net: emac: emac gigabit ethernet controller driver
Subsystems (5)
PATCH 1/1 net: emac: emac gigabit ethernet controller driver
Add support for ethernet controller HW on Qualcomm Technologies, Inc. SoC. This driver supports the following features:
1) Receive Side Scaling (RSS).
2) Checksum offload.
3) Runtime power management support.
4) Interrupt coalescing support.
5) SGMII phy.
6) SGMII direct connection without external phy.
Based on a driver by Niranjana Vishwanathapura
<[email protected]>.
--- .../devicetree/bindings/net/qcom-emac.txt | 80 + drivers/net/ethernet/qualcomm/Kconfig | 7 + drivers/net/ethernet/qualcomm/Makefile | 2 + drivers/net/ethernet/qualcomm/emac/Makefile | 7 + drivers/net/ethernet/qualcomm/emac/emac-mac.c | 2267 ++++++++++++++++++++ drivers/net/ethernet/qualcomm/emac/emac-mac.h | 341 +++ drivers/net/ethernet/qualcomm/emac/emac-phy.c | 527 +++++ drivers/net/ethernet/qualcomm/emac/emac-phy.h | 73 + drivers/net/ethernet/qualcomm/emac/emac-sgmii.c | 693 ++++++ drivers/net/ethernet/qualcomm/emac/emac-sgmii.h | 30 + drivers/net/ethernet/qualcomm/emac/emac.c | 1324 ++++++++++++ drivers/net/ethernet/qualcomm/emac/emac.h | 427 ++++ 12 files changed, 5778 insertions(+) create mode 100644 Documentation/devicetree/bindings/net/qcom-emac.txt create mode 100644 drivers/net/ethernet/qualcomm/emac/Makefile create mode 100644 drivers/net/ethernet/qualcomm/emac/emac-mac.c create mode 100644 drivers/net/ethernet/qualcomm/emac/emac-mac.h create mode 100644 drivers/net/ethernet/qualcomm/emac/emac-phy.c create mode 100644 drivers/net/ethernet/qualcomm/emac/emac-phy.h create mode 100644 drivers/net/ethernet/qualcomm/emac/emac-sgmii.c create mode 100644 drivers/net/ethernet/qualcomm/emac/emac-sgmii.h create mode 100644 drivers/net/ethernet/qualcomm/emac/emac.c create mode 100644 drivers/net/ethernet/qualcomm/emac/emac.h
Documentation/devicetree/bindings/net/qcom-emac.txt new file +80 -0
@@ -0,0 +1,80 @@ +Qualcomm EMAC Gigabit Ethernet Controller + +Required properties: +- cell-index : EMAC controller instance number. +- compatible : Should be "qcom,emac". +- reg : Offset and length of the register regions for the device +- reg-names : Register region names referenced in 'reg' above. + Required register resource entries are: + "base" : EMAC controller base register block. + "csr" : EMAC wrapper register block. + Optional register resource entries are: + "ptp" : EMAC PTP (1588) register block. + Required if 'qcom,emac-tstamp-en' is present. + "sgmii" : EMAC SGMII PHY register block. +- interrupts : Interrupt numbers used by this controller +- interrupt-names : Interrupt resource names referenced in 'interrupts' above. + Required interrupt resource entries are: + "core0_irq" : EMAC core0 interrupt. + "sgmii_irq" : EMAC SGMII interrupt. + Optional interrupt resource entries are: + "core1_irq" : EMAC core1 interrupt. + "core2_irq" : EMAC core2 interrupt. + "core3_irq" : EMAC core3 interrupt. + "wol_irq" : EMAC Wake-On-LAN (WOL) interrupt. Required if WOL is used. +- qcom,emac-gpio-mdc : GPIO pin number of the MDC line of MDIO bus. +- qcom,emac-gpio-mdio : GPIO pin number of the MDIO line of MDIO bus. +- phy-addr : Specifies phy address on MDIO bus. + Required if the optional property "qcom,no-external-phy" + is not specified. + +Optional properties: +- qcom,emac-tstamp-en : Enables the PTP (1588) timestamping feature. + Include this only if PTP (1588) timestamping + feature is needed. If included, "ptp" register + base should be specified. +- mac-address : The 6-byte MAC address. If present, it is the + default MAC address. +- qcom,no-external-phy : Indicates there is no external PHY connected to + EMAC. Include this only if the EMAC is directly + connected to the peer end without EPHY. +- qcom,emac-ptp-grandmaster : Enable the PTP (1588) grandmaster mode. + Include this only if PTP (1588) is configured as + grandmaster. +- qcom,emac-ptp-frac-ns-adj : The vector table to adjust the fractional ns per + RTC clock cycle. + Include this only if there is accuracy loss of + fractional ns per RTC clock cycle. For individual + table entry, the first field indicates the RTC + reference clock rate. The second field indicates + the number of adjustment in 2 ^ -26 ns. +Example: + emac0: qcom,emac@feb20000 { + cell-index = <0>; + compatible = "qcom,emac"; + reg-names = "base", "csr", "ptp", "sgmii"; + reg = <0xfeb20000 0x10000>, + <0xfeb36000 0x1000>, + <0xfeb3c000 0x4000>, + <0xfeb38000 0x400>; + #address-cells = <0>; + interrupt-parent = <&emac0>; + #interrupt-cells = <1>; + interrupts = <0 1 2 3 4 5>; + interrupt-map-mask = <0xffffffff>; + interrupt-map = <0 &intc 0 76 0 + 1 &intc 0 77 0 + 2 &intc 0 78 0 + 3 &intc 0 79 0 + 4 &intc 0 80 0>; + interrupt-names = "core0_irq", + "core1_irq", + "core2_irq", + "core3_irq", + "sgmii_irq"; + qcom,emac-gpio-mdc = <&msmgpio 123 0>; + qcom,emac-gpio-mdio = <&msmgpio 124 0>; + qcom,emac-tstamp-en; + qcom,emac-ptp-frac-ns-adj = <125000000 1>; + phy-addr = <0>; + };
drivers/net/ethernet/qualcomm/Kconfig +7 -0
@@ -24,4 +24,11 @@ config QCA7000 To compile this driver as a module, choose M here. The module will be called qcaspi. +config QCOM_EMAC + tristate "MSM EMAC Gigabit Ethernet support" + default n + select CRC32 + ---help--- + This driver supports the Qualcomm EMAC Gigabit Ethernet controller. + endif # NET_VENDOR_QUALCOMM
drivers/net/ethernet/qualcomm/Makefile +2 -0
@@ -4,3 +4,5 @@ obj-$(CONFIG_QCA7000) += qcaspi.o qcaspi-objs := qca_spi.o qca_framing.o qca_7k.o qca_debug.o + +obj-$(CONFIG_QCOM_EMAC) += emac/
drivers/net/ethernet/qualcomm/emac/Makefile new file +7 -0
@@ -0,0 +1,7 @@ +# +# Makefile for the Qualcomm Technologies Inc EMAC Gigabit Ethernet driver +# + +obj-$(CONFIG_QCOM_EMAC) += qcom-emac.o + +qcom-emac-objs := emac.o emac-mac.o emac-phy.o emac-sgmii.o
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +2267 -0
2,267 lines not on this page · raw message
drivers/net/ethernet/qualcomm/emac/emac-mac.h new file +341 -0
341 lines not on this page · raw message
drivers/net/ethernet/qualcomm/emac/emac-phy.c new file +527 -0
288 lines not on this page · raw message
drivers/net/ethernet/qualcomm/emac/emac-phy.h new file +73 -0
73 lines not on this page · raw message
drivers/net/ethernet/qualcomm/emac/emac-sgmii.c new file +693 -0
693 lines not on this page · raw message
drivers/net/ethernet/qualcomm/emac/emac-sgmii.h new file +30 -0
Lines of this file are not shown. View the raw message.
drivers/net/ethernet/qualcomm/emac/emac.c new file +1324 -0
1,324 lines not on this page · raw message
drivers/net/ethernet/qualcomm/emac/emac.h new file +427 -0
427 lines not on this page · raw message
The rest of this diff is not shown. View the raw message.
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by The Linux Foundation
Part of the diff is not shown here. View the raw message.
drivers/net/ethernet/qualcomm/emac/emac-mac.h line 66
Quoted 10 lines · Gilad Avidov
drivers/net/ethernet/qualcomm/emac/emac-mac.h new file +10
-0+/* RRD (Receive Return Descriptor) */ +union emac_rrd { +struct { +/* 32bit word 0 */ +u32 xsum:16 ; +u32 nor:4 ;/* number of RFD */ +u32 si:12 ;/* start index of rfd-ring */ +/* 32bit word 1 */ +u32 hash; +/* 32bit word 2 */
You should never use bitfields for hardware structs.
