Here's a series of fixes and changes for the DPAA 1.x drivers.
Fixing some boot time dependency issues, removing some dead code,
changing the buffers used for reception, fixing the DMA devices,
some cleanups.
Laurentiu Tudor (3):
fsl/fman: don't touch liodn base regs reserved on non-PAMU SoCs
dpaa_eth: defer probing after qbman
fsl/fman: add API to get the device behind a fman port
Madalin Bucur (17):
dpaa_eth: remove redundant code
dpaa_eth: change DMA device
fsl/fman: remove unused struct member
dpaa_eth: use only one buffer pool per interface
dpaa_eth: use page backed rx buffers
dpaa_eth: perform DMA unmapping before read
dpaa_eth: avoid timestamp read on error paths
dpaa_eth: simplify variables used in dpaa_cleanup_tx_fd()
dpaa_eth: use fd information in dpaa_cleanup_tx_fd()
dpaa_eth: use a page to store the SGT
soc: fsl: qbman: allow registering a device link for the portal user
dpaa_eth: register a device link for the qman portal used
dpaa_eth: add dropped frames to percpu ethtool stats
dpaa_eth: remove netdev_err() for user errors
dpaa_eth: extend delays in ndo_stop
dpaa_eth: add dpaa_dma_to_virt()
dpaa_eth: cleanup skb_to_contig_fd()
drivers/net/ethernet/freescale/dpaa/dpaa_eth.c | 407 +++++++++++----------
drivers/net/ethernet/freescale/dpaa/dpaa_eth.h | 12 +-
.../net/ethernet/freescale/dpaa/dpaa_eth_sysfs.c | 6 +-
drivers/net/ethernet/freescale/dpaa/dpaa_ethtool.c | 56 ++-
drivers/net/ethernet/freescale/fman/fman.c | 6 +-
drivers/net/ethernet/freescale/fman/fman_port.c | 17 +-
drivers/net/ethernet/freescale/fman/fman_port.h | 2 +
drivers/soc/fsl/qbman/qman.c | 7 +
include/soc/fsl/qman.h | 11 +
9 files changed, 285 insertions(+), 239 deletions(-)
--
2.1.0
Prior to this change, the frames dropped on receive or transmit
were not displayed in the ethtool statistics, leaving the dropped
frames unaccounted for.
Signed-off-by: Madalin Bucur <[email protected]>
---
drivers/net/ethernet/freescale/dpaa/dpaa_ethtool.c | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/drivers/net/ethernet/freescale/dpaa/dpaa_ethtool.c b/drivers/net/ethernet/freescale/dpaa/dpaa_ethtool.c
index bc6ed1df53ca..1c689e11c61f 100644
--- a/drivers/net/ethernet/freescale/dpaa/dpaa_ethtool.c
+++ b/drivers/net/ethernet/freescale/dpaa/dpaa_ethtool.c
@@ -47,6 +47,8 @@ static const char dpaa_stats_percpu[][ETH_GSTRING_LEN] = {
"tx S/G",
"tx error",
"rx error",
+ "rx dropped",
+ "tx dropped",
};
static char dpaa_stats_global[][ETH_GSTRING_LEN] = {
@@ -262,6 +264,12 @@ static void copy_stats(struct dpaa_percpu_priv *percpu_priv, int num_cpus,
data[crr * num_values + crr_cpu] = percpu_priv->stats.rx_errors;
data[crr++ * num_values + num_cpus] += percpu_priv->stats.rx_errors;
+ data[crr * num_values + crr_cpu] = percpu_priv->stats.rx_dropped;
+ data[crr++ * num_values + num_cpus] += percpu_priv->stats.rx_dropped;
+
+ data[crr * num_values + crr_cpu] = percpu_priv->stats.tx_dropped;
+ data[crr++ * num_values + num_cpus] += percpu_priv->stats.tx_dropped;
+
data[crr * num_values + crr_cpu] = bp_count;
data[crr++ * num_values + num_cpus] += bp_count;
}
--
2.1.0
Before this change, unbinding the QMan portals did not trigger a
corresponding unbinding of the dpaa_eth making use of it; the first
QMan portal related operation issued afterwards crashed the kernel.
The device link ensures the dpaa_eth dependency upon the qman portal
used is honoured at the QMan portal removal.
