_p_irqsource_add(struct bman_portal *p, u32 bits)
unsigned long irqflags;

local_irq_save(irqflags);
- set_bits(bits & BM_PIRQ_VISIBLE, &p->irq_sources);
+ p->irq_sources |= bits & BM_PIRQ_VISIBLE;
bm_out(&p->p, BM_REG_IER, p->irq_sources);
local_irq_restore(irqflags);
return 0;
diff --git a/drivers/soc/fsl/qbman/qman.c b/drivers/soc/fsl/qbman/qman.c
index 1bcfc51..25419e1 100644
--- a/drivers/soc/fsl/qbman/qman.c
+++ b/drivers/soc/fsl/qbman/qman.c
@@ -908,12 +908,12 @@ static inline int qm_mc_result_timeout(struct qm_portal *portal,

static inline void fq_set(struct qman_fq *fq, u32 mask)
{
- set_bits(mask, &fq->flags);
+ fq->flags |= mask;
}

static inline void fq_clear(struct qman_fq *fq, u32 mask)
{
- clear_bits(mask, &fq->flags);
+ fq->flags &= ~mask;
}

static inline int fq_isset(struct qman_fq *fq, u32 mask)
@@ -1574,7 +1574,7 @@ void qman_p_irqsource_add(struct qman_portal *p, u32 bits)
unsigned long irqflags;

local_irq_save(irqflags);
- set_bits(bits & QM_PIRQ_VISIBLE, &p->irq_sources);
+ p->irq_sources |= bits & QM_PIRQ_VISIBLE;
qm_out(&p->p, QM_REG_IER, p->irq_sources);
local_irq_restore(irqflags);
}
@@ -1597,7 +1597,7 @@ void qman_p_irqsource_remove(struct qman_portal *p, u32 bits)
*/
local_irq_save(irqflags);
bits &= QM_PIRQ_VISIBLE;
- clear_bits(bits, &p->irq_sources);
+ p->irq_sources &= ~bits;
qm_out(&p->p, QM_REG_IER, p->irq_sources);
ier = qm_in(&p->p, QM_REG_IER);
/*
--
2.7.4

2017-09-18 20:39:57

by Roy Pledge

[permalink] [raw]
Subject: [v5 01/12] soc/fsl/qbman: Add common routine for QBMan private allocations

The QBMan device uses several memory regions to manage frame
queues and buffers. Add a common routine for extracting and
initializing these reserved memory areas.

Signed-off-by: Roy Pledge <[email protected]>
---
drivers/soc/fsl/qbman/Makefile | 2 +-
drivers/soc/fsl/qbman/dpaa_sys.c | 78 ++++++++++++++++++++++++++++++++++++++++
drivers/soc/fsl/qbman/dpaa_sys.h | 4 +++
3 files changed, 83 insertions(+), 1 deletion(-)
create mode 100644 drivers/soc/fsl/qbman/dpaa_sys.c

diff --git a/drivers/soc/fsl/qbman/Makefile b/drivers/soc/fsl/qbman/Makefile
index 7ae199f..3cbd08a 100644
--- a/drivers/soc/fsl/qbman/Makefile
+++ b/drivers/soc/fsl/qbman/Makefile
@@ -1,6 +1,6 @@
obj-$(CONFIG_FSL_DPAA) += bman_ccsr.o qman_ccsr.o \
bman_portal.o qman_portal.o \
- bman.o qman.o
+ bman.o qman.o dpaa_sys.o

