* Re: serdev: How to attach serdev devices to USB based tty devices?
From: Johan Hovold @ 2018-08-21 14:29 UTC (permalink / raw)
To: Sebastian Reichel
Cc: Andreas Färber, Johan Hovold, Rob Herring,
linux-serial@vger.kernel.org, linux-usb, Linux-MIPS, Xue Liu,
Ben Whitten, devicetree, netdev@vger.kernel.org, Oliver Neukum,
Alexander Graf, LoRa_Community_Support@semtech.com, Jian-Hong Pan,
Stefan Rehm, linux-arm-kernel@lists.infradead.org
In-Reply-To: <20180815182150.wsd5oxlucsox2qig@earth.universe>
On Wed, Aug 15, 2018 at 08:21:50PM +0200, Sebastian Reichel wrote:
> Hi,
>
> +cc Johan Hovold <johan@kernel.org>
>
> Johan told me, that he is working on this at ELCE 2017. Also he is
> the subsystem maintainer of the USB serial subsystem.
I haven't done much work on this; it's more of a low-priority background
task that keeps popping up. ;)
Rob already linked to Ricardo's series in which this was recently
discussed [1][2].
In one of those threads I also posted to some code I've been using to
test serdev with USB-serial devices [3]. There are some known issues
blocking this from being merged (e.g. serdev not supporting hangups and
agreement on DT bindings), but it would otherwise allow you to use
serdev for fixed topologies (i.e. you know beforehand which port you'll
be plugging your USB-serial device into). So that might still be useful
for development purposes as is.
With DT-overlay support this could be extended also to the dynamic case
(e.g. loading overlays from userspace or passing the equivalent data
from a tty driver).
Johan
[1] https://lkml.kernel.org/r/CAPybu_0RRNMsdzv4CKyw922hX3_EF=-LKD_QWZV0DoQmjG0aRQ@mail.gmail.com
[2] https://lkml.kernel.org/r/20180611115240.32606-1-ricardo.ribalda@gmail.com
[3] https://git.kernel.org/pub/scm/linux/kernel/git/johan/usb-serial.git/log/?h=usb-serial-of
^ permalink raw reply
* [PATCH] rds: tcp: remove duplicated include from tcp.c
From: Yue Haibing @ 2018-08-21 14:05 UTC (permalink / raw)
To: Santosh Shilimkar, David S. Miller
Cc: Yue Haibing, netdev, linux-rdma, rds-devel, kernel-janitors
Remove duplicated include.
Signed-off-by: Yue Haibing <yuehaibing@huawei.com>
---
net/rds/tcp.c | 1 -
1 file changed, 1 deletion(-)
diff --git a/net/rds/tcp.c b/net/rds/tcp.c
index 2c7b7c3..b9bbcf3 100644
--- a/net/rds/tcp.c
+++ b/net/rds/tcp.c
@@ -37,7 +37,6 @@
#include <net/tcp.h>
#include <net/net_namespace.h>
#include <net/netns/generic.h>
-#include <net/tcp.h>
#include <net/addrconf.h>
#include "rds.h"
^ permalink raw reply related
* Re: [PATCH v2 06/29] mtd: Add support for reading MTD devices via the nvmem API
From: Boris Brezillon @ 2018-08-21 13:57 UTC (permalink / raw)
To: Srinivas Kandagatla, Rob Herring
Cc: Alban, Bartosz Golaszewski, Jonathan Corbet, Sekhar Nori,
Kevin Hilman, Russell King, Arnd Bergmann, Greg Kroah-Hartman,
David Woodhouse, Brian Norris, Marek Vasut, Richard Weinberger,
Grygorii Strashko, David S . Miller, Naren, Mauro Carvalho Chehab,
Andrew Morton, Lukas Wunner, Dan Carpenter, Florian Fainelli
In-Reply-To: <e4ea4614-4673-020c-641d-30c94e7b761a@linaro.org>
On Tue, 21 Aug 2018 14:37:37 +0100
Srinivas Kandagatla <srinivas.kandagatla@linaro.org> wrote:
> On 21/08/18 14:34, Srinivas Kandagatla wrote:
> >
> >
> > On 21/08/18 12:31, Boris Brezillon wrote:
> >>> * struct nvmem_config - NVMEM device configuration
> >>> @@ -58,6 +62,7 @@ struct nvmem_config {
> >>> bool root_only;
> >>> nvmem_reg_read_t reg_read;
> >>> nvmem_reg_write_t reg_write;
> >>> + nvmem_match_t match;
> >>> int size;
> >>> int word_size;
> >>> int stride;
> >>>
> >> That might work if nvmem cells are defined directly under the mtdnode.
> > Layout should not matter! which is the purpose of this callback.
> >
> > The only purpose of this callback is to tell nvmem core that the
> > node(nvmem cell) belongs to that provider or not, if it is then we
> > successfully found the provider. Its up to the provider on which layout
> > it describes nvmem cells. Additionally the provider can add additional
> > sanity checks in this match function to ensure that cell is correctly
> > represented.
> >
> >
> >> If we go for this approach, I'd recommend replacing this ->match() hook
> >> by ->is_nvmem_cell() and pass it the cell node instead of the nvmem
> >> node, because what we're really after here is knowing which subnode is
> >> an nvmem cell and which subnode is not.
> >
> > I agree on passing cell node instead of its parent. Regarding basic
> > validating if its nvmem cell or not, we can check compatible string in
> > nvmem core if we decide to use "nvmem-cell" compatible.
> >
> > Also just in case if you missed this, nvmem would not iterate the
> Sorry !! i hit send button too quickly I guess.
>
> What I meant to say here, is that nvmem core would not iterate the
> provider node in any case.
>
> Only time it looks at the cell node is when a consumer requests for the
> cell.
I did miss that, indeed. Thanks for the heads up.