Signed-off-by: Madalin Bucur <[email protected]>
---
drivers/net/ethernet/freescale/dpaa/dpaa_eth.c | 5 +++--
drivers/soc/fsl/qbman/qman.c | 6 ------
include/soc/fsl/qman.h | 7 -------
3 files changed, 3 insertions(+), 15 deletions(-)
diff --git a/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c b/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c
index e2385c2fa81a..97fc067ae269 100644
--- a/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c
+++ b/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c
@@ -748,7 +748,7 @@ static void dpaa_release_channel(void)
qman_release_pool(rx_pool_channel);
}
-static void dpaa_eth_add_channel(u16 channel)
+static void dpaa_eth_add_channel(u16 channel, struct device *dev)
{
u32 pool = QM_SDQCR_CHANNELS_POOL_CONV(channel);
const cpumask_t *cpus = qman_affine_cpus();
@@ -758,6 +758,7 @@ static void dpaa_eth_add_channel(u16 channel)
for_each_cpu_and(cpu, cpus, cpu_online_mask) {
portal = qman_get_affine_portal(cpu);
qman_p_static_dequeue_add(portal, pool);
+ qman_start_using_portal(portal, dev);
}
}
@@ -2871,7 +2872,7 @@ static int dpaa_eth_probe(struct platform_device *pdev)
/* Walk the CPUs with affine portals
* and add this pool channel to each's dequeue mask.
*/
- dpaa_eth_add_channel(priv->channel);
+ dpaa_eth_add_channel(priv->channel, &pdev->dev);
dpaa_fq_setup(priv, &dpaa_fq_cbs, priv->mac_dev->port[TX]);
diff --git a/drivers/soc/fsl/qbman/qman.c b/drivers/soc/fsl/qbman/qman.c
index bc75a5882b9e..1e164e03410a 100644
--- a/drivers/soc/fsl/qbman/qman.c
+++ b/drivers/soc/fsl/qbman/qman.c
@@ -1756,12 +1756,6 @@ int qman_start_using_portal(struct qman_portal *p, struct device *dev)
}
EXPORT_SYMBOL(qman_start_using_portal);
-void qman_stop_using_portal(struct qman_portal *p, struct device *dev)
-{
- device_link_remove(dev, p->config->dev);
-}
-EXPORT_SYMBOL(qman_stop_using_portal);
-
int qman_p_poll_dqrr(struct qman_portal *p, unsigned int limit)
{
return __poll_portal_fast(p, limit);
diff --git a/include/soc/fsl/qman.h b/include/soc/fsl/qman.h
index c499c5cfa7c9..cfe00e08e85b 100644
--- a/include/soc/fsl/qman.h
+++ b/include/soc/fsl/qman.h
@@ -926,13 +926,6 @@ struct qman_portal *qman_get_affine_portal(int cpu);
int qman_start_using_portal(struct qman_portal *p, struct device *dev);
/**
- * qman_stop_using_portal - deregister a device link for the portal user
- * @p: the portal that will no longer be in use
- * @dev: the device that uses the portal
- */
-void qman_stop_using_portal(struct qman_portal *p, struct device *dev);
-
-/**
* qman_p_poll_dqrr - process DQRR (fast-path) entries
* @limit: the maximum number of DQRR entries to process
*
--
2.1.0
From: Laurentiu Tudor <[email protected]>
If the DPAA 1 Ethernet driver gets probed before the QBMan driver it will
cause a boot crash. Add predictability in the probing order by deferring
the Ethernet driver probe after QBMan and portals by using the recently
introduced QBMan APIs.
Signed-off-by: Laurentiu Tudor <[email protected]>
Signed-off-by: Madalin Bucur <[email protected]>
---
drivers/net/ethernet/freescale/dpaa/dpaa_eth.c | 31 ++++++++++++++++++++++++++
1 file changed, 31 insertions(+)
diff --git a/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c b/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c
index b4b82b9c5cd6..75eeb2ef409f 100644
--- a/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c
+++ b/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c
@@ -2773,6 +2773,37 @@ static int dpaa_eth_probe(struct platform_device *pdev)
int err = 0, i, channel;
struct device *dev;
+ err = bman_is_probed();
+ if (!err)
+ return -EPROBE_DEFER;
+ if (err < 0) {
+ dev_err(&pdev->dev, "failing probe due to bman probe error\n");
+ return -ENODEV;
+ }
+ err = qman_is_probed();
+ if (!err)
+ return -EPROBE_DEFER;
+ if (err < 0) {
+ dev_err(&pdev->dev, "failing probe due to qman probe error\n");
+ return -ENODEV;
+ }
+ err = bman_portals_probed();
+ if (!err)
+ return -EPROBE_DEFER;
+ if (err < 0) {
+ dev_err(&pdev->dev,
+ "failing probe due to bman portals probe error\n");
+ return -ENODEV;
+ }
+ err = qman_portals_probed();
+ if (!err)
+ return -EPROBE_DEFER;
+ if (err < 0) {
+ dev_err(&pdev->dev,
+ "failing probe due to qman portals probe error\n");
+ return -ENODEV;
+ }
+
/* device used for DMA mapping */
dev = pdev->dev.parent;
err = dma_coerce_mask_and_coherent(dev, DMA_BIT_MASK(40));
--
2.1.0
Condition was previously checked, removing duplicate code.