obj-$(CONFIG_FSL_BMAN_TEST) += bman-test.o
bman-test-y = bman_test.o
diff --git a/drivers/soc/fsl/qbman/dpaa_sys.c b/drivers/soc/fsl/qbman/dpaa_sys.c
new file mode 100644
index 0000000..9436aa8
--- /dev/null
+++ b/drivers/soc/fsl/qbman/dpaa_sys.c
@@ -0,0 +1,78 @@
+/* Copyright 2017 NXP Semiconductor, Inc.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in the
+ * documentation and/or other materials provided with the distribution.
+ * * Neither the name of NXP Semiconductor nor the
+ * names of its contributors may be used to endorse or promote products
+ * derived from this software without specific prior written permission.
+ *
+ * ALTERNATIVELY, this software may be distributed under the terms of the
+ * GNU General Public License ("GPL") as published by the Free Software
+ * Foundation, either version 2 of that License or (at your option) any
+ * later version.
+ *
+ * THIS SOFTWARE IS PROVIDED BY NXP Semiconductor ``AS IS'' AND ANY
+ * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+ * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+ * DISCLAIMED. IN NO EVENT SHALL NXP Semiconductor BE LIABLE FOR ANY
+ * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+ * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+ * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+ * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+ * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <linux/dma-mapping.h>
+#include "dpaa_sys.h"
+
+/*
+ * Initialize a devices private memory region
+ */
+int qbman_init_private_mem(struct device *dev, int idx, dma_addr_t *addr,
+ size_t *size)
+{
+ int ret;
+ struct device_node *mem_node;
+ u64 size64;
+
+ ret = of_reserved_mem_device_init_by_idx(dev, dev->of_node, idx);
+ if (ret) {
+ dev_err(dev,
+ "of_reserved_mem_device_init_by_idx(%d) failed 0x%x\n",
+ idx, ret);
+ return -ENODEV;
+ }
+ mem_node = of_parse_phandle(dev->of_node, "memory-region", 0);
+ if (mem_node) {
+ ret = of_property_read_u64(mem_node, "size", &size64);
+ if (ret) {
+ dev_err(dev, "of_address_to_resource fails 0x%x\n",
+ ret);
+ return -ENODEV;
+ }
+ *size = size64;
+ } else {
+ dev_err(dev, "No memory-region found for index %d\n", idx);
+ return -ENODEV;
+ }
+
+ if (!dma_zalloc_coherent(dev, *size, addr, 0)) {
+ dev_err(dev, "DMA Alloc memory failed\n");
+ return -ENODEV;
+ }
+
+ /*
+ * Disassociate the reserved memory area from the device
+ * because a device can only have one DMA memory area. This
+ * should be fine since the memory is allocated and initialized
+ * and only ever accessed by the QBMan device from now on
+ */
+ of_reserved_mem_device_release(dev);
+ return 0;
+}
diff --git a/drivers/soc/fsl/qbman/dpaa_sys.h b/drivers/soc/fsl/qbman/dpaa_sys.h
index 2ce394a..676af82 100644
--- a/drivers/soc/fsl/qbman/dpaa_sys.h
+++ b/drivers/soc/fsl/qbman/dpaa_sys.h
@@ -102,4 +102,8 @@ static inline u8 dpaa_cyc_diff(u8 ringsize, u8 first, u8 last)
/* Offset applied to genalloc pools due to zero being an error return */
#define DPAA_GENALLOC_OFF 0x80000000

+/* Initialize the devices private memory region */
+int qbman_init_private_mem(struct device *dev, int idx, dma_addr_t *addr,
+ size_t *size);
+
#endif /* __DPAA_SYS_H */
--
2.7.4

2017-09-18 20:42:12

by Roy Pledge

[permalink] [raw]
Subject: [v5 04/12] dt-bindings: soc/fsl: Update reserved memory binding for QBMan

Updates the QMan and BMan device tree bindings for reserved memory
nodes. This makes the reserved memory allocation compatible with
the shared-dma-pool usage.

Signed-off-by: Roy Pledge <[email protected]>
---
Documentation/devicetree/bindings/soc/fsl/bman.txt | 12 +++++-----
Documentation/devicetree/bindings/soc/fsl/qman.txt | 26 ++++++++++++++++------
2 files changed, 26 insertions(+), 12 deletions(-)

diff --git a/Documentation/devicetree/bindings/soc/fsl/bman.txt b/Documentation/devicetree/bindings/soc/fsl/bman.txt
index 47ac834..48eed14 100644
--- a/Documentation/devicetree/bindings/soc/fsl/bman.txt
+++ b/Documentation/devicetree/bindings/soc/fsl/bman.txt
@@ -65,8 +65,8 @@ to the respective BMan instance
BMan Private Memory Node

BMan requires a contiguous range of physical memory used for the backing store
-for BMan Free Buffer Proxy Records (FBPR). This memory is reserved/allocated as a
-node under the /reserved-memory node
+for BMan Free Buffer Proxy Records (FBPR). This memory is reserved/allocated as
+a node under the /reserved-memory node.