So, the "old partitions being considered as nvmem cells" is not really
a problem, because those parts shouldn't be referenced.
This leaves us with the config->force_compat_check topic, which I'd
like to have to ensure that nvmem cells under MTD nodes actually have
compatible = "nvmem-cell" and prevent people from inadvertently
omitting this prop.
And of course, we need Rob's approval on this new binding :-).
^ permalink raw reply
* [PATCH] netfilter: conntrack: remove duplicated include from nf_conntrack_proto_udp.c
From: Yue Haibing @ 2018-08-21 14:03 UTC (permalink / raw)
To: Pablo Neira Ayuso, Jozsef Kadlecsik, Florian Westphal,
David S. Miller
Cc: Yue Haibing, netfilter-devel, coreteam, netdev, kernel-janitors
Remove duplicated include.
Signed-off-by: Yue Haibing <yuehaibing@huawei.com>
---
net/netfilter/nf_conntrack_proto_udp.c | 1 -
1 file changed, 1 deletion(-)
diff --git a/net/netfilter/nf_conntrack_proto_udp.c b/net/netfilter/nf_conntrack_proto_udp.c
index 7a1b898..9272a2c 100644
--- a/net/netfilter/nf_conntrack_proto_udp.c
+++ b/net/netfilter/nf_conntrack_proto_udp.c
@@ -393,4 +393,3 @@ static struct nf_proto_net *udp_get_net_proto(struct net *net)
};
EXPORT_SYMBOL_GPL(nf_conntrack_l4proto_udplite6);
#endif
-#include <net/netfilter/nf_conntrack_timeout.h>
^ permalink raw reply related
* [PATCH] sch_cake: Remove unused including <linux/version.h>
From: Yue Haibing @ 2018-08-21 13:58 UTC (permalink / raw)
To: Jamal Hadi Salim, Cong Wang, Jiri Pirko, David S. Miller
Cc: Yue Haibing, netdev, kernel-janitors
Remove including <linux/version.h> that don't need it.
Signed-off-by: Yue Haibing <yuehaibing@huawei.com>
---
net/sched/sch_cake.c | 1 -
1 file changed, 1 deletion(-)
diff --git a/net/sched/sch_cake.c b/net/sched/sch_cake.c
index 35fc725..4d26b08 100644
--- a/net/sched/sch_cake.c
+++ b/net/sched/sch_cake.c
@@ -64,7 +64,6 @@
#include <linux/vmalloc.h>
#include <linux/reciprocal_div.h>
#include <net/netlink.h>
-#include <linux/version.h>
#include <linux/if_vlan.h>
#include <net/pkt_sched.h>
#include <net/pkt_cls.h>
^ permalink raw reply related
* Re: virtio_net failover and initramfs
From: Harald Hoyer @ 2018-08-21 13:44 UTC (permalink / raw)
To: Samudrala, Sridhar, Siwei Liu
Cc: Jiri Pirko, initramfs, Michael S. Tsirkin, Netdev,
vijay.balakrishna, si-wei liu, liran.alon
In-Reply-To: <132a4610-a59f-19e4-a602-ead91325fb47@intel.com>
On 17.08.2018 21:09, Samudrala, Sridhar wrote:
> On 8/17/2018 2:56 AM, Harald Hoyer wrote:
>> On 17.08.2018 11:51, Harald Hoyer wrote:
>>> On 16.08.2018 00:17, Siwei Liu wrote:
>>>> On Wed, Aug 15, 2018 at 12:05 PM, Samudrala, Sridhar
>>>> <sridhar.samudrala@intel.com> wrote:
>>>>> On 8/14/2018 5:03 PM, Siwei Liu wrote:
>>>>>> Are we sure all userspace apps skip and ignore slave interfaces by
>>>>>> just looking at "IFLA_MASTER" attribute?
>>>>>>
>>>>>> When STANDBY is enabled on virtio-net, a failover master interface
>>>>>> will appear, which automatically enslaves the virtio device. But it is
>>>>>> found out that iSCSI (or any network boot) cannot boot strap over the
>>>>>> new failover interface together with a standby virtio (without any VF
>>>>>> or PT device in place).
>>>>>>
>>>>>> Dracut (initramfs) ends up with timeout and dropping into emergency shell:
>>>>>>
>>>>>> [ 228.170425] dracut-initqueue[377]: Warning: dracut-initqueue
>>>>>> timeout - starting timeout scripts
>>>>>> [ 228.171788] dracut-initqueue[377]: Warning: Could not boot.
>>>>>> Starting Dracut Emergency Shell...
>>>>>> Generating "/run/initramfs/rdsosreport.txt"
>>>>>> Entering emergency mode. Exit the shell to continue.
>>>>>> Type "journalctl" to view system logs.
>>>>>> You might want to save "/run/initramfs/rdsosreport.txt" to a USB stick or
>>>>>> /boot
>>>>>> after mounting them and attach it to a bug report.
>>>>>> dracut:/# ip l sh
>>>>>> 1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN
>>>>>> mode DEFAULT group default qlen 1000
>>>>>> link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
>>>>>> 2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue
>>>>>> state UP mode DEFAULT group default qlen 1000
>>>>>> link/ether 9a:46:22:ae:33:54 brd ff:ff:ff:ff:ff:ff\
>>>>>> 3: eth1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast
>>>>>> master eth0 state UP mode DEFAULT group default qlen 1000
>>>>>> link/ether 9a:46:22:ae:33:54 brd ff:ff:ff:ff:ff:ff
>>>>>> dracut:/#
>>>>>>
>>>>>> If changing dracut code to ignore eth1 (with IFLA_MASTER attr),
>>>>>> network boot starts to work.
>>>>>
>>>>> Does dracut by default tries to use all the interfaces that are UP?