I think in general, kernel code should be made endian safe, even if you only care about one particular endian type for your platform.
- Felix
Quoted 15 lines · On Tue, 8 Dec 2015 00:33:04 +0100 Felix Fietkau <[email protected]> wrote:
drivers/net/ethernet/qualcomm/emac/emac-mac.h line 66
Quoted 10 lines · Gilad Avidov
drivers/net/ethernet/qualcomm/emac/emac-mac.h new file +10
-0+/* RRD (Receive Return Descriptor) */ +union emac_rrd { +struct { +/* 32bit word 0 */ +u32 xsum:16 ; +u32 nor:4 ;/* number of RFD */ +u32 si:12 ;/* start index of rfd-ring */ +/* 32bit word 1 */ +u32 hash; +/* 32bit word 2 */ You should never use bitfields for hardware structs.
I think in general, kernel code should be made endian safe, even if you only care about one particular endian type for your platform.
Quoted 1 line
- Felix
Thank you Felix,
I will change the bit fields to bitwise operations and macros.
Gilad
kbuild test robot
build botreproduce: wget https://git.kernel.org/cgit/linux/kernel/git/wfg/lkp-tests.git/plain/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross # save the attached .config to linux build tree…
Hi Gilad,
[auto build test ERROR on net/master]
[also build test ERROR on v4.4-rc4 next-20151207]
url: https://github.com/0day-ci/linux/commits/Gilad-Avidov/net-emac-emac-gigabit-ethernet-controller-driver/20151208-070026 config: openrisc-allyesconfig (attached as .config)
reproduce:
wget https://git.kernel.org/cgit/linux/kernel/git/wfg/lkp-tests.git/plain/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross
# save the attached .config to linux build tree
make.cross ARCH=openrisc
All errors (new ones prefixed by >>):
drivers/net/ethernet/qualcomm/emac/emac-mac.c: In function 'emac_tso_csum':
>> drivers/net/ethernet/qualcomm/emac/emac-mac.c:2086:4: error: implicit declaration of function 'csum_ipv6_magic'
vim +/csum_ipv6_magic +2086 drivers/net/ethernet/qualcomm/emac/emac-mac.c
2080 union emac_tpd extra_tpd;
2081
2082 memset(tpd, 0, sizeof(*tpd));
2083 memset(&extra_tpd, 0, sizeof(extra_tpd));
2084
2085 ipv6_hdr(skb)->payload_len = 0;
> 2086 tcp_hdr(skb)->check = ~csum_ipv6_magic(
2087 &ipv6_hdr(skb)->saddr,
2088 &ipv6_hdr(skb)->daddr,
2089 0, IPPROTO_TCP, 0);
---
0-DAY kernel test infrastructure Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all Intel Corporation
So first of all, thanks for posting this. I know it's missing a bunch of stuff that's necessary for Qualcomm's Server chip, but it's a start.
Unfortunately, 6,000 lines is a lot to review at once. Any chance you can break up the next version into smaller patches?
Documentation/devicetree/bindings/net/qcom-emac.txt line 76
Quoted 2 lines · Gilad Avidov
Documentation/devicetree/bindings/net/qcom-emac.txt new file +2
-0+ qcom,emac-gpio-mdc = <&msmgpio 123 0>; + qcom,emac-gpio-mdio = <&msmgpio 124 0>;
Is there any chance you can remove all references to "MSM" throughout the entire driver, since the EMAC exists on non-MSM parts?
Quoted 9 lines
Documentation/devicetree/bindings/net/qcom-emac.txt new file +4
-0+ qcom,emac-tstamp-en; + qcom,emac-ptp-frac-ns-adj = <125000000 1>; + phy-addr = <0>; + };drivers/net/ethernet/qualcomm/Kconfig +0
-0
drivers/net/ethernet/qualcomm/Kconfig config QCA7000 · line 29
Quoted 6 lines
drivers/net/ethernet/qualcomm/Kconfig +3
-0@@ -24,4 +24,11 @@ config QCA7000 To compile this driver as a module, choose M here. The module will be called qcaspi. +config QCOM_EMAC + tristate "MSM EMAC Gigabit Ethernet support" + default n
"default n" is redundant
drivers/net/ethernet/qualcomm/Kconfig config QCA7000 · line 32
Quoted 3 lines
drivers/net/ethernet/qualcomm/Kconfig +3
-0+ select CRC32 + ---help--- + This driver supports the Qualcomm EMAC Gigabit Ethernet controller.
I think should be longer, perhaps by adding some more info about the controller itself?
drivers/net/ethernet/qualcomm/Makefile line 8
Quoted 13 lines
drivers/net/ethernet/qualcomm/Kconfig +1
-0+ endif # NET_VENDOR_QUALCOMMdrivers/net/ethernet/qualcomm/Makefile +2
-0@@ -4,3 +4,5 @@ obj-$(CONFIG_QCA7000) += qcaspi.o qcaspi-objs := qca_spi.o qca_framing.o qca_7k.o qca_debug.o + +obj-$(CONFIG_QCOM_EMAC) += emac/
Please fix
drivers/net/ethernet/qualcomm/Makefile
Quoted 5 lines
drivers/net/ethernet/qualcomm/Makefile +5
-0+/* RSS */ +static void emac_mac_rss_config(struct emac_adapter *adpt) +{ + int key_len_by_u32 = sizeof(adpt->rss_key) / sizeof(u32); + int idt_len_by_u32 = sizeof(adpt->rss_idt) / sizeof(u32);
Can you use ARRAY_SIZE here?
Quoted 40 lines
drivers/net/ethernet/qualcomm/Makefile +40
-0+ u32 rxq0; + int i; + + /* Fill out hash function keys */ + for (i = 0; i < key_len_by_u32; i++) { + u32 key, idx_base; + + idx_base = (key_len_by_u32 - i) * 4; + key = ((adpt->rss_key[idx_base - 1]) | + (adpt->rss_key[idx_base - 2] << 8) | + (adpt->rss_key[idx_base - 3] << 16) | + (adpt->rss_key[idx_base - 4] << 24)); + writel_relaxed(key, adpt->base + EMAC_RSS_KEY(i, u32)); + } + + /* Fill out redirection table */ + for (i = 0; i < idt_len_by_u32; i++) + writel_relaxed(adpt->rss_idt[i], + adpt->base + EMAC_RSS_TBL(i, u32)); + + writel_relaxed(adpt->rss_base_cpu, adpt->base + EMAC_BASE_CPU_NUMBER); + + rxq0 = readl_relaxed(adpt->base + EMAC_RXQ_CTRL_0); + if (adpt->rss_hstype & EMAC_RSS_HSTYP_IPV4_EN) + rxq0 |= RXQ0_RSS_HSTYP_IPV4_EN; + else + rxq0 &= ~RXQ0_RSS_HSTYP_IPV4_EN; + + if (adpt->rss_hstype & EMAC_RSS_HSTYP_TCP4_EN) + rxq0 |= RXQ0_RSS_HSTYP_IPV4_TCP_EN; + else + rxq0 &= ~RXQ0_RSS_HSTYP_IPV4_TCP_EN; + + if (adpt->rss_hstype & EMAC_RSS_HSTYP_IPV6_EN) + rxq0 |= RXQ0_RSS_HSTYP_IPV6_EN; + else + rxq0 &= ~RXQ0_RSS_HSTYP_IPV6_EN; + + if (adpt->rss_hstype & EMAC_RSS_HSTYP_TCP6_EN) + rxq0 |= RXQ0_RSS_HSTYP_IPV6_TCP_EN;
drivers/net/ethernet/qualcomm/Makefile
Quoted 10 lines
drivers/net/ethernet/qualcomm/Makefile +10
-0+ else + rxq0 &= ~RXQ0_RSS_HSTYP_IPV6_TCP_EN; + + rxq0 |= ((adpt->rss_idt_size << IDT_TABLE_SIZE_SHFT) & + IDT_TABLE_SIZE_BMSK); + rxq0 |= RSS_HASH_EN; + + wmb(); /* ensure all parameters are written before enabling RSS */ + + writel_relaxed(rxq0, adpt->base + EMAC_RXQ_CTRL_0);
Why not just use writel(), which already includes a wmb()
Quoted 7 lines
drivers/net/ethernet/qualcomm/Makefile +7
-0+/* Power Management */ +void emac_mac_pm(struct emac_adapter *adpt, u32 speed, bool wol_en, bool rx_en) +{ + u32 dma_mas, mac; + + dma_mas = readl_relaxed(adpt->base + EMAC_DMA_MAS_CTRL); + dma_mas &= ~LPW_CLK_SEL;
drivers/net/ethernet/qualcomm/Makefile
Quoted 10 lines
drivers/net/ethernet/qualcomm/Makefile +10
-0+ dma_mas |= LPW_STATE; + + mac = readl_relaxed(adpt->base + EMAC_MAC_CTRL); + mac &= ~(FULLD | RXEN | TXEN); + mac = (mac & ~SPEED_BMSK) | + (((u32)emac_mac_speed_10_100 << SPEED_SHFT) & SPEED_BMSK); + + if (wol_en) { + if (rx_en) + mac |= (RXEN | BROAD_EN);
You don't need the parentheses.
drivers/net/ethernet/qualcomm/Makefile
Quoted 8 lines
drivers/net/ethernet/qualcomm/Makefile +8
-0+/* Config descriptor rings */ +static void emac_mac_dma_rings_config(struct emac_adapter *adpt) +{ + if (adpt->timestamp_en) + emac_reg_update32(adpt->csr + EMAC_EMAC_WRAPPER_CSR1, + 0, ENABLE_RRD_TIMESTAMP); + + /* TPD */
What does TPD stand for?