The BMan FBPR memory node must be named "bman-fbpr"

@@ -75,7 +75,9 @@ PROPERTIES
- compatible
Usage: required
Value type: <stringlist>
- Definition: Must inclide "fsl,bman-fbpr"
+ Definition: PPC platforms: Must include "fsl,bman-fbpr"
+ ARM platforms: Must include "shared-dma-pool"
+ as well as the "no-map" property

The following constraints are relevant to the FBPR private memory:
- The size must be 2^(size + 1), with size = 11..33. That is 4 KiB to
@@ -100,10 +102,10 @@ The example below shows a BMan FBPR dynamic allocation memory node
ranges;

bman_fbpr: bman-fbpr {
- compatible = "fsl,bman-fbpr";
- alloc-ranges = <0 0 0x10 0>;
+ compatible = "shared-mem-pool";
size = <0 0x1000000>;
alignment = <0 0x1000000>;
+ no-map;
};
};

diff --git a/Documentation/devicetree/bindings/soc/fsl/qman.txt b/Documentation/devicetree/bindings/soc/fsl/qman.txt
index 556ebb8..ee96afd 100644
--- a/Documentation/devicetree/bindings/soc/fsl/qman.txt
+++ b/Documentation/devicetree/bindings/soc/fsl/qman.txt
@@ -60,6 +60,12 @@ are located at offsets 0xbf8 and 0xbfc
Value type: <prop-encoded-array>
Definition: Reference input clock. Its frequency is half of the
platform clock
+- memory-regions
+ Usage: Required for ARM
+ Value type: <phandle array>
+ Definition: List of phandles referencing the QMan private memory
+ nodes (described below). The qman-fqd node must be
+ first followed by qman-pfdr node. Only used on ARM

Devices connected to a QMan instance via Direct Connect Portals (DCP) must link
to the respective QMan instance
@@ -74,7 +80,9 @@ QMan Private Memory Nodes

QMan requires two contiguous range of physical memory used for the backing store
for QMan Frame Queue Descriptor (FQD) and Packed Frame Descriptor Record (PFDR).
-This memory is reserved/allocated as a nodes under the /reserved-memory node
+This memory is reserved/allocated as a node under the /reserved-memory node.
+
+For additional details about reserved memory regions see reserved-memory.txt

The QMan FQD memory node must be named "qman-fqd"

@@ -83,7 +91,9 @@ PROPERTIES
- compatible
Usage: required
Value type: <stringlist>
- Definition: Must inclide "fsl,qman-fqd"
+ Definition: PPC platforms: Must include "fsl,qman-fqd"
+ ARM platforms: Must include "shared-dma-pool"
+ as well as the "no-map" property

The QMan PFDR memory node must be named "qman-pfdr"

@@ -92,7 +102,9 @@ PROPERTIES
- compatible
Usage: required
Value type: <stringlist>
- Definition: Must inclide "fsl,qman-pfdr"
+ Definition: PPC platforms: Must include "fsl,qman-pfdr"
+ ARM platforms: Must include "shared-dma-pool"
+ as well as the "no-map" property

The following constraints are relevant to the FQD and PFDR private memory:
- The size must be 2^(size + 1), with size = 11..29. That is 4 KiB to
@@ -117,16 +129,16 @@ The example below shows a QMan FQD and a PFDR dynamic allocation memory nodes
ranges;

qman_fqd: qman-fqd {
- compatible = "fsl,qman-fqd";
- alloc-ranges = <0 0 0x10 0>;
+ compatible = "shared-dma-pool";
size = <0 0x400000>;
alignment = <0 0x400000>;
+ no-map;
};
qman_pfdr: qman-pfdr {
- compatible = "fsl,qman-pfdr";
- alloc-ranges = <0 0 0x10 0>;
+ compatible = "shared-dma-pool";
size = <0 0x2000000>;
alignment = <0 0x2000000>;
+ no-map;
};
};

--
2.7.4

2017-09-18 20:42:51

by Roy Pledge

[permalink] [raw]
Subject: [v5 03/12] soc/fsl/qbman: Use shared-dma-pool for QMan private memory allocations

Use the shared-memory-pool mechanism for frame queue descriptor and
packed frame descriptor record area allocations.