>>>>>
>>>> Yes. The specific dracut cmdline of our case is "ip=dhcp
>>>> netroot=iscsi:... ", but it's not specific to iscsi boot. And because
>>>> of same MAC address for failover and standby, while dracut tries to
>>>> run DHCP on all interfaces that are up it eventually gets same route
>>>> for each interface. Those conflict route entries kill off the network
>>>> connection.
>>>>
>>>>>> The reason is that dracut has its own means to differentiate virtual
>>>>>> interfaces for network boot: it does not look at IFLA_MASTER and
>>>>>> ignores slave interfaces. Instead, users have to provide explicit
>>>>>> option e.g. bond=eth0,eth1 in the boot line, then dracut would know
>>>>>> the config and ignore the slave interfaces.
>>>>>
>>>>> Isn't it possible to specify the interface that should be used for network
>>>>> boot?
>>>> As I understand it, one can only specify interface name for running
>>>> DHCP but not select interface for network boot. We want DHCP to run
>>>> on every NIC that is up (excluding the enslaved interfaces), and only
>>>> one of them can get a route entry to the network boot server (ie.g.
>>>> iSCSI target).
>>>>
>>>>>
>>>>>> However, with automatic creation of failover interface that assumption
>>>>>> is no longer true. Can we change dracut to ignore all slave interface
>>>>>> by checking IFLA_MASTER? I don't think so. It has a large impact to
>>>>>> existing configs.
>>>>>
>>>>> What is the issue with checking for IFLA_MASTER? I guess this is used with
>>>>> team/bonding setups.
>>>> That should be discussed within and determined by the dracut
>>>> community. But the current dracut code doesn't check IFLA_MASTER for
>>>> team or bonding specifically. I guess this change might have broader
>>>> impact to existing userspace that might be already relying on the
>>>> current behaviour.
>>>>
>>>> Thanks,
>>>> -Siwei
>>> Is there a sysfs flag for IFF_SLAVE? Or any "ip" output I can use to detect, that it is a IFF_SLAVE?
>>>
>> Oh, it's the other way around.. dracut should ignore "master" (eth1).
> In the above example eth0 is the net_failover device and eth1 is the lower virtio_net device.
> "ip" output of eth1 shows "master eth0". It indicates that eth0 is its upper/master device.
> This information can also be obtained via sysfs too. /sys/class/net/eth1/upper_eth0
>>
>> Can the master enslave the "eth0", if it is already "UP" and busy later on?
> eth0 is the master/failover device and eth1 gets registered as its slave via NETDEV_REGISTER event.
> dracut should ignore eth1 in this setup.
Care to test, if that fixes your case?
https://github.com/dracutdevs/dracut/pull/450/files
^ permalink raw reply
* RE: [PATCH 1/2] staging: rtl8192e: Fix compiler warning about strncpy
From: David Laight @ 2018-08-21 13:41 UTC (permalink / raw)
To: 'Larry Finger', gregkh@linuxfoundation.org
Cc: netdev@vger.kernel.org, devel@driverdev.osuosl.org
In-Reply-To: <20180820175124.23863-2-Larry.Finger@lwfinger.net>
From: Larry Finger
> Sent: 20 August 2018 18:51
> When strncpy() is called with source and destination strings the same
> length, gcc 8 warns that there may be an unterminated string. Using
> strlcpy() rather than strncpy() forces a null at the end and quiets the
> warning.
>
> Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
> ---
> drivers/staging/rtl8192e/rtllib_softmac.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/staging/rtl8192e/rtllib_softmac.c b/drivers/staging/rtl8192e/rtllib_softmac.c
> index 919231fec09c..95a8390cb7ac 100644
> --- a/drivers/staging/rtl8192e/rtllib_softmac.c
> +++ b/drivers/staging/rtl8192e/rtllib_softmac.c
> @@ -1684,14 +1684,14 @@ inline void rtllib_softmac_new_net(struct rtllib_device *ieee,
> * essid provided by the user.
> */
> if (!ssidbroad) {
> - strncpy(tmp_ssid, ieee->current_network.ssid,
> + strlcpy(tmp_ssid, ieee->current_network.ssid,
> IW_ESSID_MAX_SIZE);
> tmp_ssid_len = ieee->current_network.ssid_len;
If there is a length, why not use it?
Depending on where the data came from the length might need validating.
Depending on how tmp_ssid is used it might need zero filling.
> }
> memcpy(&ieee->current_network, net,
> sizeof(struct rtllib_network));
Gah - should be sizeof(ieee->current_network).
Or better still a structure assignment.
> if (!ssidbroad) {
> - strncpy(ieee->current_network.ssid, tmp_ssid,
> + strlcpy(ieee->current_network.ssid, tmp_ssid,
> IW_ESSID_MAX_SIZE);
> ieee->current_network.ssid_len = tmp_ssid_len;
Hmmm... this looks like it is restoring the fields.
So why not have a temporary ssid buffer that is the size of the
actual buffer and user memcpy().
David
-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)
^ permalink raw reply
* Re: [PATCH v2 06/29] mtd: Add support for reading MTD devices via the nvmem API
From: Srinivas Kandagatla @ 2018-08-21 13:37 UTC (permalink / raw)
To: Boris Brezillon
Cc: Alban, Bartosz Golaszewski, Jonathan Corbet, Sekhar Nori,
Kevin Hilman, Russell King, Arnd Bergmann, Greg Kroah-Hartman,
David Woodhouse, Brian Norris, Marek Vasut, Richard Weinberger,
Grygorii Strashko, David S . Miller, Naren, Mauro Carvalho Chehab,
Andrew Morton, Lukas Wunner, Dan Carpenter, Florian Fainelli
In-Reply-To: <6fb36da4-c985-6d6e-f9e1-572f5cd7609b@linaro.org>
On 21/08/18 14:34, Srinivas Kandagatla wrote:
>
>
> On 21/08/18 12:31, Boris Brezillon wrote:
>>> * struct nvmem_config - NVMEM device configuration
>>> @@ -58,6 +62,7 @@ struct nvmem_config {
>>> bool root_only;
>>> nvmem_reg_read_t reg_read;
>>> nvmem_reg_write_t reg_write;
>>> + nvmem_match_t match;
>>> int size;
>>> int word_size;
>>> int stride;
>>>
>> That might work if nvmem cells are defined directly under the mtdnode.