Quoted 15 lines
drivers/net/ethernet/qualcomm/Makefile +15
-0+ writel_relaxed(EMAC_DMA_ADDR_HI(adpt->tx_q[0].tpd.p_addr), + adpt->base + EMAC_DESC_CTRL_1); + switch (adpt->tx_q_cnt) { + case 4: + writel_relaxed(EMAC_DMA_ADDR_LO(adpt->tx_q[3].tpd.p_addr), + adpt->base + EMAC_H3TPD_BASE_ADDR_LO); + /* fall through */ + case 3: + writel_relaxed(EMAC_DMA_ADDR_LO(adpt->tx_q[2].tpd.p_addr), + adpt->base + EMAC_H2TPD_BASE_ADDR_LO); + /* fall through */ + case 2: + writel_relaxed(EMAC_DMA_ADDR_LO(adpt->tx_q[1].tpd.p_addr), + adpt->base + EMAC_H1TPD_BASE_ADDR_LO); + /* fall through */
drivers/net/ethernet/qualcomm/Makefile
Quoted 10 lines
drivers/net/ethernet/qualcomm/Makefile +10
-0+ case 1: + writel_relaxed(EMAC_DMA_ADDR_LO(adpt->tx_q[0].tpd.p_addr), + adpt->base + EMAC_DESC_CTRL_8); + break; + default: + netdev_err(adpt->netdev, + "error: invalid number of TX queues (%d)\n", + adpt->tx_q_cnt); + return; + }
This is not time-critical code. Why not just create a for-loop?
drivers/net/ethernet/qualcomm/Makefile
Quoted 7 lines
drivers/net/ethernet/qualcomm/Makefile +7
-0+/* Config transmit parameters */ +static void emac_mac_tx_config(struct emac_adapter *adpt) +{ + u16 tx_offload_thresh = EMAC_MAX_TX_OFFLOAD_THRESH; + u32 val; + + writel_relaxed((tx_offload_thresh >> 3) &
Why is tx_offload_thresh a u16 if you're going to use writel anyway? Make it a u32.
drivers/net/ethernet/qualcomm/Makefile
Quoted 10 lines
drivers/net/ethernet/qualcomm/Makefile +10
-0+void emac_mac_reset(struct emac_adapter *adpt) +{ + writel_relaxed(0, adpt->base + EMAC_INT_MASK); + writel_relaxed(DIS_INT, adpt->base + EMAC_INT_STATUS); + + emac_mac_stop(adpt); + + emac_reg_update32(adpt->base + EMAC_DMA_MAS_CTRL, 0, SOFT_RST); + wmb(); /* ensure mac is fully reset */ + usleep_range(100, 150);
Please add a comment explaiing why this delay is necessary.
drivers/net/ethernet/qualcomm/Makefile
Quoted 7 lines
drivers/net/ethernet/qualcomm/Makefile +7
-0+void emac_mac_stop(struct emac_adapter *adpt) +{ + emac_reg_update32(adpt->base + EMAC_RXQ_CTRL_0, RXQ_EN, 0); + emac_reg_update32(adpt->base + EMAC_TXQ_CTRL_0, TXQ_EN, 0); + emac_reg_update32(adpt->base + EMAC_MAC_CTRL, (TXEN | RXEN), 0); + wmb(); /* ensure mac is stopped before we proceed */ + usleep_range(1000, 1050);
Same here.
drivers/net/ethernet/qualcomm/Makefile
Quoted 8 lines
drivers/net/ethernet/qualcomm/Makefile +8
-0+/* set MAC address */ +void emac_mac_addr_clear(struct emac_adapter *adpt, u8 *addr) +{ + u32 sta; + + /* for example: 00-A0-C6-11-22-33 + * 0<-->C6112233, 1<-->00A0. + */
/*
* Multi-line comments
* look like this.
*/
drivers/net/ethernet/qualcomm/Makefile
Quoted 6 lines
drivers/net/ethernet/qualcomm/Makefile +6
-0+/* Free all descriptors of given transmit queue */ +static void emac_tx_q_descs_free(struct emac_adapter *adpt, + struct emac_tx_queue *tx_q) +{ + unsigned long size; + u32 i;
Since 'i' is used as an index, it should be an unsized integer. And 'size' should be a 'size_t'
drivers/net/ethernet/qualcomm/Makefile
Quoted 3 lines
drivers/net/ethernet/qualcomm/Makefile +3
-0+static void emac_tx_q_descs_free_all(struct emac_adapter *adpt) +{ + u8 i;
'int'. Personally, I'd prefer "unsigned int", but 'int' is what you use elsewhere.
drivers/net/ethernet/qualcomm/Makefile
Quoted 7 lines
drivers/net/ethernet/qualcomm/Makefile +7
-0+/* Free all descriptors of given receive queue */ +static void emac_rx_q_free_descs(struct emac_adapter *adpt, + struct emac_rx_queue *rx_q) +{ + struct device *dev = adpt->netdev->dev.parent; + unsigned long size; + u32 i;
size_t size;
int i;
drivers/net/ethernet/qualcomm/Makefile
Quoted 6 lines
drivers/net/ethernet/qualcomm/Makefile +6
-0+/* Allocate TX descriptor ring for the given transmit queue */ +static int emac_tx_q_desc_alloc(struct emac_adapter *adpt, + struct emac_tx_queue *tx_q) +{ + struct emac_ring_header *ring_header = &adpt->ring_header; + unsigned long size;
size_t
drivers/net/ethernet/qualcomm/Makefile
Quoted 13 lines
drivers/net/ethernet/qualcomm/Makefile +13
-0+ + size = sizeof(struct emac_buffer) * tx_q->tpd.count; + tx_q->tpd.tpbuff = kzalloc(size, GFP_KERNEL); + if (!tx_q->tpd.tpbuff) + return -ENOMEM; + + tx_q->tpd.size = tx_q->tpd.count * (adpt->tpd_size * 4); + tx_q->tpd.p_addr = ring_header->p_addr + ring_header->used; + tx_q->tpd.v_addr = ring_header->v_addr + ring_header->used; + ring_header->used += ALIGN(tx_q->tpd.size, 8); + tx_q->tpd.produce_idx = 0; + tx_q->tpd.consume_idx = 0; + return 0;
blank line above "return".
drivers/net/ethernet/qualcomm/Makefile
Quoted 6 lines
drivers/net/ethernet/qualcomm/Makefile +6
-0+} + +static int emac_tx_q_desc_alloc_all(struct emac_adapter *adpt) +{ + int retval = 0; + u8 i;
int i;
drivers/net/ethernet/qualcomm/Makefile
Quoted 3 lines
drivers/net/ethernet/qualcomm/Makefile +3
-0+static void emac_rx_q_free_bufs_all(struct emac_adapter *adpt) +{ + u8 i;
int i;
drivers/net/ethernet/qualcomm/Makefile
Quoted 6 lines
drivers/net/ethernet/qualcomm/Makefile +6
-0+/* Allocate RX descriptor rings for the given receive queue */ +static int emac_rx_descs_alloc(struct emac_adapter *adpt, + struct emac_rx_queue *rx_q) +{ + struct emac_ring_header *ring_header = &adpt->ring_header; + unsigned long size;
size_t
drivers/net/ethernet/qualcomm/Makefile
Quoted 4 lines
drivers/net/ethernet/qualcomm/Makefile +4
-0+static int emac_rx_descs_allocs_all(struct emac_adapter *adpt) +{ + int retval = 0; + u8 i;
int i;
drivers/net/ethernet/qualcomm/Makefile
Quoted 9 lines
drivers/net/ethernet/qualcomm/Makefile +9
-0+ + for (i = 0; i < adpt->rx_q_cnt; i++) { + retval = emac_rx_descs_alloc(adpt, &adpt->rx_q[i]); + if (retval) + break; + } + + if (retval) { + netdev_err(adpt->netdev, "error: Rx Queue %u alloc failed\n",
%d
Quoted 5 lines
drivers/net/ethernet/qualcomm/Makefile +5
-0+/* Produce new receive free descriptor */ +static bool emac_mac_rx_rfd_create(struct emac_adapter *adpt, + struct emac_rx_queue *rx_q, + union emac_rfd *rfd) +{
drivers/net/ethernet/qualcomm/Makefile
Quoted 10 lines
drivers/net/ethernet/qualcomm/Makefile +10
-0+ u32 *hw_rfd = EMAC_RFD(rx_q, adpt->rfd_size, + rx_q->rfd.produce_idx); + + *(hw_rfd++) = rfd->word[0]; + *hw_rfd = rfd->word[1]; + + if (++rx_q->rfd.produce_idx == rx_q->rfd.count) + rx_q->rfd.produce_idx = 0; + + return true;
You never check the return value, so just make this a void function.
