[LEDE-DEV] MT7621 support Jumbo frames

Jaap Buurman jaapbuurman at gmail.com
Sun Apr 23 05:42:19 PDT 2017


The mediatek folder is not there by default. It gets generated in the
during the "make" command when compiling the image by the patches. It
can be found at the following full path:

source/build_dir/target-mipsel_24kc_musl/linux-ramips_mt7621/linux-4.4.61/drivers/net/ethernet/mediatek

I will try if I can patch the 17.01 branch. Does anyone have an idea
why gsw_mt7620.h is missing from the master branch?

On Sun, Apr 23, 2017 at 2:30 PM, Alberto Bursi <alberto.bursi at outlook.it> wrote:
> Weird, I'm not seeing the whole "mediatek" folder if I try to follow
> your steps from a fresh source download with LEDE master (trunk/HEAD,
> whatever).
>
> I've checked the LEDE 17.01 branch and yes, there the "mediatek" folder
> is there and also that file. Kernel version is the same for both,
> 4.4.61, and that "mediatek" folder isn't there in the source package
> downloaded so it must come from LEDE patches.
>
> So go in main source folder and write "git checkout lede-17.01" then it
> should work.
>
> -Alberto
>
>
> On 04/23/2017 12:52 PM, Jaap Buurman wrote:
>> Thank you all for the suggestions. I've tried the following steps:
>>
>> 1) Add the patch in a file with a name so that it will be added last:
>> 999-mtu.patch. This failed during the build.
>> 2) Next, I wanted to write these changes manually to a patch with
>> quilt. However, the file /drivers/net/ethernet/mediatek/gsw_mt7620.h
>> was not there, so I could not apply those changes manually either.
>> That was the only file that gave me issues. The other files could be
>> found in the build_dir, and thus I could apply all the changes
>> manually. So that's where I'm stuck at the moment.
>>
>> On Sun, Apr 23, 2017 at 11:34 AM, Alberto Bursi
>> <alberto.bursi at outlook.it> wrote:
>>> That patch is a raw kernel patch, you need to integrate it in LEDE's
>>> buildsystem and hope that he is working with LEDE patches already applied.
>>>
>>> Copy the patch's text in a text file and place this file in the /patches-*
>>> folder in your device's source folder
>>>
>>> Your SoC is a ramips so it should be target/linux/ramips/patches-4.4 as he
>>> said it is for kernel 4.4.7
>>>
>>> Give it a name that makes sure it will be added last, like 999-mypatch.patch
>>>
>>> Then you can select your device with make menuconfig and then start a
>>> compilation with make and see if it goes well.
>>>
>>> If that fails, you need to follow the tutorial as linked, and create a new
>>> kernel patch where you write these changes manually to each file he changed,
>>> then save the new patch, and then move the patch file you created from the
>>> quilt kernel patch folder (platform/) to the patch folder as said above.
>>>
>>> There might be better ways, but that's what I did when I had to add new
>>> kernel patches.
>>>
>>> -Alberto
>>>
>>>
>>> On 04/22/2017 01:53 PM, Jaap Buurman wrote:
>>>
>>> I have never applied patches before, so I am probably making mistakes.
>>> The first problem I'm running into, is that it is trying to patch the
>>> following file:
>>>
>>> /drivers/net/ethernet/mediatek/gsw_mt7620.h
>>>
>>> However, such a file does not exist in that path in my build_dir. I
>>> can find gsw_mt7620.c and gsw_mt7620.o files just fine in the
>>> aforementioned path, but not the .h file it is trying to patch. To
>>> which branch does the patch apply cleanly? I am currently trying to
>>> patch the master branch. Should I try the 17.01 branch instead?
>>>
>>> On Wed, Apr 19, 2017 at 10:32 AM, Jaap Buurman <jaapbuurman at gmail.com>
>>> wrote:
>>>
>>> Ah, that sounds even better :) I will try to compile and test this
>>> patch tomorrow or the day after tomorrow. Will let you know if it
>>> works. Thanks again for the effort you're putting into this!
>>>
>>> On Wed, Apr 19, 2017 at 10:23 AM, Gaetano Catalli
>>> <gaetano.catalli at gmail.com> wrote:
>>>
>>> I'm still working on this since I would like to raise the limit up to
>>> 9KB if possible. Please, let me know if this works for you.