Signed-off-by: Madalin Bucur <[email protected]>
---
drivers/net/ethernet/freescale/dpaa/dpaa_eth.c | 4 ----
1 file changed, 4 deletions(-)
diff --git a/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c b/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c
index 75eeb2ef409f..8d5686d88d30 100644
--- a/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c
+++ b/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c
@@ -2304,10 +2304,6 @@ static enum qman_cb_dqrr_result rx_default_dqrr(struct qman_portal *portal,
return qman_cb_dqrr_consume;
}
- dpaa_bp = dpaa_bpid2pool(fd->bpid);
- if (!dpaa_bp)
- return qman_cb_dqrr_consume;
-
dma_unmap_single(dpaa_bp->dev, addr, dpaa_bp->size, DMA_FROM_DEVICE);
/* prefetch the first 64 bytes of the frame or the SGT start */
--
2.1.0
Remove unused struct member second_largest_buf_size. Also, an out of
bounds access would have occurred in the removed code if there was only
one buffer pool in use.
Signed-off-by: Madalin Bucur <[email protected]>
---
drivers/net/ethernet/freescale/fman/fman_port.c | 3 ---
1 file changed, 3 deletions(-)
diff --git a/drivers/net/ethernet/freescale/fman/fman_port.c b/drivers/net/ethernet/freescale/fman/fman_port.c
index bd76c9730692..87b26f063cc8 100644
--- a/drivers/net/ethernet/freescale/fman/fman_port.c
+++ b/drivers/net/ethernet/freescale/fman/fman_port.c
@@ -435,7 +435,6 @@ struct fman_port_cfg {
struct fman_port_rx_pools_params {
u8 num_of_pools;
- u16 second_largest_buf_size;
u16 largest_buf_size;
};
@@ -946,8 +945,6 @@ static int set_ext_buffer_pools(struct fman_port *port)
port->rx_pools_params.num_of_pools = ext_buf_pools->num_of_pools_used;
port->rx_pools_params.largest_buf_size =
sizes_array[ordered_array[ext_buf_pools->num_of_pools_used - 1]];
- port->rx_pools_params.second_largest_buf_size =
- sizes_array[ordered_array[ext_buf_pools->num_of_pools_used - 2]];
/* FMBM_RMPD reg. - pool depletion */
if (buf_pool_depletion->pools_grp_mode_enable) {
--
2.1.0
The dpaa_cleanup_tx_fd() function is called by the frame transmit
confirmation callback but also on several error paths. This function
is reading the transmit timestamp value. Avoid reading an invalid
timestamp value on the error paths.
Fixes: 4664856e9ca2 ("dpaa_eth: add support for hardware timestamping")
Signed-off-by: Madalin Bucur <[email protected]>
---
drivers/net/ethernet/freescale/dpaa/dpaa_eth.c | 17 ++++++++++-------
1 file changed, 10 insertions(+), 7 deletions(-)
diff --git a/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c b/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c
index c3f19485739b..14ebdb10fcd4 100644
--- a/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c
+++ b/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c
@@ -1571,13 +1571,15 @@ static int dpaa_eth_refill_bpools(struct dpaa_priv *priv)
* Skb freeing is not handled here.
*
* This function may be called on error paths in the Tx function, so guard
- * against cases when not all fd relevant fields were filled in.
+ * against cases when not all fd relevant fields were filled in. To avoid
+ * reading the invalid transmission timestamp for the error paths set ts to
+ * false.
*
* Return the skb backpointer, since for S/G frames the buffer containing it
* gets freed here.
