[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CAMdnO-K8CPMihDwJnzy1KcTXNT51FGeTAYRQFHMdG6fG45wR-g@mail.gmail.com>
Date: Fri, 30 Aug 2024 16:35:22 -0700
From: Jitendra Vegiraju <jitendra.vegiraju@...adcom.com>
To: Serge Semin <fancer.lancer@...il.com>
Cc: netdev@...r.kernel.org, alexandre.torgue@...s.st.com, joabreu@...opsys.com,
davem@...emloft.net, edumazet@...gle.com, kuba@...nel.org, pabeni@...hat.com,
mcoquelin.stm32@...il.com, bcm-kernel-feedback-list@...adcom.com,
richardcochran@...il.com, ast@...nel.org, daniel@...earbox.net,
hawk@...nel.org, john.fastabend@...il.com, rmk+kernel@...linux.org.uk,
ahalaney@...hat.com, xiaolei.wang@...driver.com, rohan.g.thomas@...el.com,
Jianheng.Zhang@...opsys.com, linux-kernel@...r.kernel.org,
linux-stm32@...md-mailman.stormreply.com,
linux-arm-kernel@...ts.infradead.org, bpf@...r.kernel.org, andrew@...n.ch,
linux@...linux.org.uk, horms@...nel.org, florian.fainelli@...adcom.com
Subject: Re: [net-next v4 3/5] net: stmmac: Integrate dw25gmac into stmmac
hwif handling
On Thu, Aug 29, 2024 at 3:52 AM Serge Semin <fancer.lancer@...il.com> wrote:
>
> Hi Jitendra
>
> On Mon, Aug 26, 2024 at 11:53:13AM -0700, Jitendra Vegiraju wrote:
> > Hi Serge(y)
> > Thank you for reviewing the patch.
> >
> > On Fri, Aug 23, 2024 at 6:49 AM Serge Semin <fancer.lancer@...il.com> wrote:
> > >
> > > Hi Jitendra
> > >
> > > On Wed, Aug 14, 2024 at 03:18:16PM -0700, jitendra.vegiraju@...adcom.com wrote:
> > > > From: Jitendra Vegiraju <jitendra.vegiraju@...adcom.com>
> > > >
> > > > Integrate dw25gmac support into stmmac hardware interface handling.
> > > > Added a new entry to the stmmac_hw table in hwif.c.
> > > > Define new macros DW25GMAC_CORE_4_00 and DW25GMAC_ID to identify 25GMAC
> > > > device.
> > > > Since BCM8958x is an early adaptor device, the synopsis_id reported in HW
> > > > is 0x32 and device_id is DWXGMAC_ID. Provide override support by defining
> > > > synopsys_dev_id member in struct stmmac_priv so that driver specific setup
> > > > functions can override the hardware reported values.
> > > >
> > > > Signed-off-by: Jitendra Vegiraju <jitendra.vegiraju@...adcom.com>
> > > > ---
> > > > + }, {
> > > > + .gmac = false,
> > > > + .gmac4 = false,
> > > > + .xgmac = true,
> > > > + .min_id = DW25GMAC_CORE_4_00,
> > > > + .dev_id = DW25GMAC_ID,
> > > > + .regs = {
> > > > + .ptp_off = PTP_XGMAC_OFFSET,
> > > > + .mmc_off = MMC_XGMAC_OFFSET,
> > > > + .est_off = EST_XGMAC_OFFSET,
> > > > + },
> > > > + .desc = &dwxgmac210_desc_ops,
> > > > + .dma = &dw25gmac400_dma_ops,
> > > > + .mac = &dwxgmac210_ops,
> > > > + .hwtimestamp = &stmmac_ptp,
> > > > + .mode = NULL,
> > > > + .tc = &dwmac510_tc_ops,
> > > > + .mmc = &dwxgmac_mmc_ops,
> > > > + .est = &dwmac510_est_ops,
> > > > + .setup = dwxgmac2_setup,
> > > > + .quirks = NULL,
> > > > },
> > >
> > > This can be replaced with just:
> > >
> > > + }, {
> > > + .gmac = false,
> > > + .gmac4 = false,
> > > + .xgmac = true,
> > > + .min_id = DW25GMAC_CORE_4_00,
> > > + .dev_id = DWXGMAC_ID, /* Early DW 25GMAC IP-core had XGMAC ID */
> > > + .regs = {
> > > + .ptp_off = PTP_XGMAC_OFFSET,
> > > + .mmc_off = MMC_XGMAC_OFFSET,
> > > + .est_off = EST_XGMAC_OFFSET,
> > > + },
> > > + .desc = &dwxgmac210_desc_ops,
> > > + .dma = &dw25gmac400_dma_ops,
> > > + .mac = &dwxgmac210_ops,
> > > + .hwtimestamp = &stmmac_ptp,
> > > + .mode = NULL,
> > > + .tc = &dwmac510_tc_ops,
> > > + .mmc = &dwxgmac_mmc_ops,
> > > + .est = &dwmac510_est_ops,
> > > + .setup = dw25gmac_setup,
> > > + .quirks = NULL,
> > > }
> > >
> > > and you won't need to pre-define the setup() method in the
> > > glue driver. Instead you can define a new dw25xgmac_setup() method in
> > > the dwxgmac2_core.c as it's done for the DW XGMAC/LXGMAC IP-cores.