Signed-off-by: Roy Pledge <[email protected]>
---
drivers/soc/fsl/qbman/qman_ccsr.c | 93 ++++++++++++++++++++++++++-------------
drivers/soc/fsl/qbman/qman_priv.h | 2 -
drivers/soc/fsl/qbman/qman_test.h | 2 -
3 files changed, 63 insertions(+), 34 deletions(-)

diff --git a/drivers/soc/fsl/qbman/qman_ccsr.c b/drivers/soc/fsl/qbman/qman_ccsr.c
index 835ce94..607355b9 100644
--- a/drivers/soc/fsl/qbman/qman_ccsr.c
+++ b/drivers/soc/fsl/qbman/qman_ccsr.c
@@ -401,21 +401,42 @@ static int qm_init_pfdr(struct device *dev, u32 pfdr_start, u32 num)
}

/*
- * Ideally we would use the DMA API to turn rmem->base into a DMA address
- * (especially if iommu translations ever get involved). Unfortunately, the
- * DMA API currently does not allow mapping anything that is not backed with
- * a struct page.
+ * QMan needs two global memory areas initialized at boot time:
+ * 1) FQD: Frame Queue Descriptors used to manage frame queues
+ * 2) PFDR: Packed Frame Queue Descriptor Records used to store frames
+ * Both areas are reserved using the device tree reserved memory framework
+ * and the addresses and sizes are initialized when the QMan device is probed
*/
static dma_addr_t fqd_a, pfdr_a;
static size_t fqd_sz, pfdr_sz;

+#ifdef CONFIG_PPC
+/*
+ * Support for PPC Device Tree backward compatibility when compatible
+ * string is set to fsl-qman-fqd and fsl-qman-pfdr
+ */
+static int zero_priv_mem(phys_addr_t addr, size_t sz)
+{
+ /* map as cacheable, non-guarded */
+ void __iomem *tmpp = ioremap_prot(addr, sz, 0);
+
+ if (!tmpp)
+ return -ENOMEM;
+
+ memset_io(tmpp, 0, sz);
+ flush_dcache_range((unsigned long)tmpp,
+ (unsigned long)tmpp + sz);
+ iounmap(tmpp);
+
+ return 0;
+}
+
static int qman_fqd(struct reserved_mem *rmem)
{
fqd_a = rmem->base;
fqd_sz = rmem->size;

WARN_ON(!(fqd_a && fqd_sz));
-
return 0;
}
RESERVEDMEM_OF_DECLARE(qman_fqd, "fsl,qman-fqd", qman_fqd);
@@ -431,32 +452,13 @@ static int qman_pfdr(struct reserved_mem *rmem)
}
RESERVEDMEM_OF_DECLARE(qman_pfdr, "fsl,qman-pfdr", qman_pfdr);

+#endif
+
static unsigned int qm_get_fqid_maxcnt(void)
{
return fqd_sz / 64;
}

-/*
- * Flush this memory range from data cache so that QMAN originated
- * transactions for this memory region could be marked non-coherent.
- */
-static int zero_priv_mem(struct device *dev, struct device_node *node,
- phys_addr_t addr, size_t sz)
-{
- /* map as cacheable, non-guarded */
- void __iomem *tmpp = ioremap_prot(addr, sz, 0);
-
- if (!tmpp)
- return -ENOMEM;
-
- memset_io(tmpp, 0, sz);
- flush_dcache_range((unsigned long)tmpp,
- (unsigned long)tmpp + sz);
- iounmap(tmpp);
-
- return 0;
-}
-
static void log_edata_bits(struct device *dev, u32 bit_count)
{
u32 i, j, mask = 0xffffffff;
@@ -727,10 +729,41 @@ static int fsl_qman_probe(struct platform_device *pdev)
qm_channel_caam = QMAN_CHANNEL_CAAM_REV3;
}

- ret = zero_priv_mem(dev, node, fqd_a, fqd_sz);
- WARN_ON(ret);
- if (ret)
- return -ENODEV;
+ if (fqd_a) {
+#ifdef CONFIG_PPC
+ /*
+ * For PPC backward DT compatibility
+ * FQD memory MUST be zero'd by software
+ */
+ zero_priv_mem(fqd_a, fqd_sz);
+#else
+ WARN(1, "Unexpected architecture using non shared-dma-mem reservations");
+#endif
+ } else {
+ /*
+ * Order of memory regions is assumed as FQD followed by PFDR
+ * in order to ensure allocations from the correct regions the
+ * driver initializes then allocates each piece in order
+ */
+ ret = qbman_init_private_mem(dev, 0, &fqd_a, &fqd_sz);
+ if (ret) {
+ dev_err(dev, "qbman_init_private_mem() for FQD failed 0x%x\n",
+ ret);
+ return -ENODEV;
+ }
+ }
+ dev_dbg(dev, "Allocated FQD 0x%llx 0x%zx\n", fqd_a, fqd_sz);
+
+ if (!pfdr_a) {
+ /* Setup PFDR memory */
+ ret = qbman_init_private_mem(dev, 1, &pfdr_a, &pfdr_sz);
+ if (ret) {
+ dev_err(dev, "qbman_init_private_mem() for PFDR failed 0x%x\n",
+ ret);
+ return -ENODEV;
+ }
+ }
+ dev_dbg(dev, "Allocated PFDR 0x%llx 0x%zx\n", pfdr_a, pfdr_sz);