>>>
>>> On Wed, Apr 19, 2017 at 10:18 AM, Jaap Buurman <jaapbuurman at gmail.com>
>>> wrote:
>>>
>>> Wow, this is perfect. Thank you very much. I will try to use this
>>> patch and compile my own image with up to 2kb frame support. Do you
>>> have any plans on submitting this as a PR to the LEDE git?
>>>
>>> On Wed, Apr 19, 2017 at 9:32 AM, Gaetano Catalli
>>> <gaetano.catalli at gmail.com> wrote:
>>>
>>> I've been working on this for a while. Apparently the embedded 5-port
>>> gigabit switch is able to handle packets with size up to 15KB. On the
>>> contrary, the GMAC, to which the switch is attached, has a limit of
>>> 2KB. The following is a patch that changes the max recv frame length
>>> to 2KB and allows to set the MTU up to that value. It is based on
>>> kernel 4.4.7.
>>>
>>> Signed-off-by: Gaetano Catalli <gaetano.catalli at gmail.com>
>>> ---
>>>   drivers/net/ethernet/mediatek/gsw_mt7620.h  |  6 +++++
>>>   drivers/net/ethernet/mediatek/gsw_mt7621.c  | 13 +++++++----
>>>   drivers/net/ethernet/mediatek/mtk_eth_soc.c | 36
>>> ++++++++++++++++++-----------
>>>   drivers/net/ethernet/mediatek/mtk_eth_soc.h |  1 +
>>>   drivers/net/ethernet/mediatek/soc_mt7621.c  |  2 +-
>>>   5 files changed, 39 insertions(+), 19 deletions(-)
>>>
>>> diff --git a/drivers/net/ethernet/mediatek/gsw_mt7620.h
>>> b/drivers/net/ethernet/mediatek/gsw_mt7620.h
>>> index dcef9a8..ce3cd10 100644
>>> --- a/drivers/net/ethernet/mediatek/gsw_mt7620.h
>>> +++ b/drivers/net/ethernet/mediatek/gsw_mt7620.h
>>> @@ -45,6 +45,12 @@
>>>   #define GSW_REG_ISR 0x700c
>>>   #define GSW_REG_GPC1 0x7014
>>>
>>> +#define GSW_REG_MAC_P0_MCR 0x100
>>> +#define GSW_REG_MAC_P1_MCR 0x200
>>> +
>>> +// Global MAC control register
>>> +#define GSW_REG_GMACCR 0x30E0
>>> +
>>>   #define SYSC_REG_CHIP_REV_ID 0x0c
>>>   #define SYSC_REG_CFG1 0x14
>>>   #define RST_CTRL_MCM BIT(2)
>>> diff --git a/drivers/net/ethernet/mediatek/gsw_mt7621.c
>>> b/drivers/net/ethernet/mediatek/gsw_mt7621.c
>>> index 96280b4..db5d56d 100644
>>> --- a/drivers/net/ethernet/mediatek/gsw_mt7621.c
>>> +++ b/drivers/net/ethernet/mediatek/gsw_mt7621.c
>>> @@ -99,17 +99,20 @@ static void mt7621_hw_init(struct mt7620_gsw *gsw,
>>> struct device_node *np)
>>>    usleep_range(10, 20);
>>>
>>>    if ((rt_sysc_r32(SYSC_REG_CHIP_REV_ID) & 0xFFFF) == 0x0101) {
>>> - /* (GE1, Force 1000M/FD, FC ON, MAX_RX_LENGTH 1536) */
>>> - mtk_switch_w32(gsw, 0x2105e30b, 0x100);
>>> + /* (GE1, Force 1000M/FD, FC ON, MAX_RX_LENGTH 2k) */
>>> + mtk_switch_w32(gsw, 0x2305e30b, GSW_REG_MAC_P0_MCR);
>>>    mt7530_mdio_w32(gsw, 0x3600, 0x5e30b);
>>>    } else {
>>> - /* (GE1, Force 1000M/FD, FC ON, MAX_RX_LENGTH 1536) */
>>> - mtk_switch_w32(gsw, 0x2105e33b, 0x100);
>>> + /* (GE1, Force 1000M/FD, FC ON, MAX_RX_LENGTH 2k) */
>>> + mtk_switch_w32(gsw, 0x2305e33b, GSW_REG_MAC_P0_MCR);
>>>    mt7530_mdio_w32(gsw, 0x3600, 0x5e33b);
>>>    }
>>>
>>>    /* (GE2, Link down) */
>>> - mtk_switch_w32(gsw, 0x8000, 0x200);
>>> + mtk_switch_w32(gsw, 0x8000, GSW_REG_MAC_P1_MCR);
>>> +
>>> + /* Set switch max RX frame length to 2k */