> > >
> > > Note if your device is capable to work with up to 10Gbps speed, then
> > > just set the plat_stmmacenet_data::max_speed field to SPEED_10000.
> > > Alternatively if you really need to specify the exact MAC
> > > capabilities, then you can implement what Russell suggested here
> > > sometime ago:
> > > https://lore.kernel.org/netdev/Zf3ifH%2FCjyHtmXE3@shell.armlinux.org.uk/
> > >
> > I like your suggestion to add one stmmac_hw[] array entry (entry_a) for this
> > "early release" DW25GMAC IP and another entry (entry_b) for final DW25MAC
> > IP, in the process eliminate the need for a new member variable in struct
> > stmmac_priv.
> >
>
> > However, I would like to bring to your attention that this device requires
> > special handling for both synopsys_id and dev_id.
> > This device is reporting 0x32 for synopsys_id and 0x76(XGMAC) for dev_id.
> > The final 25GMAC spec will have 0x40 for synopsys_id and 0x55(25GMAC) for
> > dev_id.
>
> For some reason I was thinking that your device had only the device ID
> pre-defined with the XGMAC value meanwhile the Synopsys ID was 0x40.
> Indeed you get to override both of these data in the platform-specific
> setup() method.
>
> >
> > So, in order to avoid falsely qualifying other XGMAC devices with
> > synopsis_id >=0x32 as DW25GMAC, I am thinking we will have to overwrite the
> > synopsys_id to 0x40 (DW25GMAC_CORE_4_00) in glue driver using existing
> > glue driver override mechanism.
> >
> > We can implement dw25gmac_setup() in dwxgmac2_core.c for generic 25GMAC
> > case. But, this glue driver will have to rely on its own setup function
> > to override the synopsys_id as DW25GMAC_CORE_4_00.
> >
> > Do you think it looks reasonable?
>
> What I was trying to avoid was the setup() method re-definition just
> for the sake of the IP-core version override. Because if not for that
> you could have created and used the generic DW 25GMAC dw25gmac_setup()
> function.
>
> One of the possible solutions I was thinking was to introduce the
> plat_stmmacenet_data::{snps_id,dev_id} fields and use their values in
> the stmmac_hwif_init() procedure instead of the data read from the
> MAC.VERSION CSR.
>
Hi Serge(y),
Thanks for the suggestions, I will implement this option since the
code change is mostly local.
We will have to add following check in hwif.c
@@ -313,7 +313,10 @@ int stmmac_hwif_init(struct stmmac_priv *priv)
u32 id, dev_id = 0;
int i, ret;
- if (needs_gmac) {
+ if (priv->plat->snps_id && priv->plat->snps_dev_id) {
+ id = priv->plat->snps_id;
+ dev_id = priv->plat->snps_dev_id;
+ } else if (needs_gmac) {
id = stmmac_get_id(priv, GMAC_VERSION);
> Another solution could be to add the plat_stmmacenet_data::has_25gmac
> field and fix the generic driver code to using it where it's relevant.
> Then you won't need to think about what actual Synopsys ID/Device ID
> since it would mean a whole new IP-core.
>
> -Serge(y)
>
Powered by blists - more mailing lists