ret = qman_init_ccsr(dev);
if (ret) {
diff --git a/drivers/soc/fsl/qbman/qman_priv.h b/drivers/soc/fsl/qbman/qman_priv.h
index 5fe9faf..b1e2cbf 100644
--- a/drivers/soc/fsl/qbman/qman_priv.h
+++ b/drivers/soc/fsl/qbman/qman_priv.h
@@ -28,8 +28,6 @@
* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/

-#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
-
#include "dpaa_sys.h"

#include <soc/fsl/qman.h>
diff --git a/drivers/soc/fsl/qbman/qman_test.h b/drivers/soc/fsl/qbman/qman_test.h
index d5f8cb2..41bdbc48 100644
--- a/drivers/soc/fsl/qbman/qman_test.h
+++ b/drivers/soc/fsl/qbman/qman_test.h
@@ -30,7 +30,5 @@

#include "qman_priv.h"

-#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
-
int qman_test_stash(void);
int qman_test_api(void);
--
2.7.4

2017-09-18 20:43:07

by Roy Pledge

[permalink] [raw]
Subject: [v5 02/12] soc/fsl/qbman: Use shared-dma-pool for BMan private memory allocations

Use the shared-memory-pool mechanism for free buffer proxy record
area allocation.

Signed-off-by: Roy Pledge <[email protected]>
---
drivers/soc/fsl/qbman/bman_ccsr.c | 15 +++++++++++++++
1 file changed, 15 insertions(+)

diff --git a/drivers/soc/fsl/qbman/bman_ccsr.c b/drivers/soc/fsl/qbman/bman_ccsr.c
index eaa9585..05c4223 100644
--- a/drivers/soc/fsl/qbman/bman_ccsr.c
+++ b/drivers/soc/fsl/qbman/bman_ccsr.c
@@ -201,6 +201,21 @@ static int fsl_bman_probe(struct platform_device *pdev)
return -ENODEV;
}

+ /*
+ * If FBPR memory wasn't defined using the qbman compatible string
+ * try using the of_reserved_mem_device method
+ */
+ if (!fbpr_a) {
+ ret = qbman_init_private_mem(dev, 0, &fbpr_a, &fbpr_sz);
+ if (ret) {
+ dev_err(dev, "qbman_init_private_mem() failed 0x%x\n",
+ ret);
+ return -ENODEV;
+ }
+ }
+
+ dev_dbg(dev, "Allocated FBPR 0x%llx 0x%zx\n", fbpr_a, fbpr_sz);
+
bm_set_memory(fbpr_a, fbpr_sz);

err_irq = platform_get_irq(pdev, 0);
--
2.7.4

2017-09-21 14:29:59

by Catalin Marinas

[permalink] [raw]
Subject: Re: [v5 01/12] soc/fsl/qbman: Add common routine for QBMan private allocations

On Mon, Sep 18, 2017 at 04:39:36PM -0400, Roy Pledge wrote:
> --- /dev/null
> +++ b/drivers/soc/fsl/qbman/dpaa_sys.c
> @@ -0,0 +1,78 @@
> +/* Copyright 2017 NXP Semiconductor, Inc.
> + *
> + * Redistribution and use in source and binary forms, with or without
> + * modification, are permitted provided that the following conditions are met:
> + * * Redistributions of source code must retain the above copyright
> + * notice, this list of conditions and the following disclaimer.
> + * * Redistributions in binary form must reproduce the above copyright
> + * notice, this list of conditions and the following disclaimer in the
> + * documentation and/or other materials provided with the distribution.
> + * * Neither the name of NXP Semiconductor nor the
> + * names of its contributors may be used to endorse or promote products
> + * derived from this software without specific prior written permission.
> + *
> + * ALTERNATIVELY, this software may be distributed under the terms of the
> + * GNU General Public License ("GPL") as published by the Free Software
> + * Foundation, either version 2 of that License or (at your option) any
> + * later version.
> + *
> + * THIS SOFTWARE IS PROVIDED BY NXP Semiconductor ``AS IS'' AND ANY
> + * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
> + * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
> + * DISCLAIMED. IN NO EVENT SHALL NXP Semiconductor BE LIABLE FOR ANY
> + * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
> + * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
> + * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
> + * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
> + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
> + * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
> + */

