Message ID | 1391205045-1751-2-git-send-email-jgunthorpe@obsidianresearch.com |
---|---|
State | Not Applicable, archived |
Delegated to: | David Miller |
Headers | show |
Hi, 2014-01-31 Jason Gunthorpe <jgunthorpe@obsidianresearch.com>: > This makes the generic of_mdiobus_register parse the DT compatible string for > the pattern ethernet-phy-idAAAA.BBBB. If present it should be a value that > matches the phy-id register normally readable through MDIO. > > When the ID is given the phy autoprobing is defeated and the phy is > created directly. > > This is necessary to support phy's that cannot be autoprobed when > of_mdiobus_register is called. Specifically, my case has the phy in reset at > of_mdiobus_register, the reset is only released once the ethernet driver > starts, before it attaches to the phy. Who is responsible for bringing the PHY out of reset, is it the Ethernet/MDIO bus driver? Is the PHY put into reset using, e.g a GPIO line or any sort of reset controller, if so, should not we have some sort of reset handle node and handle that in a generic manner? Is your DTS or DTB hardcoding the PHY id, or are you having your bootloader detect the exact PHY for you, then putting back the PHY into reset to save power, until someone uses that PHY again? > > Tested on ARM Kirkwood with phy id 0x01410e90 (Marvell 88E1318) > > Signed-off-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com> > --- > drivers/of/of_mdio.c | 30 +++++++++++++++++++++++++++++- > 1 file changed, 29 insertions(+), 1 deletion(-) > > diff --git a/drivers/of/of_mdio.c b/drivers/of/of_mdio.c > index d5a57a9..c93287e 100644 > --- a/drivers/of/of_mdio.c > +++ b/drivers/of/of_mdio.c > @@ -22,6 +22,30 @@ > MODULE_AUTHOR("Grant Likely <grant.likely@secretlab.ca>"); > MODULE_LICENSE("GPL"); > > +/* Extract the clause 22 phy ID from the compatible string of the form > + * ethernet-phy-idAAAA.BBBB */ > +static int of_get_phy_id(struct device_node *device, u32 *phy_id) > +{ > + const char *cp; > + int cplen, l; > + unsigned int upper, lower; > + > + cp = of_get_property(device, "compatible", &cplen); > + if (cp == NULL) > + return -EINVAL; > + while (cplen > 0) { > + if (sscanf(cp, "ethernet-phy-id%4x.%4x", &upper, &lower) == 2) { You might want to guard against 0x0 and 0xffff just in case whoever fills this information in the Device Tree was reading bogus data out of the MDIO bus, otherwise, chances are that the "Generic PHY" driver will be picked up, and it might still not be appropriate for driving your PHY chip. > + *phy_id = ((upper & 0xFFFF) << 16) | (lower & 0xFFFF); > + return 0; > + } > + > + l = strlen(cp) + 1; > + cp += l; > + cplen -= l; > + } > + return -EINVAL; > +} > + > /** > * of_mdiobus_register - Register mii_bus and create PHYs from the device tree > * @mdio: pointer to mii_bus structure > @@ -36,6 +60,7 @@ int of_mdiobus_register(struct mii_bus *mdio, struct device_node *np) > struct device_node *child; > const __be32 *paddr; > u32 addr; > + u32 phy_id; > bool is_c45, scanphys = false; > int rc, i, len; > > @@ -81,7 +106,10 @@ int of_mdiobus_register(struct mii_bus *mdio, struct device_node *np) > > is_c45 = of_device_is_compatible(child, > "ethernet-phy-ieee802.3-c45"); > - phy = get_phy_device(mdio, addr, is_c45); > + if (!is_c45 && !of_get_phy_id(child, &phy_id)) > + phy = phy_device_create(mdio, addr, phy_id, 0, NULL); > + else > + phy = get_phy_device(mdio, addr, is_c45); > > if (!phy || IS_ERR(phy)) { > dev_err(&mdio->dev, > -- > 1.8.1.2 > > -- > To unsubscribe from this list: send the line "unsubscribe linux-kernel" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html > Please read the FAQ at http://www.tux.org/lkml/