>>> + mt7530_mdio_w32(gsw, GSW_REG_GMACCR, 0x3F0B);
>>>
>>>    /* Enable Port 6, P5 as GMAC5, P5 disable */
>>>    val = mt7530_mdio_r32(gsw, 0x7804);
>>> diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
>>> b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
>>> index 48eda44..5edafa6 100644
>>> --- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
>>> +++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
>>> @@ -37,7 +37,7 @@
>>>   #include "mdio.h"
>>>   #include "ethtool.h"
>>>
>>> -#define MAX_RX_LENGTH 1536
>>> +#define MAX_RX_LENGTH 2048
>>>   #define FE_RX_ETH_HLEN (VLAN_ETH_HLEN + VLAN_HLEN + ETH_FCS_LEN)
>>>   #define FE_RX_HLEN (NET_SKB_PAD + FE_RX_ETH_HLEN + NET_IP_ALIGN)
>>>   #define DMA_DUMMY_DESC 0xffffffff
>>> @@ -1358,11 +1358,29 @@ static int fe_do_ioctl(struct net_device *dev,
>>> struct ifreq *ifr, int cmd)
>>>    return -EOPNOTSUPP;
>>>   }
>>>
>>> +static int fe_set_mtu(struct fe_priv* priv, int new_mtu)
>>> +{
>>> + int frag_size = fe_max_frag_size(new_mtu);
>>> + u32 fwd_cfg;
>>> +
>>> + fe_stop(priv->netdev);
>>> + fwd_cfg = fe_r32(FE_GDMA1_FWD_CFG);
>>> + if (new_mtu <= ETH_DATA_LEN) {
>>> + fwd_cfg &= ~FE_GDM1_JMB_EN;
>>> + } else {
>>> + fwd_cfg &= ~(FE_GDM1_JMB_LEN_MASK << FE_GDM1_JMB_LEN_SHIFT);
>>> + fwd_cfg |= (DIV_ROUND_UP(frag_size, 1024) <<
>>> + FE_GDM1_JMB_LEN_SHIFT) | FE_GDM1_JMB_EN;
>>> + }
>>> + fe_w32(fwd_cfg, FE_GDMA1_FWD_CFG);
>>> +
>>> + return fe_open(priv->netdev);
>>> +}
>>> +
>>>   static int fe_change_mtu(struct net_device *dev, int new_mtu)
>>>   {
>>>    struct fe_priv *priv = netdev_priv(dev);
>>>    int frag_size, old_mtu;
>>> - u32 fwd_cfg;
>>>
>>>    if (!(priv->flags & FE_FLAG_JUMBO_FRAME))
>>>    return eth_change_mtu(dev, new_mtu);
>>> @@ -1389,18 +1407,10 @@ static int fe_change_mtu(struct net_device
>>> *dev, int new_mtu)
>>>    if (!netif_running(dev))
>>>    return 0;
>>>
>>> - fe_stop(dev);
>>> - fwd_cfg = fe_r32(FE_GDMA1_FWD_CFG);
>>> - if (new_mtu <= ETH_DATA_LEN) {
>>> - fwd_cfg &= ~FE_GDM1_JMB_EN;
>>> - } else {
>>> - fwd_cfg &= ~(FE_GDM1_JMB_LEN_MASK << FE_GDM1_JMB_LEN_SHIFT);
>>> - fwd_cfg |= (DIV_ROUND_UP(frag_size, 1024) <<
>>> - FE_GDM1_JMB_LEN_SHIFT) | FE_GDM1_JMB_EN;
>>> - }
>>> - fe_w32(fwd_cfg, FE_GDMA1_FWD_CFG);
>>> + if (priv->soc->set_mtu)
>>> + return priv->soc->set_mtu(priv, new_mtu);
>>>
>>> - return fe_open(dev);
>>> + return fe_set_mtu(priv, new_mtu);
>>>   }
>>>
>>>   static const struct net_device_ops fe_netdev_ops = {
>>> diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.h
>>> b/drivers/net/ethernet/mediatek/mtk_eth_soc.h
>>> index d5f8b87..3acc2c1 100644
>>> --- a/drivers/net/ethernet/mediatek/mtk_eth_soc.h
>>> +++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.h
>>> @@ -381,6 +381,7 @@ struct fe_soc_data {
>>>    void (*reset_fe)(void);
>>>    void (*set_mac)(struct fe_priv *priv, unsigned char *mac);
>>>    int (*fwd_config)(struct fe_priv *priv);
>>> + int (*set_mtu)(struct fe_priv *priv, int new_mtu);
>>>    void (*tx_dma)(struct fe_tx_dma *txd);
>>>    int (*switch_init)(struct fe_priv *priv);