*/
static struct sk_buff *dpaa_cleanup_tx_fd(const struct dpaa_priv *priv,
- const struct qm_fd *fd)
+ const struct qm_fd *fd, bool ts)
{
const enum dma_data_direction dma_dir = DMA_TO_DEVICE;
struct device *dev = priv->net_dev->dev.parent;
@@ -1619,7 +1621,8 @@ static struct sk_buff *dpaa_cleanup_tx_fd(const struct dpaa_priv *priv,
}
/* DMA unmapping is required before accessing the HW provided info */
- if (priv->tx_tstamp && skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) {
+ if (ts && priv->tx_tstamp &&
+ skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) {
memset(&shhwtstamps, 0, sizeof(shhwtstamps));
if (!fman_port_get_tstamp(priv->mac_dev->port[TX], (void *)skbh,
@@ -2085,7 +2088,7 @@ dpaa_start_xmit(struct sk_buff *skb, struct net_device *net_dev)
if (likely(dpaa_xmit(priv, percpu_stats, queue_mapping, &fd) == 0))
return NETDEV_TX_OK;
- dpaa_cleanup_tx_fd(priv, &fd);
+ dpaa_cleanup_tx_fd(priv, &fd, false);
skb_to_fd_failed:
enomem:
percpu_stats->tx_errors++;
@@ -2131,7 +2134,7 @@ static void dpaa_tx_error(struct net_device *net_dev,
percpu_priv->stats.tx_errors++;
- skb = dpaa_cleanup_tx_fd(priv, fd);
+ skb = dpaa_cleanup_tx_fd(priv, fd, false);
dev_kfree_skb(skb);
}
@@ -2171,7 +2174,7 @@ static void dpaa_tx_conf(struct net_device *net_dev,
percpu_priv->tx_confirm++;
- skb = dpaa_cleanup_tx_fd(priv, fd);
+ skb = dpaa_cleanup_tx_fd(priv, fd, true);
consume_skb(skb);
}
@@ -2398,7 +2401,7 @@ static void egress_ern(struct qman_portal *portal,
percpu_priv->stats.tx_fifo_errors++;
count_ern(percpu_priv, msg);
- skb = dpaa_cleanup_tx_fd(priv, fd);
+ skb = dpaa_cleanup_tx_fd(priv, fd, false);
dev_kfree_skb_any(skb);
}
--
2.1.0
Change the buffers used for reception from netdev_frags to pages.
Signed-off-by: Madalin Bucur <[email protected]>
---
drivers/net/ethernet/freescale/dpaa/dpaa_eth.c | 51 +++++++++++---------------
1 file changed, 22 insertions(+), 29 deletions(-)
diff --git a/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c b/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c
index 7bfe437b0a50..ee267918aa70 100644
--- a/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c
+++ b/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c
@@ -180,13 +180,7 @@ static struct dpaa_bp *dpaa_bp_array[BM_MAX_NUM_OF_POOLS];
#define DPAA_BP_RAW_SIZE 4096
-/* FMan-DMA requires 16-byte alignment for Rx buffers, but SKB_DATA_ALIGN is
- * even stronger (SMP_CACHE_BYTES-aligned), so we just get away with that,
- * via SKB_WITH_OVERHEAD(). We can't rely on netdev_alloc_frag() giving us
- * half-page-aligned buffers, so we reserve some more space for start-of-buffer
- * alignment.