On Fri, Jan 31, 2014 at 02:24:52PM -0800, Florian Fainelli wrote: > > This is necessary to support phy's that cannot be autoprobed when > > of_mdiobus_register is called. Specifically, my case has the phy in reset at > > of_mdiobus_register, the reset is only released once the ethernet driver > > starts, before it attaches to the phy. > > Who is responsible for bringing the PHY out of reset, is it the > Ethernet/MDIO bus driver? Is the PHY put into reset using, e.g a GPIO > line or any sort of reset controller, if so, should not we have some > sort of reset handle node and handle that in a generic manner? The Phy Reset is connected to a GPIO, and the ethernet driver has code to switch the GPIO out of reset. The phy is kept in reset until the ethernet device is opened, and Linux is booted with the phy reset asserted. > Is your DTS or DTB hardcoding the PHY id, or are you having your > bootloader detect the exact PHY for you, then putting back the PHY > into reset to save power, until someone uses that PHY again? For our uses the Phy ID is hardcoded. There is only a single part that will fit on the board. So the bootloader doesn't touch the phy. If there were alternate parts we'd get the part kind from the EEPROM that stores the MAC address/etc. > > + while (cplen > 0) { > > + if (sscanf(cp, "ethernet-phy-id%4x.%4x", &upper, &lower) == 2) { > > You might want to guard against 0x0 and 0xffff just in case whoever > fills this information in the Device Tree was reading bogus data out > of the MDIO bus, otherwise, chances are that the "Generic PHY" driver > will be picked up, and it might still not be appropriate for driving > your PHY chip. Having the bootloader read the phy ID just to fill in this compatible string isn't really the point. In every normal case I think it makes sense to let Linux autoprobe the phy id. The use for this compatible string is to defeat the autoprobe for situations where it is not appropriate. Thanks, Jason -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
2014-01-31 Jason Gunthorpe <jgunthorpe@obsidianresearch.com>: > On Fri, Jan 31, 2014 at 02:24:52PM -0800, Florian Fainelli wrote: > >> > This is necessary to support phy's that cannot be autoprobed when >> > of_mdiobus_register is called. Specifically, my case has the phy in reset at >> > of_mdiobus_register, the reset is only released once the ethernet driver >> > starts, before it attaches to the phy. >> >> Who is responsible for bringing the PHY out of reset, is it the >> Ethernet/MDIO bus driver? Is the PHY put into reset using, e.g a GPIO >> line or any sort of reset controller, if so, should not we have some >> sort of reset handle node and handle that in a generic manner? > > The Phy Reset is connected to a GPIO, and the ethernet driver has code > to switch the GPIO out of reset. The phy is kept in reset until the > ethernet device is opened, and Linux is booted with the phy reset > asserted. > >> Is your DTS or DTB hardcoding the PHY id, or are you having your >> bootloader detect the exact PHY for you, then putting back the PHY >> into reset to save power, until someone uses that PHY again? > > For our uses the Phy ID is hardcoded. There is only a single part that > will fit on the board. So the bootloader doesn't touch the phy. If > there were alternate parts we'd get the part kind from the EEPROM that > stores the MAC address/etc. > >> > + while (cplen > 0) { >> > + if (sscanf(cp, "ethernet-phy-id%4x.%4x", &upper, &lower) == 2) { >> >> You might want to guard against 0x0 and 0xffff just in case whoever >> fills this information in the Device Tree was reading bogus data out >> of the MDIO bus, otherwise, chances are that the "Generic PHY" driver >> will be picked up, and it might still not be appropriate for driving >> your PHY chip. > > Having the bootloader read the phy ID just to fill in this compatible > string isn't really the point. In every normal case I think it makes > sense to let Linux autoprobe the phy id. The use for this compatible > string is to defeat the autoprobe for situations where it is not > appropriate. This is well understood, I just think there a few different ways to proceed with your use case here: - you allow for a compatible string to defeat auto-probing like you just did - you tell of_mdiobus_register() to look for a reset phandle and have a reset controller release the PHY from reset before it tries to probe for it, because doing that could avoid doing the PHY out of reset sequence in a few dozen Ethernet drivers The auto-probing bypass logic looks simple enough, and it does not require burying the reset logic behind a reset controller, which uses regmap and a few dozens layers down the road to toggle a bit somewhere in a register, although this might be the desired way to move forward if we want that to be generalized. Anyway: Acked-by: Florian Fainelli <f.fainelli@gmail.com>