drivers/net/ethernet/qualcomm/Makefile
Quoted 12 lines
drivers/net/ethernet/qualcomm/Makefile +12
-0+} + +/* Fill up receive queue's RFD with preallocated receive buffers */ +static int emac_mac_rx_descs_refill(struct emac_adapter *adpt, + struct emac_rx_queue *rx_q) +{ + struct emac_buffer *curr_rxbuf; + struct emac_buffer *next_rxbuf; + union emac_rfd rfd; + struct sk_buff *skb; + void *skb_data = NULL; + u32 count = 0;
int count = 0;
The type should match the return value of the function.
drivers/net/ethernet/qualcomm/Makefile
Quoted 13 lines
drivers/net/ethernet/qualcomm/Makefile +13
-0+ u32 next_produce_idx; + + next_produce_idx = rx_q->rfd.produce_idx; + if (++next_produce_idx == rx_q->rfd.count) + next_produce_idx = 0; + curr_rxbuf = GET_RFD_BUFFER(rx_q, rx_q->rfd.produce_idx); + next_rxbuf = GET_RFD_BUFFER(rx_q, next_produce_idx); + + /* this always has a blank rx_buffer*/ + while (!next_rxbuf->dma) { + skb = dev_alloc_skb(adpt->rxbuf_size + NET_IP_ALIGN); + if (unlikely(!skb)) + break;
I don't think this is time-critical code, so don't use unlikely().
Quoted 25 lines
drivers/net/ethernet/qualcomm/Makefile +25
-0+/* Consume next received packet descriptor */ +static bool emac_rx_process_rrd(struct emac_adapter *adpt, + struct emac_rx_queue *rx_q, + union emac_rrd *rrd) +{ + u32 *hw_rrd = EMAC_RRD(rx_q, adpt->rrd_size, + rx_q->rrd.consume_idx); + + /* If time stamping is enabled, it will be added in the beginning of + * the hw rrd (hw_rrd). In sw rrd (rrd), 32bit words 4 & 5 are reserved + * for the time stamp; hence the conversion. + * Also, read the rrd word with update flag first; read rest of rrd + * only if update flag is set. + */ + if (adpt->timestamp_en) + rrd->word[3] = *(hw_rrd + 5); + else + rrd->word[3] = *(hw_rrd + 3); + rmb(); /* ensure hw receive returned descriptor timestamp is read */ + + if (!rrd->genr.update) + return false; + + if (adpt->timestamp_en) { + rrd->word[4] = *(hw_rrd++);
drivers/net/ethernet/qualcomm/Makefile
Quoted 10 lines
drivers/net/ethernet/qualcomm/Makefile +10
-0+ rrd->word[5] = *(hw_rrd++); + } else { + rrd->word[4] = 0; + rrd->word[5] = 0; + } + + rrd->word[0] = *(hw_rrd++); + rrd->word[1] = *(hw_rrd++); + rrd->word[2] = *(hw_rrd++); + mb(); /* ensure descriptor is read */
Why do you use mb() here, but rmb() above? The comment is the same.
drivers/net/ethernet/qualcomm/Makefile
Quoted 6 lines
drivers/net/ethernet/qualcomm/Makefile +6
-0+static void emac_rx_rfd_clean(struct emac_rx_queue *rx_q, + union emac_rrd *rrd) +{ + struct emac_buffer *rfbuf = rx_q->rfd.rfbuff; + u32 consume_idx = rrd->genr.si; + u16 i;
int i;
drivers/net/ethernet/qualcomm/Makefile
Quoted 6 lines
drivers/net/ethernet/qualcomm/Makefile +6
-0+static inline bool emac_skb_cb_expired(struct sk_buff *skb) +{ + if (time_is_after_jiffies(EMAC_SKB_CB(skb)->jiffies + + msecs_to_jiffies(100))) + return false; + return true;
return time_is_before_jiffies(EMAC_SKB_CB(skb)->jiffies + msecs_to_jiffies(100));
drivers/net/ethernet/qualcomm/Makefile
Quoted 12 lines
drivers/net/ethernet/qualcomm/Makefile +12
-0+/* Process receive event */ +void emac_mac_rx_process(struct emac_adapter *adpt, struct emac_rx_queue *rx_q, + int *num_pkts, int max_pkts) +{ + struct net_device *netdev = adpt->netdev; + + union emac_rrd rrd; + struct emac_buffer *rfbuf; + struct sk_buff *skb; + + u32 hw_consume_idx, num_consume_pkts; + u32 count = 0;
unsigned int count;
Quoted 59 lines
drivers/net/ethernet/qualcomm/Makefile +59
-0+ u32 proc_idx; + u32 reg = readl_relaxed(adpt->base + rx_q->consume_reg); + + hw_consume_idx = (reg & rx_q->consume_mask) >> rx_q->consume_shft; + num_consume_pkts = (hw_consume_idx >= rx_q->rrd.consume_idx) ? + (hw_consume_idx - rx_q->rrd.consume_idx) : + (hw_consume_idx + rx_q->rrd.count - rx_q->rrd.consume_idx); + + while (1) { + if (!num_consume_pkts) + break; + + if (!emac_rx_process_rrd(adpt, rx_q, &rrd)) + break; + + if (likely(rrd.genr.nor == 1)) { + /* good receive */ + rfbuf = GET_RFD_BUFFER(rx_q, rrd.genr.si); + dma_unmap_single(adpt->netdev->dev.parent, rfbuf->dma, + rfbuf->length, DMA_FROM_DEVICE); + rfbuf->dma = 0; + skb = rfbuf->skb; + } else { + netdev_err(adpt->netdev, + "error: multi-RFD not support yet!\n"); + break; + } + emac_rx_rfd_clean(rx_q, &rrd); + num_consume_pkts--; + count++; + + /* Due to a HW issue in L4 check sum detection (UDP/TCP frags + * with DF set are marked as error), drop packets based on the + * error mask rather than the summary bit (ignoring L4F errors) + */ + if (rrd.word[EMAC_RRD_STATS_DW_IDX] & EMAC_RRD_ERROR) { + netif_dbg(adpt, rx_status, adpt->netdev, + "Drop error packet[RRD: 0x%x:0x%x:0x%x:0x%x]\n", + rrd.word[0], rrd.word[1], + rrd.word[2], rrd.word[3]); + + dev_kfree_skb(skb); + continue; + } + + skb_put(skb, rrd.genr.pkt_len - ETH_FCS_LEN); + skb->dev = netdev; + skb->protocol = eth_type_trans(skb, skb->dev); + if (netdev->features & NETIF_F_RXCSUM) + skb->ip_summed = ((rrd.genr.l4f) ? + CHECKSUM_NONE : CHECKSUM_UNNECESSARY); + else + skb_checksum_none_assert(skb); + + if (test_bit(EMAC_STATUS_TS_RX_EN, &adpt->status)) { + struct skb_shared_hwtstamps *hwts = skb_hwtstamps(skb); + + hwts->hwtstamp = ktime_set(rrd.genr.ts_high, + rrd.genr.ts_low);
drivers/net/ethernet/qualcomm/Makefile
Quoted 10 lines
drivers/net/ethernet/qualcomm/Makefile +10
-0+ } + + emac_receive_skb(rx_q, skb, (u16)rrd.genr.cvlan_tag, + (bool)rrd.genr.cvlan_flag); + + netdev->last_rx = jiffies; + (*num_pkts)++; + if (*num_pkts >= max_pkts) + break; + }
How about
do {
...
} while (*num_pkts < max_pkts);
drivers/net/ethernet/qualcomm/Makefile
Quoted 6 lines
drivers/net/ethernet/qualcomm/Makefile +6
-0+/* Check if enough transmit descriptors are available */ +static bool emac_tx_has_enough_descs(struct emac_tx_queue *tx_q, + const struct sk_buff *skb) +{ + u32 num_required = 1; + u16 i;
int i;
Anyway, you got the idea. I think sized integers should be used sparingly, and general counting and index variable should be unsized integers, preferably also unsigned.