- */
-#define dpaa_bp_size(raw_size) SKB_WITH_OVERHEAD((raw_size) - SMP_CACHE_BYTES)
+#define dpaa_bp_size(raw_size) SKB_WITH_OVERHEAD(raw_size)
static int dpaa_max_frm;
@@ -1313,13 +1307,14 @@ static void dpaa_fd_release(const struct net_device *net_dev,
vaddr = phys_to_virt(qm_fd_addr(fd));
sgt = vaddr + qm_fd_get_offset(fd);
- dma_unmap_single(dpaa_bp->priv->rx_dma_dev, qm_fd_addr(fd),
- dpaa_bp->size, DMA_FROM_DEVICE);
+ dma_unmap_page(dpaa_bp->priv->rx_dma_dev, qm_fd_addr(fd),
+ DPAA_BP_RAW_SIZE, DMA_FROM_DEVICE);
dpaa_release_sgt_members(sgt);
- addr = dma_map_single(dpaa_bp->priv->rx_dma_dev, vaddr,
- dpaa_bp->size, DMA_FROM_DEVICE);
+ addr = dma_map_page(dpaa_bp->priv->rx_dma_dev,
+ virt_to_page(vaddr), 0, DPAA_BP_RAW_SIZE,
+ DMA_FROM_DEVICE);
if (dma_mapping_error(dpaa_bp->priv->rx_dma_dev, addr)) {
netdev_err(net_dev, "DMA mapping failed");
return;
@@ -1469,21 +1464,18 @@ static int dpaa_bp_add_8_bufs(const struct dpaa_bp *dpaa_bp)
struct net_device *net_dev = dpaa_bp->priv->net_dev;
struct bm_buffer bmb[8];
dma_addr_t addr;
- void *new_buf;
+ struct page *p;
u8 i;
for (i = 0; i < 8; i++) {
- new_buf = netdev_alloc_frag(dpaa_bp->raw_size);
- if (unlikely(!new_buf)) {
- netdev_err(net_dev,
- "netdev_alloc_frag() failed, size %zu\n",
- dpaa_bp->raw_size);
+ p = dev_alloc_pages(0);
+ if (unlikely(!p)) {
+ netdev_err(net_dev, "dev_alloc_pages() failed\n");
goto release_previous_buffs;
}
- new_buf = PTR_ALIGN(new_buf, SMP_CACHE_BYTES);
- addr = dma_map_single(dpaa_bp->priv->rx_dma_dev, new_buf,
- dpaa_bp->size, DMA_FROM_DEVICE);
+ addr = dma_map_page(dpaa_bp->priv->rx_dma_dev, p, 0,
+ DPAA_BP_RAW_SIZE, DMA_FROM_DEVICE);
if (unlikely(dma_mapping_error(dpaa_bp->priv->rx_dma_dev,
addr))) {
netdev_err(net_dev, "DMA map failed");
@@ -1694,7 +1686,7 @@ static struct sk_buff *contig_fd_to_skb(const struct dpaa_priv *priv,
return skb;
free_buffer:
- skb_free_frag(vaddr);
+ free_pages((unsigned long)vaddr, 0);
return NULL;
}
@@ -1741,8 +1733,8 @@ static struct sk_buff *sg_fd_to_skb(const struct dpaa_priv *priv,
goto free_buffers;
count_ptr = this_cpu_ptr(dpaa_bp->percpu_count);
- dma_unmap_single(dpaa_bp->priv->rx_dma_dev, sg_addr,
- dpaa_bp->size, DMA_FROM_DEVICE);
+ dma_unmap_page(dpaa_bp->priv->rx_dma_dev, sg_addr,
+ DPAA_BP_RAW_SIZE, DMA_FROM_DEVICE);
if (!skb) {
sz = dpaa_bp->size +
SKB_DATA_ALIGN(sizeof(struct skb_shared_info));
@@ -1794,7 +1786,7 @@ static struct sk_buff *sg_fd_to_skb(const struct dpaa_priv *priv,
WARN_ONCE(i == DPAA_SGT_MAX_ENTRIES, "No final bit on SGT\n");
/* free the SG table buffer */
- skb_free_frag(vaddr);
+ free_pages((unsigned long)vaddr, 0);
return skb;
@@ -1811,7 +1803,7 @@ static struct sk_buff *sg_fd_to_skb(const struct dpaa_priv *priv,
for (i = 0; i < DPAA_SGT_MAX_ENTRIES ; i++) {
sg_addr = qm_sg_addr(&sgt[i]);
sg_vaddr = phys_to_virt(sg_addr);
- skb_free_frag(sg_vaddr);
+ free_pages((unsigned long)sg_vaddr, 0);
dpaa_bp = dpaa_bpid2pool(sgt[i].bpid);
if (dpaa_bp) {
count_ptr = this_cpu_ptr(dpaa_bp->percpu_count);
@@ -1822,7 +1814,7 @@ static struct sk_buff *sg_fd_to_skb(const struct dpaa_priv *priv,
break;
}
/* free the SGT fragment */
- skb_free_frag(vaddr);
+ free_pages((unsigned long)vaddr, 0);
return NULL;
}
@@ -2281,8 +2273,8 @@ static enum qman_cb_dqrr_result rx_default_dqrr(struct qman_portal *portal,
return qman_cb_dqrr_consume;
}
- dma_unmap_single(dpaa_bp->priv->rx_dma_dev, addr, dpaa_bp->size,
- DMA_FROM_DEVICE);
+ dma_unmap_page(dpaa_bp->priv->rx_dma_dev, addr, DPAA_BP_RAW_SIZE,
+ DMA_FROM_DEVICE);
/* prefetch the first 64 bytes of the frame or the SGT start */
vaddr = phys_to_virt(addr);
@@ -2637,7 +2629,8 @@ static inline void dpaa_bp_free_pf(const struct dpaa_bp *bp,
{
dma_addr_t addr = bm_buf_addr(bmb);
- dma_unmap_single(bp->priv->rx_dma_dev, addr, bp->size, DMA_FROM_DEVICE);
+ dma_unmap_page(bp->priv->rx_dma_dev, addr, DPAA_BP_RAW_SIZE,
+ DMA_FROM_DEVICE);
skb_free_frag(phys_to_virt(addr));
}
--
2.1.0
Use a page to store the scatter gather table on the transmit path.