On Fri, Jan 31, 2014 at 03:28:10PM -0800, Florian Fainelli wrote: > - you tell of_mdiobus_register() to look for a reset phandle and have > a reset controller release the PHY from reset before it tries to probe > for it, because doing that could avoid doing the PHY out of reset > sequence in a few dozen Ethernet drivers I think this would be interesting, but difficult. Ideally you'd want to keep the phy in reset until an ethernet driver attaches to it, and reset it again once it detaches, so Linux would have to defer probing the phy driver until an attach. Not sure what this would do for power consumption, but reset might be lower power on phy's that don't have an internal power down. > The auto-probing bypass logic looks simple enough, and it does not > require burying the reset logic behind a reset controller, which > uses Right, it closely matches what was happening in board files. FWIW, this system was designed in 2005 for Linux 2.16 and this patch set fell out of porting forward to 3.13 w/ DT. Jason -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/of/of_mdio.c b/drivers/of/of_mdio.c index d5a57a9..c93287e 100644 --- a/drivers/of/of_mdio.c +++ b/drivers/of/of_mdio.c @@ -22,6 +22,30 @@ MODULE_AUTHOR("Grant Likely <grant.likely@secretlab.ca>"); MODULE_LICENSE("GPL"); +/* Extract the clause 22 phy ID from the compatible string of the form + * ethernet-phy-idAAAA.BBBB */ +static int of_get_phy_id(struct device_node *device, u32 *phy_id) +{ + const char *cp; + int cplen, l; + unsigned int upper, lower; + + cp = of_get_property(device, "compatible", &cplen); + if (cp == NULL) + return -EINVAL; + while (cplen > 0) { + if (sscanf(cp, "ethernet-phy-id%4x.%4x", &upper, &lower) == 2) { + *phy_id = ((upper & 0xFFFF) << 16) | (lower & 0xFFFF); + return 0; + } + + l = strlen(cp) + 1; + cp += l; + cplen -= l; + } + return -EINVAL; +} + /** * of_mdiobus_register - Register mii_bus and create PHYs from the device tree * @mdio: pointer to mii_bus structure @@ -36,6 +60,7 @@ int of_mdiobus_register(struct mii_bus *mdio, struct device_node *np) struct device_node *child; const __be32 *paddr; u32 addr; + u32 phy_id; bool is_c45, scanphys = false; int rc, i, len; @@ -81,7 +106,10 @@ int of_mdiobus_register(struct mii_bus *mdio, struct device_node *np) is_c45 = of_device_is_compatible(child, "ethernet-phy-ieee802.3-c45"); - phy = get_phy_device(mdio, addr, is_c45); + if (!is_c45 && !of_get_phy_id(child, &phy_id)) + phy = phy_device_create(mdio, addr, phy_id, 0, NULL); + else + phy = get_phy_device(mdio, addr, is_c45); if (!phy || IS_ERR(phy)) { dev_err(&mdio->dev,
This makes the generic of_mdiobus_register parse the DT compatible string for the pattern ethernet-phy-idAAAA.BBBB. If present it should be a value that matches the phy-id register normally readable through MDIO. When the ID is given the phy autoprobing is defeated and the phy is created directly. This is necessary to support phy's that cannot be autoprobed when of_mdiobus_register is called. Specifically, my case has the phy in reset at of_mdiobus_register, the reset is only released once the ethernet driver starts, before it attaches to the phy. Tested on ARM Kirkwood with phy id 0x01410e90 (Marvell 88E1318) Signed-off-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com> --- drivers/of/of_mdio.c | 30 +++++++++++++++++++++++++++++- 1 file changed, 29 insertions(+), 1 deletion(-)