Quoted 9 lines
drivers/net/ethernet/qualcomm/Makefile +9
-0+/* Fill up transmit descriptors with TSO and Checksum offload information */ +static int emac_tso_csum(struct emac_adapter *adpt, + struct emac_tx_queue *tx_q, + struct sk_buff *skb, + union emac_tpd *tpd) +{ + u8 hdr_len; + int retval; +
drivers/net/ethernet/qualcomm/Makefile
Quoted 10 lines
drivers/net/ethernet/qualcomm/Makefile +10
-0+ if (skb_is_gso(skb)) { + if (skb_header_cloned(skb)) { + retval = pskb_expand_head(skb, 0, 0, GFP_ATOMIC); + if (unlikely(retval)) + return retval; + } + + if (skb->protocol == htons(ETH_P_IP)) { + u32 pkt_len = + ((unsigned char *)ip_hdr(skb) - skb->data) +
Use void* for pointer math, instead of "unsigned char *".
Quoted 21 lines
drivers/net/ethernet/qualcomm/Makefile +21
-0+/* Transmit the packet using specified transmit queue */ +int emac_mac_tx_buf_send(struct emac_adapter *adpt, struct emac_tx_queue *tx_q, + struct sk_buff *skb) +{ + union emac_tpd tpd; + u32 prod_idx; + + if (test_bit(EMAC_STATUS_DOWN, &adpt->status)) { + dev_kfree_skb_any(skb); + return NETDEV_TX_OK; + } + + if (!emac_tx_has_enough_descs(tx_q, skb)) { + /* not enough descriptors, just stop queue */ + netif_stop_queue(adpt->netdev); + return NETDEV_TX_BUSY; + } + + memset(&tpd, 0, sizeof(tpd)); + + if (emac_tso_csum(adpt, tx_q, skb, &tpd) != 0) {
drivers/net/ethernet/qualcomm/Makefile
Quoted 10 lines
drivers/net/ethernet/qualcomm/Makefile +10
-0+ dev_kfree_skb_any(skb); + return NETDEV_TX_OK; + } + + if (skb_vlan_tag_present(skb)) { + u16 vlan = skb_vlan_tag_get(skb); + u16 tag; + + EMAC_VLAN_TO_TAG(vlan, tag); + tpd.genr.cvlan_tag = tag;
Can't you just do EMAC_VLAN_TO_TAG(vlan, tpd.genr.cvlan_tag);
Quoted 29 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.h new file +23
-0@@ -0,0 +1,341 @@ +/* Copyright (c) 2013-2015, The Linux Foundation. All rights reserved. +* +* This program is free software; you can redistribute it and/or modify +* it under the terms of the GNU General Public License version 2 and +* only version 2 as published by the Free Software Foundation. +* +* This program is distributed in the hope that it will be useful, +* but WITHOUT ANY WARRANTY; without even the implied warranty of +* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +* GNU General Public License for more details. +*/ + +/* EMAC DMA HW engine uses three rings: +* Tx: +* TPD: Transmit Packet Descriptor ring. +* Rx: +* RFD: Receive Free Descriptor ring. +* Ring of descriptors with empty buffers to be filled by Rx HW. +* RRD: Receive Return Descriptor ring. +* Ring of descriptors with buffers filled with received data. +*/ + +#ifndef _EMAC_HW_H_
drivers/net/ethernet/qualcomm/emac/emac-mac.h line 33
Quoted 10 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.h new file +10
-0+#define _EMAC_HW_H_ + +/* EMAC_CSR register offsets */ +#define EMAC_EMAC_WRAPPER_CSR1 0x000000 +#define EMAC_EMAC_WRAPPER_CSR2 0x000004 +#define EMAC_EMAC_WRAPPER_CSR3 0x000008 +#define EMAC_EMAC_WRAPPER_CSR5 0x000010 +#define EMAC_EMAC_WRAPPER_TX_TS_LO 0x000104 +#define EMAC_EMAC_WRAPPER_TX_TS_HI 0x000108 +#define EMAC_EMAC_WRAPPER_TX_TS_INX 0x00010c
Can you move some of the macros into the .c files? For example, I'm pretty sure that the EMAC_EMAC_WRAPPER_CSRx macros are used only in emac-sgmii.c.
Anyway, I'm stopping for now. I'll post more later.
Quoted 5 lines · Timur Tabi
Quoted 5 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +5
-0+/* set MAC address */ +void emac_mac_addr_clear (struct emac_adapter *adpt,u8 *addr) +{ +u32 sta; +
Quoted 8 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 859
Quoted 3 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +3
-0+/* for example: 00-A0-C6-11-22-33 +* 0<-->C6112233, 1<-->00A0. +*/ /*
* Multi-line comments
* look like this.
*/
Except in drivers/net. The convention in drivers/net is to use multi-line format as posted in this patch.
David Miller
maintainerFrom: Fabio Estevam <[email protected]>
Date: Wed, 9 Dec 2015 18:37:35 -0200
Quoted 16 lines · Timur Tabi
Quoted 14 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 859
Quoted 9 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +8
-0+/* set MAC address */ +void emac_mac_addr_clear (struct emac_adapter *adpt,u8 *addr) +{ +u32 sta; + +/* for example: 00-A0-C6-11-22-33 +* 0<-->C6112233, 1<-->00A0. +*/ /*
* Multi-line comments
* look like this.
*/
Quoted 2 lines
Except in
drivers/net. The convention indrivers/netis to use multi-line format as posted in this patch.
Correct.
Thank you Timur for the good review.
Quoted 6 lines · On Wed, 9 Dec 2015 14:09:27 -0600 Timur Tabi <[email protected]> wrote:
So first of all, thanks for posting this. I know it's missing a bunch of stuff that's necessary for Qualcomm's Server chip, but it's a start.
Unfortunately, 6,000 lines is a lot to review at once. Any chance you can break up the next version into smaller patches?
Agree its a lot to review, however:
1) This driver is the what left after I removed all unnecessary features, thus it is already minimal.
I have removed features such as support for: server, ACPI, ethtool, ptp/1588, etc.
2) This size seems comparable to first patches of existing Ethernet drivers.
Quoted 9 lines
Documentation/devicetree/bindings/net/qcom-emac.txt line 76
Quoted 2 lines · Gilad Avidov
Documentation/devicetree/bindings/net/qcom-emac.txt new file +2
-0+ qcom,emac-gpio-mdc = <&msmgpio 123 0>; + qcom,emac-gpio-mdio = <&msmgpio 124 0>;Is there any chance you can remove all references to "MSM" throughout the entire driver, since the EMAC exists on non-MSM parts?
msm appears only in this DT binding example. None in the code. I will look into removing this instance too.
Quoted 19 lines
Quoted 13 lines
Documentation/devicetree/bindings/net/qcom-emac.txt new file +4
-0+ qcom,emac-tstamp-en; + qcom,emac-ptp-frac-ns-adj = <125000000 1>; + phy-addr = <0>; + };(quoted diff) +0
-0b/drivers/net/ethernet/qualcomm/Kconfig index a76e380..ae9442d 100644 --- a/drivers/net/ethernet/qualcomm/Kconfig +++ b/drivers/net/ethernet/qualcomm/Kconfig
(quoted diff) +0
-0@@ -24,4 +24,11 @@ config QCA7000 To compile this driver as a module, choose M here. Themodule will be called qcaspi.
drivers/net/ethernet/qualcomm/Kconfig config QCA7000 · line 29
Quoted 3 lines
drivers/net/ethernet/qualcomm/Kconfig +3
-0+config QCOM_EMAC + tristate "MSM EMAC Gigabit Ethernet support" + default n"default n" is redundant
Quoted 7 lines
Quoted 4 lines
drivers/net/ethernet/qualcomm/Kconfig +3
-0+ select CRC32 + ---help--- + This driver supports the Qualcomm EMAC Gigabit Ethernetcontroller.
I think should be longer, perhaps by adding some more info about the controller itself?
ok.
Quoted 9 lines
Quoted 9 lines
drivers/net/ethernet/qualcomm/Kconfig +1
-0+ endif # NET_VENDOR_QUALCOMM(quoted diff) +0
-0b/drivers/net/ethernet/qualcomm/Makefile index 9da2d75..b14686e 100644 --- a/drivers/net/ethernet/qualcomm/Makefile +++ b/drivers/net/ethernet/qualcomm/Makefile
(quoted diff) +0
-0@@ -4,3 +4,5 @@
Quoted 7 lines
drivers/net/ethernet/qualcomm/Makefile line 8
Quoted 5 lines
drivers/net/ethernet/qualcomm/Makefile +2
-0obj-$(CONFIG_QCA7000) += qcaspi.o qcaspi-objs := qca_spi.o qca_framing.o qca_7k.o qca_debug.o + +obj-$(CONFIG_QCOM_EMAC) += emac/Please fix
ok.