Signed-off-by: Madalin Bucur <[email protected]>
---
drivers/net/ethernet/freescale/dpaa/dpaa_eth.c | 43 +++++++++++++-------------
1 file changed, 21 insertions(+), 22 deletions(-)
diff --git a/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c b/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c
index 20f0062afdec..e2385c2fa81a 100644
--- a/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c
+++ b/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c
@@ -1592,9 +1592,9 @@ static struct sk_buff *dpaa_cleanup_tx_fd(const struct dpaa_priv *priv,
int i;
if (unlikely(qm_fd_get_format(fd) == qm_fd_sg)) {
- dma_unmap_single(priv->tx_dma_dev, addr,
- qm_fd_get_offset(fd) + DPAA_SGT_SIZE,
- dma_dir);
+ dma_unmap_page(priv->tx_dma_dev, addr,
+ qm_fd_get_offset(fd) + DPAA_SGT_SIZE,
+ dma_dir);
/* The sgt buffer has been allocated with netdev_alloc_frag(),
* it's from lowmem.
@@ -1636,8 +1636,8 @@ static struct sk_buff *dpaa_cleanup_tx_fd(const struct dpaa_priv *priv,
}
if (qm_fd_get_format(fd) == qm_fd_sg)
- /* Free the page frag that we allocated on Tx */
- skb_free_frag(vaddr);
+ /* Free the page that we allocated on Tx for the SGT */
+ free_pages((unsigned long)vaddr, 0);
return skb;
}
@@ -1885,21 +1885,20 @@ static int skb_to_sg_fd(struct dpaa_priv *priv,
struct net_device *net_dev = priv->net_dev;
struct qm_sg_entry *sgt;
struct sk_buff **skbh;
- int i, j, err, sz;
- void *buffer_start;
+ void *buff_start;
skb_frag_t *frag;
dma_addr_t addr;
size_t frag_len;
- void *sgt_buf;
-
- /* get a page frag to store the SGTable */
- sz = SKB_DATA_ALIGN(priv->tx_headroom + DPAA_SGT_SIZE);
- sgt_buf = netdev_alloc_frag(sz);
- if (unlikely(!sgt_buf)) {
- netdev_err(net_dev, "netdev_alloc_frag() failed for size %d\n",
- sz);
+ struct page *p;
+ int i, j, err;
+
+ /* get a page to store the SGTable */
+ p = dev_alloc_pages(0);
+ if (unlikely(!p)) {
+ netdev_err(net_dev, "dev_alloc_pages() failed\n");
return -ENOMEM;
}
+ buff_start = page_address(p);
/* Enable L3/L4 hardware checksum computation.
*
@@ -1907,7 +1906,7 @@ static int skb_to_sg_fd(struct dpaa_priv *priv,
* need to write into the skb.