Pretty verbose header, I guess GPLv2 would do but IANAL.

Anyway:

Reviewed-by: Catalin Marinas <[email protected]>

2017-09-21 14:30:12

by Catalin Marinas

[permalink] [raw]
Subject: Re: [v5 02/12] soc/fsl/qbman: Use shared-dma-pool for BMan private memory allocations

On Mon, Sep 18, 2017 at 04:39:37PM -0400, Roy Pledge wrote:
> Use the shared-memory-pool mechanism for free buffer proxy record
> area allocation.
>
> Signed-off-by: Roy Pledge <[email protected]>

Reviewed-by: Catalin Marinas <[email protected]>

2017-09-21 14:30:37

by Catalin Marinas

[permalink] [raw]
Subject: Re: [v5 03/12] soc/fsl/qbman: Use shared-dma-pool for QMan private memory allocations

On Mon, Sep 18, 2017 at 04:39:38PM -0400, Roy Pledge wrote:
> Use the shared-memory-pool mechanism for frame queue descriptor and
> packed frame descriptor record area allocations.
>
> Signed-off-by: Roy Pledge <[email protected]>

Reviewed-by: Catalin Marinas <[email protected]>

2017-09-21 14:31:32

by Catalin Marinas

[permalink] [raw]
Subject: Re: [v5 05/12] soc/fsl/qbman: Drop set/clear_bits usage

On Mon, Sep 18, 2017 at 04:39:40PM -0400, Roy Pledge wrote:
> From: Madalin Bucur <[email protected]>
>
> Replace PPC specific set/clear_bits API with standard
> bit twiddling so driver is portalable outside PPC.
>
> Signed-off-by: Madalin Bucur <[email protected]>
> Signed-off-by: Claudiu Manoil <[email protected]>
> Signed-off-by: Roy Pledge <[email protected]>

Reviewed-by: Catalin Marinas <[email protected]>

2017-09-21 14:31:56

by Catalin Marinas

[permalink] [raw]
Subject: Re: [v5 06/12] soc/fsl/qbman: Drop L1_CACHE_BYTES compile time check

On Mon, Sep 18, 2017 at 04:39:41PM -0400, Roy Pledge wrote:
> From: Claudiu Manoil <[email protected]>
>
> Not relevant and arch dependent. Overkill for PPC.
>
> Signed-off-by: Claudiu Manoil <[email protected]>
> Signed-off-by: Roy Pledge <[email protected]>

Reviewed-by: Catalin Marinas <[email protected]>

2017-09-21 14:32:14

by Catalin Marinas

[permalink] [raw]
Subject: Re: [v5 07/12] soc/fsl/qbman: Fix ARM32 typo

On Mon, Sep 18, 2017 at 04:39:42PM -0400, Roy Pledge wrote:
> From: Valentin Rothberg <[email protected]>
>
> The Kconfig symbol for 32bit ARM is 'ARM', not 'ARM32'.
>
> Signed-off-by: Valentin Rothberg <[email protected]>
> Signed-off-by: Claudiu Manoil <[email protected]>
> Signed-off-by: Roy Pledge <[email protected]>

Reviewed-by: Catalin Marinas <[email protected]>

2017-09-21 14:32:48

by Catalin Marinas

[permalink] [raw]
Subject: Re: [v5 08/12] soc/fsl/qbman: Rework portal mapping calls for ARM/PPC

On Mon, Sep 18, 2017 at 04:39:43PM -0400, Roy Pledge wrote:
> Rework portal mapping for PPC and ARM. The PPC devices require a
> cacheable coherent mapping while ARM will work with a non-cachable/write
> combine mapping. This also eliminates the need for manual cache
> flushes on ARM. This also fixes the code so sparse checking is clean.
>
> Signed-off-by: Roy Pledge <[email protected]>