> Layout should not matter! which is the purpose of this callback.
>
> The only purpose of this callback is to tell nvmem core that the
> node(nvmem cell) belongs to that provider or not, if it is then we
> successfully found the provider. Its up to the provider on which layout
> it describes nvmem cells. Additionally the provider can add additional
> sanity checks in this match function to ensure that cell is correctly
> represented.
>
>
>> If we go for this approach, I'd recommend replacing this ->match() hook
>> by ->is_nvmem_cell() and pass it the cell node instead of the nvmem
>> node, because what we're really after here is knowing which subnode is
>> an nvmem cell and which subnode is not.
>
> I agree on passing cell node instead of its parent. Regarding basic
> validating if its nvmem cell or not, we can check compatible string in
> nvmem core if we decide to use "nvmem-cell" compatible.
>
> Also just in case if you missed this, nvmem would not iterate the
Sorry !! i hit send button too quickly I guess.
What I meant to say here, is that nvmem core would not iterate the
provider node in any case.
Only time it looks at the cell node is when a consumer requests for the
cell.
--srini
^ permalink raw reply
* Re: [PATCH 1/4] net: macb: Fix regression breaking non-MDIO fixed-link PHYs
From: Andrew Lunn @ 2018-08-21 13:34 UTC (permalink / raw)
To: Ahmad Fatoum
Cc: David S. Miller, Nicolas Ferre, kernel, netdev, mdf, Brad Mouring,
Florian Fainelli
In-Reply-To: <77901074-bb78-5860-d6bc-00a1826de8a6@pengutronix.de>
> I've traced it some more: While mdiobus_register fails to find a PHY,
> creation of the "MDIO" bus is still successful and it returns successfully,
> having claimed the reset GPIO.
>
> of_phy_fixed_link_register tries to claim the same GPIO and fails.
I don't see where this is happening. It is looking for a gpio called
'link-gpios'.
> But regardless, there shouldn't have been an of_mdiobus_register and a MDIO bus probe
> before registering the fixed-link in the first place and my patch remedies that.
There are cases where you need both, e.g a switch controller over
MDIO. So we cannot make it one or the other.
However, there are currently no such boards. So far "net", lets go
with your partial revert patch. But we also need patches for
"net-next" which put it back again, allows both fixed-link and an
mdiobus inside a subnode, and not break backwards compatibility.
Andrew
^ permalink raw reply
* Re: [PATCH v2 06/29] mtd: Add support for reading MTD devices via the nvmem API
From: Srinivas Kandagatla @ 2018-08-21 13:34 UTC (permalink / raw)
To: Boris Brezillon
Cc: Alban, Bartosz Golaszewski, Jonathan Corbet, Sekhar Nori,
Kevin Hilman, Russell King, Arnd Bergmann, Greg Kroah-Hartman,
David Woodhouse, Brian Norris, Marek Vasut, Richard Weinberger,
Grygorii Strashko, David S . Miller, Naren, Mauro Carvalho Chehab,
Andrew Morton, Lukas Wunner, Dan Carpenter, Florian Fainelli
In-Reply-To: <20180821133136.1fada1b6@bbrezillon>
On 21/08/18 12:31, Boris Brezillon wrote:
>> * struct nvmem_config - NVMEM device configuration
>> @@ -58,6 +62,7 @@ struct nvmem_config {
>> bool root_only;
>> nvmem_reg_read_t reg_read;
>> nvmem_reg_write_t reg_write;
>> + nvmem_match_t match;
>> int size;
>> int word_size;
>> int stride;
>>
> That might work if nvmem cells are defined directly under the mtdnode.
Layout should not matter! which is the purpose of this callback.
The only purpose of this callback is to tell nvmem core that the
node(nvmem cell) belongs to that provider or not, if it is then we
successfully found the provider. Its up to the provider on which layout
it describes nvmem cells. Additionally the provider can add additional
sanity checks in this match function to ensure that cell is correctly
represented.
> If we go for this approach, I'd recommend replacing this ->match() hook
> by ->is_nvmem_cell() and pass it the cell node instead of the nvmem
> node, because what we're really after here is knowing which subnode is
> an nvmem cell and which subnode is not.
I agree on passing cell node instead of its parent. Regarding basic
validating if its nvmem cell or not, we can check compatible string in
nvmem core if we decide to use "nvmem-cell" compatible.
Also just in case if you missed this, nvmem would not iterate the
--srini
^ permalink raw reply
* Re: serdev: How to attach serdev devices to USB based tty devices?
From: Frank Kunz @ 2018-08-21 16:32 UTC (permalink / raw)
To: Andreas Färber, Rob Herring, linux-serial@vger.kernel.org,
linux-usb
Cc: Linux-MIPS, Xue Liu, Ben Whitten, devicetree,
netdev@vger.kernel.org, Oliver Neukum, Alexander Graf,
LoRa_Community_Support@semtech.com, Jian-Hong Pan, Stefan Rehm,
linux-arm-kernel@lists.infradead.org
In-Reply-To: <3639955d-5990-1c82-7158-ac07b33c41f2@suse.de>
Am 14.08.2018 um 04:28 schrieb Andreas Färber:
> Hi Rob et al.,
>
> For my LoRa network driver project [1] I have found your serdev
> framework to be a valuable help for dealing with hardware modules
> exposing some textual or binary UART interface.