Quoted 8 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 401
Quoted 6 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +5
-0+/* RSS */ +static void emac_mac_rss_config (struct emac_adapter *adpt) +{ +int key_len_by_u32 =sizeof (adpt->rss_key) /sizeof (u32); +int idt_len_by_u32 =sizeof (adpt->rss_idt) /sizeof (u32);Can you use
ARRAY_SIZEhere?
Agree.
Quoted 46 lines
Quoted 46 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +13
-0+ u32 rxq0; + int i; + + /* Fill out hash function keys */ + for (i = 0; i < key_len_by_u32; i++) { + u32 key, idx_base; + + idx_base = (key_len_by_u32 - i) * 4; + key = ((adpt->rss_key[idx_base - 1]) | + (adpt->rss_key[idx_base - 2] << 8) | + (adpt->rss_key[idx_base - 3] << 16) | + (adpt->rss_key[idx_base - 4] << 24)); + writel_relaxed(key, adpt->base + EMAC_RSS_KEY(i,u32));
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +8
-0+ } + +/* Fill out redirection table */ +for (i =0 ; i < idt_len_by_u32; i++) +writel_relaxed (adpt->rss_idt[i], + adpt->base +EMAC_RSS_TBL (i, u32)); + + writel_relaxed(adpt->rss_base_cpu, adpt->base +
EMAC_BASE_CPU_NUMBER); +drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +22
-0+ rxq0 =readl_relaxed (adpt->base +EMAC_RXQ_CTRL_0 ); +if (adpt->rss_hstype &EMAC_RSS_HSTYP_IPV4_EN ) + rxq0 |=RXQ0_RSS_HSTYP_IPV4_EN ; +else + rxq0 &= ~RXQ0_RSS_HSTYP_IPV4_EN ; + +if (adpt->rss_hstype &EMAC_RSS_HSTYP_TCP4_EN ) + rxq0 |=RXQ0_RSS_HSTYP_IPV4_TCP_EN ; +else + rxq0 &= ~RXQ0_RSS_HSTYP_IPV4_TCP_EN ; + +if (adpt->rss_hstype &EMAC_RSS_HSTYP_IPV6_EN ) + rxq0 |=RXQ0_RSS_HSTYP_IPV6_EN ; +else + rxq0 &= ~RXQ0_RSS_HSTYP_IPV6_EN ; + +if (adpt->rss_hstype &EMAC_RSS_HSTYP_TCP6_EN ) + rxq0 |=RXQ0_RSS_HSTYP_IPV6_TCP_EN ; +else + rxq0 &= ~RXQ0_RSS_HSTYP_IPV6_TCP_EN ; + + rxq0 |= ((adpt->rss_idt_size << IDT_TABLE_SIZE_SHFT) &
Quoted 9 lines
Quoted 6 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +4
-0+ IDT_TABLE_SIZE_BMSK); + rxq0 |=RSS_HASH_EN ; + +wmb ();/* ensure all parameters are written before enabling RSS */ +
+writel_relaxed(rxq0, adpt->base + EMAC_RXQ_CTRL_0);Why not just use
writel(), which already includes awmb()
ok.
Quoted 11 lines
Quoted 11 lines
(quoted diff) +2
-0+/* Power Management */ +void emac_mac_pm(struct emac_adapter *adpt, u32 speed, bool
wol_en, boolrx_en) +{drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +8
-0+u32 dma_mas, mac; + + dma_mas =readl_relaxed (adpt->base +EMAC_DMA_MAS_CTRL ); + dma_mas &= ~LPW_CLK_SEL ; + dma_mas |=LPW_STATE ; + + mac =readl_relaxed (adpt->base +EMAC_MAC_CTRL ); + mac &= ~(FULLD |RXEN |TXEN );
Quoted 8 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 519
Quoted 6 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +6
-0+ mac = (mac & ~SPEED_BMSK ) | + (((u32 )emac_mac_speed_10_100 <<SPEED_SHFT ) &SPEED_BMSK ); + +if (wol_en) { +if (rx_en) + mac |= (RXEN |BROAD_EN );You don't need the parentheses.
ok.
Quoted 12 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 548
Quoted 10 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +5
-0+/* Config descriptor rings */ +static void emac_mac_dma_rings_config (struct emac_adapter *adpt) +{ +if (adpt->timestamp_en) + emac_reg_update32(adpt->csr +
EMAC_EMAC_WRAPPER_CSR1,drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +3
-0+ 0, ENABLE_RRD_TIMESTAMP); + +/* TPD */ What does TPD stand for?
TPD: Transmit Packet Descriptor ring.
See definition of TPD, RFD and RRD at the top of emac-mac.h
Quoted 28 lines
Quoted 28 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +5
-0+writel_relaxed (EMAC_DMA_ADDR_HI (adpt->tx_q[0 ].tpd.p_addr), + adpt->base +EMAC_DESC_CTRL_1 ); +switch (adpt->tx_q_cnt) { +case 4 : +writel_relaxed(EMAC_DMA_ADDR_LO(adpt->tx_q[3].tpd.p_addr),
+adpt->base+EMAC_H3TPD_BASE_ADDR_LO);drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +3
-0+/* fall through */ +case 3 : +writel_relaxed(EMAC_DMA_ADDR_LO(adpt->tx_q[2].tpd.p_addr),
+adpt->base+EMAC_H2TPD_BASE_ADDR_LO);drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +3
-0+/* fall through */ +case 2 : +writel_relaxed(EMAC_DMA_ADDR_LO(adpt->tx_q[1].tpd.p_addr),
+adpt->base+EMAC_H1TPD_BASE_ADDR_LO);drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +3
-0+/* fall through */ +case 1 : +writel_relaxed(EMAC_DMA_ADDR_LO(adpt->tx_q[0].tpd.p_addr),
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +3
-0+ adpt->base + EMAC_DESC_CTRL_8); +break ; +default :
Quoted 9 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 573
Quoted 6 lines
(quoted diff) +2
-0+ netdev_err(adpt->netdev, + "error: invalid number of TX queues(%d)\n",
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +3
-0+ adpt->tx_q_cnt); +return ; + }This is not time-critical code. Why not just create a for-loop?
the offsets are all different:EMAC_H3TPD_BASE_ADDR_LO, EMAC_H2TPD_BASE_ADDR_LO,EMAC_H1TPD_BASE_ADDR_LO, EMAC_DESC_CTRL_8
of course I can put the offset in an array, but I am not sure that it will look better.
Quoted 10 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 634
Quoted 7 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +7
-0+/* Config transmit parameters */ +static void emac_mac_tx_config (struct emac_adapter *adpt) +{ +u16 tx_offload_thresh =EMAC_MAX_TX_OFFLOAD_THRESH ; +u32 val; + + writel_relaxed((tx_offload_thresh >> 3) &Why is
tx_offload_thresha u16 if you're going to use writel anyway? Make it a u32.
agree.
Quoted 14 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 750
Quoted 12 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +8
-0+void emac_mac_reset (struct emac_adapter *adpt) +{ +writel_relaxed (0 , adpt->base +EMAC_INT_MASK ); +writel_relaxed (DIS_INT , adpt->base +EMAC_INT_STATUS ); + +emac_mac_stop (adpt); + + emac_reg_update32(adpt->base + EMAC_DMA_MAS_CTRL, 0,
SOFT_RST);drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +2
-0+wmb ();/* ensure mac is fully reset */ +usleep_range (100 ,150 );Please add a comment explaiing why this delay is necessary.
ok.
Quoted 11 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 849
Quoted 9 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +5
-0+void emac_mac_stop (struct emac_adapter *adpt) +{ +emac_reg_update32 (adpt->base +EMAC_RXQ_CTRL_0 ,RXQ_EN ,0 ); +emac_reg_update32 (adpt->base +EMAC_TXQ_CTRL_0 ,TXQ_EN ,0 ); + emac_reg_update32(adpt->base + EMAC_MAC_CTRL, (TXEN |RXEN), 0);
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +2
-0+wmb ();/* ensure mac is stopped before we proceed */ +usleep_range (1000 ,1050 );Same here.
ok.
Quoted 15 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 859
Quoted 9 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +8
-0+/* set MAC address */ +void emac_mac_addr_clear (struct emac_adapter *adpt,u8 *addr) +{ +u32 sta; + +/* for example: 00-A0-C6-11-22-33 +* 0<-->C6112233, 1<-->00A0. +*/ /*
* Multi-line comments
* look like this.
*/
Quoted 9 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 895
Quoted 6 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +6
-0+/* Free all descriptors of given transmit queue */ +static void emac_tx_q_descs_free (struct emac_adapter *adpt, +struct emac_tx_queue *tx_q) +{ +unsigned long size; +u32 i;Since 'i' is used as an index, it should be an unsized integer. And 'size' should be a '
size_t'
ok.