>>>    int (*switch_config)(struct fe_priv *priv);
>>> diff --git a/drivers/net/ethernet/mediatek/soc_mt7621.c
>>> b/drivers/net/ethernet/mediatek/soc_mt7621.c
>>> index 1609a3e..f95c354 100644
>>> --- a/drivers/net/ethernet/mediatek/soc_mt7621.c
>>> +++ b/drivers/net/ethernet/mediatek/soc_mt7621.c
>>> @@ -140,7 +140,7 @@ static void mt7621_init_data(struct fe_soc_data *data,
>>>
>>>    priv->flags = FE_FLAG_PADDING_64B | FE_FLAG_RX_2B_OFFSET |
>>>    FE_FLAG_RX_SG_DMA | FE_FLAG_NAPI_WEIGHT |
>>> - FE_FLAG_HAS_SWITCH;
>>> + FE_FLAG_HAS_SWITCH | FE_FLAG_JUMBO_FRAME;
>>>
>>>    netdev->hw_features = NETIF_F_IP_CSUM | NETIF_F_RXCSUM |
>>>    NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_SG | NETIF_F_TSO |
>>> --
>>> 2.7.4
>>>
>>> On Mon, Apr 10, 2017 at 9:21 AM, Jaap Buurman <jaapbuurman at gmail.com> wrote:
>>>
>>> I am able to test this on my DIR-860l. I could try to compile my own
>>> build to test this, but I am no C programmer myself. Am I correct in
>>> the following post about how to set the flag to allow for jumbo frames
>>> to be set?:
>>> https://forum.lede-project.org/t/build-for-the-d-link-dir-860l/948/40?u=mushoz
>>> If so, I will be able to apply the patch myself and test it. I could
>>> then send a PR if the tests are successful.
>>>
>>> Yours sincerely,
>>>
>>> Jaap Buurman
>>>
>>> On Mon, Apr 10, 2017 at 9:16 AM, Milan Kočvara <milankocvara at gmail.com>
>>> wrote:
>>>
>>> Hello all,
>>> if someone do a testing patch for jumboFrames, I can lend a helping
>>> hand with testing on EdgeRouter X. I think MT7621 know JumboFrames,
>>> because of Ubiquiti FMW allow to set MTU 1536B
>>>
>>> 2017-04-10 2:06 GMT+02:00  <rosenp at gmail.com>:
>>>
>>> Probably was not in the interest of the driver writers. Based on the
>>> copyrights though, it's mostly LEDE/OpenWRT developers. Try modifying
>>> the code to test if jumbo frames work.
>>>
>>> On Sun, 2017-04-09 at 21:20 +0200, Jaap Buurman wrote:
>>>
>>> Hello all,
>>>
>>> I found the message below in a conversation from back in August, 2016
>>> in this mailinglist. I did not find a reply to this question. Has
>>> there ever been one? Or does anyone else happen to know the answer to
>>> this question? Thank you very much in advance.
>>>
>>> Yours sincerely,
>>>
>>> Jaap Buurman
>>>
>>> August, 2016 message:
>>>
>>>
>>> Hi all,
>>>
>>> in the MT7621 ethernet driver code
>>> (drivers/net/ethernet/mediatek/soc_mt7621.c) the FE_FLAG_JUMBO_FRAME
>>> flag is not during the device initialization. This prevents the user
>>> to set an MTU greater than 1500. Is this correct? Does MT7621 support
>>> jumbo frames?
>>>
>>> Thanks.
>>>
>>> _______________________________________________
>>> Lede-dev mailing list
>>> Lede-dev at lists.infradead.org
>>> http://lists.infradead.org/mailman/listinfo/lede-dev
>>>
>>>
>>> --
>>> gaetano catalli
>>>
>>>
>>> --
>>> gaetano catalli
>>>
>>> _______________________________________________
>>> Lede-dev mailing list
>>> Lede-dev at lists.infradead.org
>>> http://lists.infradead.org/mailman/listinfo/lede-dev
>>>
>>>
>



More information about the Lede-dev mailing list