>
> In particular on arm(64) and mips this allows to define an unlimited
> number of serdev drivers [2] that are associated via their Device Tree
> compatible string and can optionally be configured via DT properties.
>
> And in theory it seems serdev has also grown support for ACPI.
>
> Now, a growing number of vendors are placing such modules on a USB stick
> for easy evaluation on x86_64 PC hardware, or are designing mPCIe or M.2
> cards using their USB pins. While I do not yet have access to such a
> device myself, it is my understanding that devices with USB-UART bridge
> chipsets (e.g., FTDI) will show up as /dev/ttyUSBx and devices with an
> MCU implementing the CDC USB protocol (e.g., Pico-cell gateway = picoGW)
> will show up as /dev/ttyACMx.
> On the Raspberry Pi I've seen that Device Tree nodes can be used to pass
> information to on-board devices such as MAC address to Ethernet chipset,
> but that does not seem all that useful for passing a serdev child node
> to hot-plugged devices at unpredictable hub/port location (where it
> should not interfere with regular USB-UART cables for debugging), nor
> would it help ACPI based platforms such as x86_64.
>
> My idea then was that if we had some unique criteria like vendor and
> product IDs (or whatever is supported in usb_device_id), we could write
> a usb_driver with suitable USB_DEVICE*() macro. In its probe function we
> could call into the existing tty driver's probe function and afterwards
> try creating and attaching the appropriate serdev device, i.e. a fixed
> USB-to-serdev driver mapping. Problem is that most devices don't seem to
> implement any unique identifier I could make this depend on - either by
> using a standard FT232/FT2232/CH340G chip or by using STMicroelectronics
> virtual com port identifiers in CDC firmware and only differing in the
> textual description [3] the usb_device_id does not seem to match on.
>
> The obvious solution would of course be if hardware vendors could revise
> their designs to configure FTDI/etc. chips uniquely. I hear that that
> may involve exchanging the chipset, increasing costs, and may impact
> existing drivers. Wouldn't help for devices out there today either.
They need to put an extra eeprom (cents) into their design and program it.
>
> For the picoGW CDC firmware, Semtech does appear to own a USB vendor ID,
> so it would seem possible to allocate their own product IDs for SX1301
> and SX1308 respectively to replace the generic STMicroelectronics IDs,
> which the various vendors could offer as firmware updates.
>
> All outside my control though.
>
> Oliver therefore suggested to not mess with USB drivers and instead use
> a line discipline (ldisc). It seems that for example the userspace tool
> slattach takes a tty device and performs an ioctl to switch the generic
> tty device into a special N_SLIP protocol mode, implemented in [4].
>
> However, the existing number of such ldisc modes appears to be below 30,
> with hardly any vendor-specific implementation, so polluting its number
> space seems undesirable? And in some cases I would like to use the same
> protocol implementation over direct UART and over USB, so would like to
> avoid duplicate serdev_device_driver and tty_ldisc_ops implementations.
>
> Long story short, has there been any thinking about a userspace
> interface to attach a given serdev driver to a tty device?
>
> Or is there, on OF_DYNAMIC platforms, a way from userspace to associate
> a DT fragment (!= DT Overlay) with a given USB device dynamically, to
> attach a serdev node with sub-nodes?
>
> Any other ideas how to cleanly solve this?
>
> In some cases we're talking about a "simple" AT-like command interface;
> the picoGW implements a semi-generic USB-SPI bridge that may host a
> choice of 2+ chipsets, which in turn has two further sub-devices with 3+
> chipset choices (theoretically clk output and rx/tx options etc.) each.
> (For the latter I'm thinking we'll need a serdev driver exposing a
> regmap_bus and then implement regmap_bus based versions of the SPI
> drivers like Ben and I refactored SX1257 in [2] last weekend.)>
There is a mPCIe module (RAK833) available by RAK wireless that uses a
FT2232 as USB-SPI bridge, not uart. I have one here for experiments. It
is detected as generic FT2232 device on usb. As far as I understood so
far the serdev does only support uart based communication, is there a
chance to get USB-SPI bridged modules also working?
Br,
Frank
> Thanks,
> Andreas
>
> [1] https://patchwork.ozlabs.org/cover/937545/
> [2]
> https://git.kernel.org/pub/scm/linux/kernel/git/afaerber/linux-lora.git/tree/drivers/net/lora?h=lora-next
> [3]
> https://github.com/Lora-net/picoGW_mcu/blob/master/src/usb_cdc/Src/usbd_desc.cpp#L59
> [4]
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/net/slip/slip.c#n1281
>
_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
^ permalink raw reply
* Re: [PATCH v2 06/29] mtd: Add support for reading MTD devices via the nvmem API
From: Boris Brezillon @ 2018-08-21 13:01 UTC (permalink / raw)
To: Srinivas Kandagatla
Cc: Alban, Bartosz Golaszewski, Jonathan Corbet, Sekhar Nori,
Kevin Hilman, Russell King, Arnd Bergmann, Greg Kroah-Hartman,
David Woodhouse, Brian Norris, Marek Vasut, Richard Weinberger,
Grygorii Strashko, David S . Miller, Naren, Mauro Carvalho Chehab,
Andrew Morton, Lukas Wunner, Dan Carpenter, Florian Fainelli
In-Reply-To: <6db14f9c-edd3-5e43-839c-953bb03097ff@linaro.org>
On Tue, 21 Aug 2018 13:00:04 +0100
Srinivas Kandagatla <srinivas.kandagatla@linaro.org> wrote:
> On 21/08/18 12:39, Alban wrote:
> > However we still have the a potential address space clash between the
> > nvmem cells and the main device binding.
> Can you elaborate?
>
Yes, I'd be interested in having a real example too, cause I don't see
what this address space clash is.