Reviewed-by: Catalin Marinas <[email protected]>

2017-09-21 14:40:33

by Catalin Marinas

[permalink] [raw]
Subject: Re: [v5 00/12] soc/fsl/qbman: Enable QBMan on ARM Platforms

On Mon, Sep 18, 2017 at 04:39:35PM -0400, Roy Pledge wrote:
> Madalin Bucur (4):
> soc/fsl/qbman: Drop set/clear_bits usage
> soc/fsl/qbman: add QMAN_REV32
> soc/fsl/qbman: different register offsets on ARM
> soc/fsl/qbman: Enable FSL_LAYERSCAPE config on ARM
>
> Roy Pledge (5):
> soc/fsl/qbman: Add common routine for QBMan private allocations
> soc/fsl/qbman: Use shared-dma-pool for BMan private memory allocations
> soc/fsl/qbman: Use shared-dma-pool for QMan private memory allocations
> dt-bindings: soc/fsl: Update reserved memory binding for QBMan
> soc/fsl/qbman: Rework portal mapping calls for ARM/PPC
>
> Valentin Rothberg (1):
> soc/fsl/qbman: Fix ARM32 typo

I reviewed the patches specific to the arm qbman buffer mappings, they
look fine to me. The DT bindings would need acks from the devicetree
maintainers. The rest of the patches are qbman-specific definitions
which I haven't checked for correctness (I guess they are fine).

How do you plan to upstream these patches? They would probably go in via
arm-soc.

--
Catalin

2017-09-21 18:36:54

by Leo Li

[permalink] [raw]
Subject: RE: [v5 04/12] dt-bindings: soc/fsl: Update reserved memory binding for QBMan


> -----Original Message-----
> From: Roy Pledge [mailto:[email protected]]
> Sent: Monday, September 18, 2017 3:40 PM
> To: Leo Li <[email protected]>; [email protected]; linux-arm-
> [email protected]; [email protected];
> [email protected]
> Cc: [email protected]; Madalin-cristian Bucur <[email protected]>;
> [email protected]; [email protected]; [email protected];
> [email protected]; Roy Pledge <[email protected]>
> Subject: [v5 04/12] dt-bindings: soc/fsl: Update reserved memory binding for
> QBMan
>
> Updates the QMan and BMan device tree bindings for reserved memory
> nodes. This makes the reserved memory allocation compatible with
> the shared-dma-pool usage.
>
> Signed-off-by: Roy Pledge <[email protected]>

Hi Rob and Mark,

Would you help to review this binding patch and ACK if it is ok?