*/
err = dpaa_enable_tx_csum(priv, skb, fd,
- sgt_buf + DPAA_TX_PRIV_DATA_SIZE);
+ buff_start + DPAA_TX_PRIV_DATA_SIZE);
if (unlikely(err < 0)) {
if (net_ratelimit())
netif_err(priv, tx_err, net_dev, "HW csum error: %d\n",
@@ -1916,7 +1915,7 @@ static int skb_to_sg_fd(struct dpaa_priv *priv,
}
/* SGT[0] is used by the linear part */
- sgt = (struct qm_sg_entry *)(sgt_buf + priv->tx_headroom);
+ sgt = (struct qm_sg_entry *)(buff_start + priv->tx_headroom);
frag_len = skb_headlen(skb);
qm_sg_entry_set_len(&sgt[0], frag_len);
sgt[0].bpid = FSL_DPAA_BPID_INV;
@@ -1954,15 +1953,15 @@ static int skb_to_sg_fd(struct dpaa_priv *priv,
/* Set the final bit in the last used entry of the SGT */
qm_sg_entry_set_f(&sgt[nr_frags], frag_len);
+ /* set fd offset to priv->tx_headroom */
qm_fd_set_sg(fd, priv->tx_headroom, skb->len);
/* DMA map the SGT page */
- buffer_start = (void *)sgt - priv->tx_headroom;
- skbh = (struct sk_buff **)buffer_start;
+ skbh = (struct sk_buff **)buff_start;
*skbh = skb;
- addr = dma_map_single(priv->tx_dma_dev, buffer_start,
- priv->tx_headroom + DPAA_SGT_SIZE, dma_dir);
+ addr = dma_map_page(priv->tx_dma_dev, p, 0,
+ priv->tx_headroom + DPAA_SGT_SIZE, dma_dir);
if (unlikely(dma_mapping_error(priv->tx_dma_dev, addr))) {
netdev_err(priv->net_dev, "DMA mapping failed");
err = -EINVAL;
@@ -1982,7 +1981,7 @@ static int skb_to_sg_fd(struct dpaa_priv *priv,
qm_sg_entry_get_len(&sgt[j]), dma_dir);
sg0_map_failed:
csum_failed:
- skb_free_frag(sgt_buf);
+ free_pages((unsigned long)buff_start, 0);
return err;
}
--
2.1.0
DMA unmapping is required before accessing the HW provided timestamping
information.
Fixes: 4664856e9ca2 ("dpaa_eth: add support for hardware timestamping")
Signed-off-by: Madalin Bucur <[email protected]>
---
drivers/net/ethernet/freescale/dpaa/dpaa_eth.c | 32 ++++++++++++++------------
1 file changed, 17 insertions(+), 15 deletions(-)
diff --git a/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c b/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c
index ee267918aa70..c3f19485739b 100644
--- a/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c
+++ b/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c
@@ -1591,18 +1591,6 @@ static struct sk_buff *dpaa_cleanup_tx_fd(const struct dpaa_priv *priv,
skbh = (struct sk_buff **)phys_to_virt(addr);
skb = *skbh;
- if (priv->tx_tstamp && skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) {
- memset(&shhwtstamps, 0, sizeof(shhwtstamps));
-
- if (!fman_port_get_tstamp(priv->mac_dev->port[TX], (void *)skbh,
- &ns)) {
- shhwtstamps.hwtstamp = ns_to_ktime(ns);
- skb_tstamp_tx(skb, &shhwtstamps);
- } else {
- dev_warn(dev, "fman_port_get_tstamp failed!\n");
- }
- }
-
if (unlikely(qm_fd_get_format(fd) == qm_fd_sg)) {
nr_frags = skb_shinfo(skb)->nr_frags;
dma_unmap_single(priv->tx_dma_dev, addr,
@@ -1625,14 +1613,28 @@ static struct sk_buff *dpaa_cleanup_tx_fd(const struct dpaa_priv *priv,
dma_unmap_page(priv->tx_dma_dev, qm_sg_addr(&sgt[i]),
qm_sg_entry_get_len(&sgt[i]), dma_dir);
}
-
- /* Free the page frag that we allocated on Tx */
- skb_free_frag(phys_to_virt(addr));
} else {
dma_unmap_single(priv->tx_dma_dev, addr,
skb_tail_pointer(skb) - (u8 *)skbh, dma_dir);
}
+ /* DMA unmapping is required before accessing the HW provided info */
+ if (priv->tx_tstamp && skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) {
+ memset(&shhwtstamps, 0, sizeof(shhwtstamps));
+
+ if (!fman_port_get_tstamp(priv->mac_dev->port[TX], (void *)skbh,
+ &ns)) {
+ shhwtstamps.hwtstamp = ns_to_ktime(ns);
+ skb_tstamp_tx(skb, &shhwtstamps);
+ } else {
+ dev_warn(dev, "fman_port_get_tstamp failed!\n");
+ }
+ }
+
+ if (qm_fd_get_format(fd) == qm_fd_sg)
+ /* Free the page frag that we allocated on Tx */
+ skb_free_frag(phys_to_virt(addr));
+
return skb;
}
--
2.1.0
From: Laurentiu Tudor <[email protected]>
The liodn base registers are specific to PAMU based NXP systems and are
reserved on SMMU based ones. Don't access them unless PAMU is compiled in.