^ permalink raw reply
* [PATCH] libbpf: Remove the duplicate checking of function storage
From: Taeung Song @ 2018-08-21 16:12 UTC (permalink / raw)
To: Daniel Borkmann, Alexei Starovoitov; +Cc: netdev, linux-kernel, Jakub Kicinski
After the commit eac7d84519a3 ("tools: libbpf: don't return '.text'
as a program for multi-function programs"), bpf_program__next()
in bpf_object__for_each_program skips the function storage such as .text,
so eliminate the duplicate checking.
Cc: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Taeung Song <treeze.taeung@gmail.com>
---
tools/lib/bpf/libbpf.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
index 2abd0f112627..8476da7f2720 100644
--- a/tools/lib/bpf/libbpf.c
+++ b/tools/lib/bpf/libbpf.c
@@ -2336,7 +2336,7 @@ int bpf_prog_load_xattr(const struct bpf_prog_load_attr *attr,
bpf_program__set_expected_attach_type(prog,
expected_attach_type);
- if (!bpf_program__is_function_storage(prog, obj) && !first_prog)
+ if (!first_prog)
first_prog = prog;
}
--
2.17.1
^ permalink raw reply related
* Re: [PATCH v2 06/29] mtd: Add support for reading MTD devices via the nvmem API
From: Alban @ 2018-08-21 12:27 UTC (permalink / raw)
To: Boris Brezillon
Cc: Alban Bedel, Srinivas Kandagatla, Bartosz Golaszewski,
Jonathan Corbet, Sekhar Nori, Kevin Hilman, Russell King,
Arnd Bergmann, Greg Kroah-Hartman, David Woodhouse, Brian Norris,
Marek Vasut, Richard Weinberger, Grygorii Strashko,
David S . Miller, Naren, Mauro Carvalho Chehab, Andrew Morton,
Lukas Wunner, Dan
In-Reply-To: <20180821074404.23aaeb6b@bbrezillon>
[-- Attachment #1: Type: text/plain, Size: 11338 bytes --]
On Tue, 21 Aug 2018 07:44:04 +0200
Boris Brezillon <boris.brezillon@bootlin.com> wrote:
> On Tue, 21 Aug 2018 00:53:27 +0200
> Alban <albeu@free.fr> wrote:
>
> > On Sun, 19 Aug 2018 18:46:09 +0200
> > Boris Brezillon <boris.brezillon@bootlin.com> wrote:
> >
> > > On Sun, 19 Aug 2018 13:31:06 +0200
> > > Alban <albeu@free.fr> wrote:
> > >
> > > > On Fri, 17 Aug 2018 18:27:20 +0200
> > > > Boris Brezillon <boris.brezillon@bootlin.com> wrote:
> > > >
> > > > > Hi Bartosz,
> > > > >
> > > > > On Fri, 10 Aug 2018 10:05:03 +0200
> > > > > Bartosz Golaszewski <brgl@bgdev.pl> wrote:
> > > > >
> > > > > > From: Alban Bedel <albeu@free.fr>
> > > > > >
> > > > > > Allow drivers that use the nvmem API to read data stored on
> > > > > > MTD devices. For this the mtd devices are registered as
> > > > > > read-only NVMEM providers.
> > > > > >
> > > > > > Signed-off-by: Alban Bedel <albeu@free.fr>
> > > > > > [Bartosz:
> > > > > > - use the managed variant of nvmem_register(),
> > > > > > - set the nvmem name]
> > > > > > Signed-off-by: Bartosz Golaszewski
> > > > > > <bgolaszewski@baylibre.com>
> > > > >
> > > > > What happened to the 2 other patches of Alban's series? I'd
> > > > > really like the DT case to be handled/agreed on in the same
> > > > > patchset, but IIRC, Alban and Srinivas disagreed on how this
> > > > > should be represented. I hope this time we'll come to an
> > > > > agreement, because the MTD <-> NVMEM glue has been floating
> > > > > around for quite some time...
> > > >
> > > > These other patches were to fix what I consider a fundamental
> > > > flaw in the generic NVMEM bindings, however we couldn't agree
> > > > on this point. Bartosz later contacted me to take over this
> > > > series and I suggested to just change the MTD NVMEM binding to
> > > > use a compatible string on the NVMEM cells as an alternative
> > > > solution to fix the clash with the old style MTD partition.
> > > >
> > > > However all this has no impact on the code needed to add NVMEM
> > > > support to MTD, so the above patch didn't change at all.
> > >
> > > It does have an impact on the supported binding though.
> > > nvmem->dev.of_node is automatically assigned to mtd->dev.of_node,
> > > which means people will be able to define their NVMEM cells
> > > directly under the MTD device and reference them from other nodes
> > > (even if it's not documented), and as you said, it conflict with
> > > the old MTD partition bindings. So we'd better agree on this
> > > binding before merging this patch.
> >
> > Unless the nvmem cell node has a compatible string, then it won't be
> > considered as a partition by the MTD code. That is were the clash
> > is, both bindings allow free named child nodes without a compatible
> > string.
>
> Except the current nvmem cells parsing code does not enforce that, and
> existing DTs rely on this behavior, so we're screwed. Or are you
> suggesting to add a new "bool check_cells_compat;" field to
> nvmem_config?
There is no nvmem cell parsing at the moment. The DT lookup just
resolve the phandle to the cell node, take the parent node and search
for the nvmem provider that has this OF node. So extending it in case
the node has a *new* compatible string would not break users of the old
binding, none of them has a compatible string.
> >
> > > I see several options:
> > >
> > > 1/ provide a way to tell the NVMEM framework not to use
> > > parent->of_node even if it's != NULL. This way we really don't
> > > support defining NVMEM cells in the DT, and also don't support
> > > referencing the nvmem device using a phandle.
> >
> > I really don't get what the point of this would be. Make the whole
> > API useless?