Regards,
Leo

> ---
> Documentation/devicetree/bindings/soc/fsl/bman.txt | 12 +++++-----
> Documentation/devicetree/bindings/soc/fsl/qman.txt | 26 ++++++++++++++++-
> -----
> 2 files changed, 26 insertions(+), 12 deletions(-)
>
> diff --git a/Documentation/devicetree/bindings/soc/fsl/bman.txt
> b/Documentation/devicetree/bindings/soc/fsl/bman.txt
> index 47ac834..48eed14 100644
> --- a/Documentation/devicetree/bindings/soc/fsl/bman.txt
> +++ b/Documentation/devicetree/bindings/soc/fsl/bman.txt
> @@ -65,8 +65,8 @@ to the respective BMan instance
> BMan Private Memory Node
>
> BMan requires a contiguous range of physical memory used for the backing
> store
> -for BMan Free Buffer Proxy Records (FBPR). This memory is reserved/allocated
> as a
> -node under the /reserved-memory node
> +for BMan Free Buffer Proxy Records (FBPR). This memory is reserved/allocated
> as
> +a node under the /reserved-memory node.
>
> The BMan FBPR memory node must be named "bman-fbpr"
>
> @@ -75,7 +75,9 @@ PROPERTIES
> - compatible
> Usage: required
> Value type: <stringlist>
> - Definition: Must inclide "fsl,bman-fbpr"
> + Definition: PPC platforms: Must include "fsl,bman-fbpr"
> + ARM platforms: Must include "shared-dma-pool"
> + as well as the "no-map" property
>
> The following constraints are relevant to the FBPR private memory:
> - The size must be 2^(size + 1), with size = 11..33. That is 4 KiB to
> @@ -100,10 +102,10 @@ The example below shows a BMan FBPR dynamic
> allocation memory node
> ranges;
>
> bman_fbpr: bman-fbpr {
> - compatible = "fsl,bman-fbpr";
> - alloc-ranges = <0 0 0x10 0>;
> + compatible = "shared-mem-pool";
> size = <0 0x1000000>;
> alignment = <0 0x1000000>;
> + no-map;
> };
> };
>
> diff --git a/Documentation/devicetree/bindings/soc/fsl/qman.txt
> b/Documentation/devicetree/bindings/soc/fsl/qman.txt
> index 556ebb8..ee96afd 100644
> --- a/Documentation/devicetree/bindings/soc/fsl/qman.txt
> +++ b/Documentation/devicetree/bindings/soc/fsl/qman.txt
> @@ -60,6 +60,12 @@ are located at offsets 0xbf8 and 0xbfc
> Value type: <prop-encoded-array>
> Definition: Reference input clock. Its frequency is half of the
> platform clock
> +- memory-regions
> + Usage: Required for ARM
> + Value type: <phandle array>
> + Definition: List of phandles referencing the QMan private memory
> + nodes (described below). The qman-fqd node must be
> + first followed by qman-pfdr node. Only used on ARM
>
> Devices connected to a QMan instance via Direct Connect Portals (DCP) must
> link
> to the respective QMan instance
> @@ -74,7 +80,9 @@ QMan Private Memory Nodes
>
> QMan requires two contiguous range of physical memory used for the backing
> store
> for QMan Frame Queue Descriptor (FQD) and Packed Frame Descriptor Record
> (PFDR).
> -This memory is reserved/allocated as a nodes under the /reserved-memory
> node
> +This memory is reserved/allocated as a node under the /reserved-memory
> node.
> +
> +For additional details about reserved memory regions see reserved-memory.txt
>
> The QMan FQD memory node must be named "qman-fqd"
>
> @@ -83,7 +91,9 @@ PROPERTIES
> - compatible
> Usage: required
> Value type: <stringlist>
> - Definition: Must inclide "fsl,qman-fqd"
> + Definition: PPC platforms: Must include "fsl,qman-fqd"
> + ARM platforms: Must include "shared-dma-pool"
> + as well as the "no-map" property
>
> The QMan PFDR memory node must be named "qman-pfdr"
>
> @@ -92,7 +102,9 @@ PROPERTIES
> - compatible
> Usage: required
> Value type: <stringlist>
> - Definition: Must inclide "fsl,qman-pfdr"
> + Definition: PPC platforms: Must include "fsl,qman-pfdr"
> + ARM platforms: Must include "shared-dma-pool"
> + as well as the "no-map" property
>
> The following constraints are relevant to the FQD and PFDR private memory:
> - The size must be 2^(size + 1), with size = 11..29. That is 4 KiB to
> @@ -117,16 +129,16 @@ The example below shows a QMan FQD and a PFDR
> dynamic allocation memory nodes
> ranges;
>
> qman_fqd: qman-fqd {
> - compatible = "fsl,qman-fqd";
> - alloc-ranges = <0 0 0x10 0>;
> + compatible = "shared-dma-pool";
> size = <0 0x400000>;
> alignment = <0 0x400000>;
> + no-map;
> };
> qman_pfdr: qman-pfdr {
> - compatible = "fsl,qman-pfdr";
> - alloc-ranges = <0 0 0x10 0>;
> + compatible = "shared-dma-pool";
> size = <0 0x2000000>;
> alignment = <0 0x2000000>;
> + no-map;
> };
> };
>
> --
> 2.7.4


2017-09-21 23:14:57

by Rob Herring (Arm)

[permalink] [raw]
Subject: Re: [v5 04/12] dt-bindings: soc/fsl: Update reserved memory binding for QBMan

On Mon, Sep 18, 2017 at 04:39:39PM -0400, Roy Pledge wrote:
> Updates the QMan and BMan device tree bindings for reserved memory
> nodes. This makes the reserved memory allocation compatible with
> the shared-dma-pool usage.
>
> Signed-off-by: Roy Pledge <[email protected]>
> ---
> Documentation/devicetree/bindings/soc/fsl/bman.txt | 12 +++++-----
> Documentation/devicetree/bindings/soc/fsl/qman.txt | 26 ++++++++++++++++------
> 2 files changed, 26 insertions(+), 12 deletions(-)

Acked-by: Rob Herring <[email protected]>