Quoted 8 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 927
Quoted 4 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +3
-0+static void emac_tx_q_descs_free_all (struct emac_adapter *adpt) +{ +u8 i;'int'. Personally, I'd prefer "unsigned int", but 'int' is what you use elsewhere.
Ill go with unsigned int.
Quoted 10 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 940
Quoted 7 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +7
-0+/* Free all descriptors of given receive queue */ +static void emac_rx_q_free_descs (struct emac_adapter *adpt, +struct emac_rx_queue *rx_q) +{ +struct device *dev = adpt->netdev->dev.parent; +unsigned long size; +u32 i;
size_tsize;
int i;
will do.
Quoted 9 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 1008
Quoted 7 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +6
-0+/* Allocate TX descriptor ring for the given transmit queue */ +static int emac_tx_q_desc_alloc (struct emac_adapter *adpt, +struct emac_tx_queue *tx_q) +{ +struct emac_ring_header *ring_header = &adpt->ring_header; +unsigned long size;
size_t
ok.
Quoted 16 lines
Quoted 4 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +3
-0+ + size =sizeof (struct emac_buffer ) * tx_q->tpd.count; + tx_q->tpd.tpbuff =kzalloc (size,GFP_KERNEL );drivers/net/ethernet/qualcomm/emac/emac-mac.c line 1021
Quoted 10 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +10
-0+if (!tx_q->tpd.tpbuff) +return -ENOMEM ; + + tx_q->tpd.size = tx_q->tpd.count * (adpt->tpd_size *4 ); + tx_q->tpd.p_addr = ring_header->p_addr + ring_header->used; + tx_q->tpd.v_addr = ring_header->v_addr + ring_header->used; + ring_header->used +=ALIGN (tx_q->tpd.size,8 ); + tx_q->tpd.produce_idx =0 ; + tx_q->tpd.consume_idx =0 ; +return 0 ;blank line above "return".
ok.
Quoted 9 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 1027
Quoted 7 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +6
-0+} + +static int emac_tx_q_desc_alloc_all (struct emac_adapter *adpt) +{ +int retval =0 ; +u8 i;int i;
ok.
Quoted 6 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 1065
Quoted 4 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +3
-0+static void emac_rx_q_free_bufs_all (struct emac_adapter *adpt) +{ +u8 i;int i;
ok.
Quoted 9 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 1076
Quoted 7 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +6
-0+/* Allocate RX descriptor rings for the given receive queue */ +static int emac_rx_descs_alloc (struct emac_adapter *adpt, +struct emac_rx_queue *rx_q) +{ +struct emac_ring_header *ring_header = &adpt->ring_header; +unsigned long size;
size_t
ok.
Quoted 7 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 1106
Quoted 5 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +4
-0+static int emac_rx_descs_allocs_all (struct emac_adapter *adpt) +{ +int retval =0 ; +u8 i;int i;
ok.
Quoted 13 lines
Quoted 11 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +9
-0+ +for (i =0 ; i < adpt->rx_q_cnt; i++) { + retval =emac_rx_descs_alloc (adpt, &adpt->rx_q[i]); +if (retval) +break ; + } + +if (retval) { + netdev_err(adpt->netdev, "error: Rx Queue %u allocfailed\n",
%d
ok.
Quoted 10 lines
Quoted 10 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +9
-0+/* Produce new receive free descriptor */ +static bool emac_mac_rx_rfd_create (struct emac_adapter *adpt, +struct emac_rx_queue *rx_q, +union emac_rfd *rfd) +{ +u32 *hw_rfd =EMAC_RFD (rx_q, adpt->rfd_size, + rx_q->rfd.produce_idx); + + *(hw_rfd++) = rfd->word[0 ];
Quoted 9 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 1295
Quoted 6 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +6
-0+ *hw_rfd = rfd->word[1 ]; + +if (++rx_q->rfd.produce_idx == rx_q->rfd.count) + rx_q->rfd.produce_idx =0 ; + +return true;You never check the return value, so just make this a void function.
Agree.
Quoted 15 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 1307
Quoted 12 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +12
-0+} + +/* Fill up receive queue's RFD with preallocated receive buffers */ +static int emac_mac_rx_descs_refill (struct emac_adapter *adpt, +struct emac_rx_queue *rx_q) +{ +struct emac_buffer *curr_rxbuf; +struct emac_buffer *next_rxbuf; +union emac_rfd rfd; +struct sk_buff *skb; +void *skb_data =NULL ; +u32 count =0 ;int count = 0;
Quoted 1 line
The type should match the return value of the function.
Agree.
Quoted 17 lines
Quoted 13 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +11
-0+u32 next_produce_idx; + + next_produce_idx = rx_q->rfd.produce_idx; +if (++next_produce_idx == rx_q->rfd.count) + next_produce_idx =0 ; + curr_rxbuf =GET_RFD_BUFFER (rx_q, rx_q->rfd.produce_idx); + next_rxbuf =GET_RFD_BUFFER (rx_q, next_produce_idx); + +/* this always has a blank rx_buffer*/ +while (!next_rxbuf->dma) { + skb = dev_alloc_skb(adpt->rxbuf_size +
NET_IP_ALIGN);drivers/net/ethernet/qualcomm/emac/emac-mac.c line 1320
Quoted 2 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +2
-0+if (unlikely (!skb)) +break ;I don't think this is time-critical code, so don't use
unlikely().
ok.
Quoted 33 lines
Quoted 33 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +9
-0+/* Consume next received packet descriptor */ +static bool emac_rx_process_rrd (struct emac_adapter *adpt, +struct emac_rx_queue *rx_q, +union emac_rrd *rrd) +{ +u32 *hw_rrd =EMAC_RRD (rx_q, adpt->rrd_size, + rx_q->rrd.consume_idx); + +/* If time stamping is enabled, it will be added in the beginning of
+ * the hw rrd (hw_rrd). In sw rrd (rrd), 32bit words 4 & 5
are reserved(quoted diff) +2
-0+ * for the time stamp; hence the conversion. + * Also, read the rrd word with update flag first; readrest of rrd
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +7
-0+* only if update flag is set. +*/ +if (adpt->timestamp_en) + rrd->word[3 ] = *(hw_rrd +5 ); +else + rrd->word[3 ] = *(hw_rrd +3 ); +rmb ();/* ensure hw receive returned descriptor timestamp is read */ +
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +9
-0+if (!rrd->genr.update) +return false; + +if (adpt->timestamp_en) { + rrd->word[4 ] = *(hw_rrd++); + rrd->word[5 ] = *(hw_rrd++); + }else { + rrd->word[4 ] =0 ; + rrd->word[5 ] =0 ;
Quoted 8 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 1519
Quoted 6 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +6
-0+ } + + rrd->word[0 ] = *(hw_rrd++); + rrd->word[1 ] = *(hw_rrd++); + rrd->word[2 ] = *(hw_rrd++); +mb ();/* ensure descriptor is read */ Why do you use
mb()here, butrmb()above? The comment is the same.
I will change both to rmb()
Quoted 9 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 1596
Quoted 7 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +6
-0+static void emac_rx_rfd_clean (struct emac_rx_queue *rx_q, +union emac_rrd *rrd) +{ +struct emac_buffer *rfbuf = rx_q->rfd.rfbuff; +u32 consume_idx = rrd->genr.si; +u16 i;int i;
ok.
Quoted 10 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 1613
Quoted 7 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +6
-0+static inline bool emac_skb_cb_expired (struct sk_buff *skb) +{ +if (time_is_after_jiffies (EMAC_SKB_CB (skb)->jiffies + +msecs_to_jiffies (100 ))) +return false; +return true;return
time_is_before_jiffies(EMAC_SKB_CB(skb)->jiffies + msecs_to_jiffies(100));
Agree.
Quoted 16 lines
Quoted 4 lines
(quoted diff) +2
-0+/* Process receive event */ +void emac_mac_rx_process(struct emac_adapter *adpt, struct
emac_rx_queue*rx_q,drivers/net/ethernet/qualcomm/emac/emac-mac.c line 1750
Quoted 10 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +10
-0+ int *num_pkts, int max_pkts) +{ +struct net_device *netdev = adpt->netdev; + +union emac_rrd rrd; +struct emac_buffer *rfbuf; +struct sk_buff *skb; + +u32 hw_consume_idx, num_consume_pkts; +u32 count =0 ;unsigned int count;
ok.