>
> No, just allow Bartosz to get his changes merged without waiting for
> you and Srinivas to agree on how to handle the new binding. As I said
> earlier, this mtd <-> nvmem stuff has been around for quite some time,
> and instead of trying to find an approach that makes everyone happy,
> you decided to let the patchset die.
As long as that wouldn't prevent using DT in the future I'm fine with
it.
> >
> > > 2/ define a new binding where all nvmem-cells are placed in an
> > > "nvmem" subnode (just like we have this "partitions" subnode
> > > for partitions), and then add a config->of_node field so that the
> > > nvmem provider can explicitly specify the DT node representing
> > > the nvmem device. We'll also need to set this field to
> > > ERR_PTR(-ENOENT) in case this node does not exist so that the
> > > nvmem framework knows that it should not assign
> > > nvmem->dev.of_node to parent->of_node
> >
> > This is not good. First the NVMEM device is only a virtual concept
> > of the Linux kernel, it has no place in the DT.
>
> nvmem-cells is a virtual concept too, still, you define them in the
> DT.
To be honest I also think that naming this concept "nvmem" in the DT was
a bad idea. Perhaps something like "driver-data" or "data-cell" would
have been better as that would make it clear what this is about, nvmem
is just the Linux implementation of this concept.
> > Secondly the NVMEM
> > provider (here the MTD device) then has to manually parse its DT
> > node to find this subnode, pass it to the NVMEM framework to later
> > again resolve it back to the MTD device.
>
> We don't resolve it back to the MTD device, because the MTD device is
> just the parent of the nvmem device.
>
> > Not very complex but still a lot of
> > useless code, just registering the MTD device is a lot simpler and
> > much more inline with most other kernel API that register a
> > "service" available from a device.
>
> I'm not a big fan of this option either, but I thought I had to
> propose it.
>
> >
> > > 3/ only declare partitions as nvmem providers. This would solve
> > > the problem we have with partitions defined in the DT since
> > > defining sub-partitions in the DT is not (yet?) supported and
> > > partition nodes are supposed to be leaf nodes. Still, I'm not
> > > a big fan of this solution because it will prevent us from
> > > supporting sub-partitions if we ever want/need to.
> >
> > That sound like a poor workaround.
>
> Yes, that's a workaround. And the reason I propose it, is, again,
> because I don't want to block Bartosz.
>
> > Remember that this problem could
> > appear with any device that has a binding that use child nodes.
>
> I'm talking about partitions, and you're talking about mtd devices.
> Right now partitions don't have subnodes, and if we define that
> partition subnodes should describe nvmem-cells, then it becomes part
> of the official binding. So, no, the problem you mention does not
> (yet) exist.
That would add another binding that allow free named child nodes
without compatible string although experience has repeatedly shown that
this was a bad idea.
> >
> > > 4/ Add a ->of_xlate() hook that would be called if present by the
> > > framework instead of using the default parsing we have right
> > > now.
> >
> > That is a bit cleaner, but I don't think it would be worse the
> > complexity.
>
> But it's way more flexible than putting everything in the nvmem
> framework. BTW, did you notice that nvmem-cells parsing does not work
> with flashes bigger than 4GB, because the framework assumes
> #address-cells and #size-cells are always 1. That's probably something
> we'll have to fix for the MTD case.
Yes, however that's just an implementation limitation which is trivial
to solve.
> > Furthermore xlate functions are more about converting
> > from hardware parameters to internal kernel representation than to
> > hide extra DT parsing.
>
> Hm, how is that different? ->of_xlate() is just a way for drivers to
> have their own DT representation, which is exactly what we want here.
There is a big difference. DT represent the hardware and the
relationship between the devices in an OS independent format. We don't
add extra stuff in there just to map back internal Linux API details.
> >
> > > 5/ Tell the nvmem framework the name of the subnode containing
> > > nvmem cell definitions (if NULL that means cells are directly
> > > defined under the nvmem provider node). We would set it to
> > > "nvmem-cells" (or whatever you like) for the MTD case.
> >
> > If so please match on compatible and not on the node name.
>
> If you like.
>
> >
> > 6/ Extend the current NVMEM cell lookup to check if the parent node
> > of the cell has a compatible string set to "nvmem-cells". If it
> > doesn't it mean we have the current binding and this node is the
> > NVMEM device. If it does the device node is just the next parent.
> > This is trivial to implement (literally 2 lines of code) and cover
> > all the cases currently known.
>
> Except Srinivas was not happy with this solution, and this stalled the
> discussion. I'm trying to find other options and you keep rejecting
> all of them to come back to this one.
Well, I think this is the best solution :/
> >
> > 7/ Just add a compatible string to the nvmem cell. No code change is
> > needed,
>
> That's not true!!!
What is not true in this statement? The current nvmem lookup don't care
about compatible strings, so the cell lookup would just works fine. The
MTD partition parser won't consider them as a partition because of the
compatible string. Problem solved!
> What forces people to add this compatible in their
> DT? Nothing. I'll tell you what will happen: people will start
> defining their nvmem cells directly under the MTD node because that
> *works*, and even if the binding is not documented and we consider it
> invalid, we'll be stuck supporting it forever.
Do note that undocumented bindings are not allowed. DTS that use
undocumented bindings (normally) just get rejected.
> As said above, the
> very reason for option #1 to exist is to give you and Srinivas some
> more time to sort this out, while unblocking Bartosz in the meantime.
I'm fine with #1, I just didn't understood what it was useful for.
> > however as the nvmem cells have an address space (the offset in
> > byte in the storage) it might still clash with another address space
> > used by the main device biding (for example a number of child
> > functions).
> >
> > > There are probably other options (some were proposed by Alban and
> > > Srinivas already), but I'd like to get this sorted out before we
> > > merge this patch.
> > >
> > > Alban, Srinivas, any opinion?