Signed-off-by: Laurentiu Tudor <[email protected]>
Signed-off-by: Madalin Bucur <[email protected]>
---
drivers/net/ethernet/freescale/fman/fman.c | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/drivers/net/ethernet/freescale/fman/fman.c b/drivers/net/ethernet/freescale/fman/fman.c
index 210749bf1eac..934111def0be 100644
--- a/drivers/net/ethernet/freescale/fman/fman.c
+++ b/drivers/net/ethernet/freescale/fman/fman.c
@@ -634,6 +634,9 @@ static void set_port_liodn(struct fman *fman, u8 port_id,
{
u32 tmp;
+ iowrite32be(liodn_ofst, &fman->bmi_regs->fmbm_spliodn[port_id - 1]);
+ if (!IS_ENABLED(CONFIG_FSL_PAMU))
+ return;
/* set LIODN base for this port */
tmp = ioread32be(&fman->dma_regs->fmdmplr[port_id / 2]);
if (port_id % 2) {
@@ -644,7 +647,6 @@ static void set_port_liodn(struct fman *fman, u8 port_id,
tmp |= liodn_base << DMA_LIODN_SHIFT;
}
iowrite32be(tmp, &fman->dma_regs->fmdmplr[port_id / 2]);
- iowrite32be(liodn_ofst, &fman->bmi_regs->fmbm_spliodn[port_id - 1]);
}
static void enable_rams_ecc(struct fman_fpm_regs __iomem *fpm_rg)
@@ -1942,6 +1944,8 @@ static int fman_init(struct fman *fman)
fman->liodn_offset[i] =
ioread32be(&fman->bmi_regs->fmbm_spliodn[i - 1]);
+ if (!IS_ENABLED(CONFIG_FSL_PAMU))
+ continue;
liodn_base = ioread32be(&fman->dma_regs->fmdmplr[i / 2]);
if (i % 2) {
/* FMDM_PLR LSB holds LIODN base for odd ports */
--
2.1.0
On Tue, 8 Oct 2019 15:10:21 +0300, Madalin Bucur wrote:
> Here's a series of fixes and changes for the DPAA 1.x drivers.
> Fixing some boot time dependency issues, removing some dead code,
> changing the buffers used for reception, fixing the DMA devices,
> some cleanups.
Hi Madalin!
The title of this series says "DPAA fixes", are these bug fixes?
If so they should target the net tree ([PATCH net]) and contain
appropriate Fixes tags.
Cleanups should go into the net-next tree ([PATCH net-next]).
Please try to not post more than 15 patches in one series, it clogs
up the review flow.
For some of those and other best practices please see:
https://www.kernel.org/doc/html/latest/networking/netdev-FAQ.html
https://www.kernel.org/doc/html/latest/process/submitting-patches.html
:)
IMHO there is also no need at all to CC linux-kernel on networking patches..
> -----Original Message-----
> From: Jakub Kicinski <[email protected]>
> Sent: Wednesday, October 9, 2019 7:02 AM
> Subject: Re: [PATCH 00/20] DPAA fixes
>
> On Tue, 8 Oct 2019 15:10:21 +0300, Madalin Bucur wrote:
> > Here's a series of fixes and changes for the DPAA 1.x drivers.
> > Fixing some boot time dependency issues, removing some dead code,
> > changing the buffers used for reception, fixing the DMA devices,
> > some cleanups.
>
> Hi Madalin!
>
> The title of this series says "DPAA fixes", are these bug fixes?
> If so they should target the net tree ([PATCH net]) and contain
> appropriate Fixes tags.
>
> Cleanups should go into the net-next tree ([PATCH net-next]).
>
> Please try to not post more than 15 patches in one series, it clogs
> up the review flow.
>
> For some of those and other best practices please see:
>
> https://www.kernel.org/doc/html/latest/process/submitting-patches.html
> https://www.kernel.org/doc/html/latest/networking/netdev-FAQ.html
>
> :)
>
> IMHO there is also no need at all to CC linux-kernel on networking
> patches..
Hi Jakub,
I should have added the net-next target in the subject.
I can split the (minor) fixes from the other changes.
I've added "Fixes" tags to the fix patches already.
In regards to the patch count, I can send two patch
sets but it won't reduce the total...
Regards,
Madalin