Quoted 84 lines
Quoted 71 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +4
-0+u32 proc_idx; +u32 reg =readl_relaxed (adpt->base + rx_q->consume_reg); + + hw_consume_idx = (reg & rx_q->consume_mask) >>
rx_q->consume_shft;
+num_consume_pkts= (hw_consume_idx>=rx_q->rrd.consume_idx) ?(quoted diff) +2
-0+ (hw_consume_idx - rx_q->rrd.consume_idx) : + (hw_consume_idx + rx_q->rrd.count -
rx_q->rrd.consume_idx); +drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +11
-0+while (1 ) { +if (!num_consume_pkts) +break ; + +if (!emac_rx_process_rrd (adpt, rx_q, &rrd)) +break ; + +if (likely (rrd.genr.nor ==1 )) { +/* good receive */ + rfbuf =GET_RFD_BUFFER (rx_q, rrd.genr.si); + dma_unmap_single(adpt->netdev->dev.parent,
rfbuf->dma,
+rfbuf->length,DMA_FROM_DEVICE);drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +5
-0+ rfbuf->dma =0 ; + skb = rfbuf->skb; + }else { + netdev_err(adpt->netdev, + "error: multi-RFD not supportyet!\n");
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +7
-0+break ; + } +emac_rx_rfd_clean (rx_q, &rrd); + num_consume_pkts--; + count++; + +/* Due to a HW issue in L4 check sum detection (UDP/TCP frags
+ * with DF set are marked as error), drop packets
based on the
+ * error mask rather than the summary bit (ignoring
L4F errors)(quoted diff) +2
-0+ */ + if (rrd.word[EMAC_RRD_STATS_DW_IDX] &
EMAC_RRD_ERROR) {(quoted diff) +2
-0+ netif_dbg(adpt, rx_status, adpt->netdev, + "Drop error packet[RRD:0x%x:0x%x:0x%x:0x%x]\n",
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +13
-0+ rrd.word[0 ], rrd.word[1 ], + rrd.word[2], rrd.word[3]); + +dev_kfree_skb (skb); +continue ; + } + +skb_put (skb, rrd.genr.pkt_len -ETH_FCS_LEN ); + skb->dev = netdev; + skb->protocol =eth_type_trans (skb, skb->dev); +if (netdev->features &NETIF_F_RXCSUM ) + skb->ip_summed = ((rrd.genr.l4f) ? + CHECKSUM_NONE :
CHECKSUM_UNNECESSARY);drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +5
-0+else +skb_checksum_none_assert (skb); + +if (test_bit (EMAC_STATUS_TS_RX_EN , &adpt->status)) { + struct skb_shared_hwtstamps *hwts =
skb_hwtstamps(skb); +drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +2
-0+ hwts->hwtstamp =ktime_set (rrd.genr.ts_high, + rrd.genr.ts_low);drivers/net/ethernet/qualcomm/emac/emac-mac.c line 1819
Quoted 10 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +10
-0+ } + +emac_receive_skb (rx_q, skb, (u16 )rrd.genr.cvlan_tag, + (bool )rrd.genr.cvlan_flag); + + netdev->last_rx = jiffies; + (*num_pkts)++; +if (*num_pkts >= max_pkts) +break ; + }How about
Quoted 3 lines
do {
...
} while (*num_pkts<max_pkts);
Agree.
Quoted 10 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 2020
Quoted 7 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +6
-0+/* Check if enough transmit descriptors are available */ +static bool emac_tx_has_enough_descs (struct emac_tx_queue *tx_q, +const struct sk_buff *skb) +{ +u32 num_required =1 ; +u16 i;int i;
Quoted 4 lines
Anyway, you got the idea. I think sized integers should be used sparingly, and general counting and index variable should be unsized integers, preferably also unsigned.
ok. I will change it throughout the driver.
Quoted 15 lines
Quoted 15 lines
+/* Fill up transmit descriptors with TSO and Checksum offload information */ +static int emac_tso_csum(struct
emac_adapter*adpt,drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +10
-0+struct emac_tx_queue *tx_q, + struct sk_buff *skb, + union emac_tpd *tpd) +{ + u8 hdr_len; +int retval; + + if (skb_is_gso(skb)) { + if (skb_header_cloned(skb)) { + retval = pskb_expand_head(skb, 0, 0,
GFP_ATOMIC);drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +2
-0+if (unlikely (retval)) +return retval;
Quoted 8 lines
Quoted 6 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +5
-0+ } + +if (skb->protocol ==htons (ETH_P_IP )) { + u32 pkt_len = + ((unsigned char *)ip_hdr(skb) -
skb->data) +Use void* for pointer math, instead of "unsigned char *".
pointer math on void* ?
what is the size of void ?
Quoted 27 lines
Quoted 27 lines
(quoted diff) +2
-0+/* Transmit the packet using specified transmit queue */ +int emac_mac_tx_buf_send(struct emac_adapter *adpt, struct
emac_tx_queue*tx_q,drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +23
-0+ struct sk_buff *skb) +{ +union emac_tpd tpd; +u32 prod_idx; + +if (test_bit (EMAC_STATUS_DOWN , &adpt->status)) { +dev_kfree_skb_any (skb); +return NETDEV_TX_OK ; + } + +if (!emac_tx_has_enough_descs (tx_q, skb)) { +/* not enough descriptors, just stop queue */ +netif_stop_queue (adpt->netdev); +return NETDEV_TX_BUSY ; + } + +memset (&tpd,0 ,sizeof (tpd)); + +if (emac_tso_csum (adpt, tx_q, skb, &tpd) !=0 ) { +dev_kfree_skb_any (skb); +return NETDEV_TX_OK ; + } +
Quoted 10 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c line 2246
Quoted 6 lines
drivers/net/ethernet/qualcomm/emac/emac-mac.c new file +6
-0+if (skb_vlan_tag_present (skb)) { +u16 vlan =skb_vlan_tag_get (skb); +u16 tag; + +EMAC_VLAN_TO_TAG (vlan, tag); + tpd.genr.cvlan_tag = tag;Can't you just do
EMAC_VLAN_TO_TAG(vlan, tpd.genr.cvlan_tag);
Should have been the way you suggested.
However, per Felix's comment I will change the bit fields to bitwise macros. This will force code slow similar to the original.
Quoted 46 lines
Quoted 46 lines
(quoted diff) +0
-0
b/drivers/net/ethernet/qualcomm/emac/emac-mac.hnew file mode 100644 index 0000000..a6761afdrivers/net/ethernet/qualcomm/emac/emac-mac.h new file +1
-0@@ -0,0 +1,341 @@ +/* Copyright (c) 2013-2015, The Linux Foundation. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 and
+ * only version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ */drivers/net/ethernet/qualcomm/emac/emac-mac.h new file +7
-0+ +/* EMAC DMA HW engine uses three rings: +* Tx: +* TPD: Transmit Packet Descriptor ring. +* Rx: +* RFD: Receive Free Descriptor ring. +* Ring of descriptors with empty buffers to be filled by Rx HW.
+ * RRD: Receive Return Descriptor ring.
+ * Ring of descriptors with buffers filled with received data.
+ */drivers/net/ethernet/qualcomm/emac/emac-mac.h new file +6
-0+ +#ifndef _EMAC_HW_H_ +#define _EMAC_HW_H_ + +/* EMAC_CSR register offsets */ +#define EMAC_EMAC_WRAPPER_CSR1
0x000000+#defineEMAC_EMAC_WRAPPER_CSR20x000004
+#defineEMAC_EMAC_WRAPPER_CSR30x000008+#defineEMAC_EMAC_WRAPPER_CSR50x000010
+#defineEMAC_EMAC_WRAPPER_TX_TS_LO
Quoted 8 lines
Quoted 4 lines
0x000104+#defineEMAC_EMAC_WRAPPER_TX_TS_HI0x000108
+#defineEMAC_EMAC_WRAPPER_TX_TS_INX0x00010cCan you move some of the macros into the .c files? For example, I'm pretty sure that the
EMAC_EMAC_WRAPPER_CSRxmacros are used only inemac-sgmii.c.
I have moved all that made sense to .c files already. If I find anything else that can be moved I will move it too.
For the case of EMAC_EMAC_WRAPPER_CSRx:EMAC_EMAC_WRAPPER_CSR1 used in emac-mac.c EMAC_EMAC_WRAPPER_CSR1 used in emac-mac.c and emac-sgmii.c andEMAC_EMAC_WRAPPER_TX_xxx used in emac-mac.c
Since they are all part of EMAC_CSR register space I think that it is better to keep them togther.
Quoted 2 lines
Anyway, I'm stopping for now. I'll post more later.
Thank you again. Looking forward for the rest of it. Gilad
Quoted 4 lines
--
To unsubscribe from this list: send the line "unsubscribe linux-arm-msm" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
Quoted 2 lines · Gilad Avidov
pointer math on void* ?
what is the size of void ?
I'm talking about adding and subtracting pointer values, so
u32 pkt_len =((void *)ip_hdr(skb) - skb->data)
Signature · 3 lines
Sent by an employee of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, hosted by The Linux Foundation.