> >
> > My preference goes to 6/ as it is trivial to implement, solves all
> > known shortcomings and is backward compatible with the current
> > binding. All other solutions have limitations and/or require too
> > complex implementations compared to what they try to solve.
>
> So we're back to square 1, and you're again blocking everything
> because you refuse to consider other options.
As I'm not a maintainer so I just can't block anything. But I won't lie
and pretend that I support a solution with known shortcomings.
Alban
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]
^ permalink raw reply
* Re: [PATCH v2 06/29] mtd: Add support for reading MTD devices via the nvmem API
From: Srinivas Kandagatla @ 2018-08-21 12:00 UTC (permalink / raw)
To: Alban
Cc: Boris Brezillon, Bartosz Golaszewski, Jonathan Corbet,
Sekhar Nori, Kevin Hilman, Russell King, Arnd Bergmann,
Greg Kroah-Hartman, David Woodhouse, Brian Norris, Marek Vasut,
Richard Weinberger, Grygorii Strashko, David S . Miller, Naren,
Mauro Carvalho Chehab, Andrew Morton, Lukas Wunner, Dan Carpenter,
Flori
In-Reply-To: <20180821133916.3a1c51b1@eos>
On 21/08/18 12:39, Alban wrote:
> However we still have the a potential address space clash between the
> nvmem cells and the main device binding.
Can you elaborate?
--srini
^ permalink raw reply
* Linux Plumbers BPF micro-conference CFP
From: Daniel Borkmann @ 2018-08-21 15:11 UTC (permalink / raw)
To: netdev-u79uwXL29TY76Z2rM5mHXA,
iovisor-dev-9jONkmmOlFHEE9lA1F8Ukti2O/JbrIOy,
xdp-newbies-u79uwXL29TY76Z2rM5mHXA,
containers-cunTk1MwBs98uUxBSJOaYoYkZiVZrdSR2LY78lusg7I,
linux-kernel-u79uwXL29TY76Z2rM5mHXA
Cc: lpc-bpf-u79uwXL29TY76Z2rM5mHXA,
alexei.starovoitov-Re5JQEeQqe8AvxtiuMwx3w
This is a call for proposals for the BPF micro-conference at the
Linux Plumbers Conference (LPC) 2018 which will be held in Vancouver,
Canada from the 13th to the 15th of November, co-located with the
Linux Kernel Summit.
The goal of the BPF micro-conference is to bring BPF developers
together to discuss topics around Linux kernel work related to
the BPF core infrastructure as well as its many subsystems under
tracing, networking, security, and BPF user space tooling. The
format of the micro-conference has a main focus on discussion,
therefore each accepted topic will provide a short 1-2 slide
introduction with subsequent discussion for the rest of the
allocated time slot. The expected time for one discussion slot is
approximately 15 min. The whole BPF micro-conference is a 3 hours
long session which will run on the third day of LPC (so that it
does not overlap with the LPC's Networking Track).
The BPF micro-conference is a community-driven event and open to
all LPC attendees, there is no additional registration required.
Please submit your discussion proposals to the LPC BPF micro-conference
organizers at:
lpc-bpf-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
Proposals must be submitted until October 1st, and submitters will
be notified of acceptance by October 5th. (Please note that proposals
must not be sent as html mail as they are otherwise dropped by vger.)
The format of the submission and many other details can be found at:
http://vger.kernel.org/lpc-bpf.html
Looking forward to seeing you all in Vancouver!
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#1460): https://lists.iovisor.org/g/iovisor-dev/message/1460
Mute This Topic: https://lists.iovisor.org/mt/24877036/1132507
Group Owner: iovisor-dev+owner-9jONkmmOlFHEE9lA1F8Ukti2O/JbrIOy@public.gmane.org
Unsubscribe: https://lists.iovisor.org/g/iovisor-dev/unsub [glki-iovisor-dev@m.gmane.org]
-=-=-=-=-=-=-=-=-=-=-=-
^ permalink raw reply
* Re: [PATCH] bluetooth/{bnep,cmtp,hidp}: Remove unnecessary smp_mb__{before,after}_atomic
From: Marcel Holtmann @ 2018-08-21 14:55 UTC (permalink / raw)
To: Andrea Parri
Cc: Johan Hedberg, David S. Miller, linux-bluetooth, netdev,
linux-kernel, Jeffy Chen, Brian Norris, AL Yu-Chen Cho
In-Reply-To: <1534272066-9366-1-git-send-email-andrea.parri@amarulasolutions.com>
Hi Andrea,
> The barriers are unneeded; wait_woken() and woken_wake_function()
> already provide us with the required synchronization: remove them
> and document that we're relying on the (implicit) synchronization
> provided by wait_woken() and woken_wake_function().
>
> Signed-off-by: Andrea Parri <andrea.parri@amarulasolutions.com>
> ---
> net/bluetooth/bnep/core.c | 7 ++++---
> net/bluetooth/cmtp/core.c | 14 ++++++++------
> net/bluetooth/hidp/core.c | 13 ++++++++-----
> 3 files changed, 20 insertions(+), 14 deletions(-)
patch has been applied to bluetooth-next tree.
Regards
Marcel
^ permalink raw reply
* Re: [PATCH] strparser: remove any offset before parsing messages
From: Doron Roberts-Kedes @ 2018-08-21 14:53 UTC (permalink / raw)
To: Dominique Martinet
Cc: Tom Herbert, Dave Watson, David S. Miller, netdev, linux-kernel
In-Reply-To: <1534855906-22870-1-git-send-email-asmadeus@codewreck.org>
On Tue, Aug 21, 2018 at 02:51:46PM +0200, Dominique Martinet wrote:
> Offset is not well handled by strparser users right now.
>
> Out of the current strparser users, we have:
> - tls, that handles offset properly in parse and rcv callbacks
> - kcm, that handles offset in rcv but not in parse
> - bpf sockmap, that does not seem to handle offset anywhere
>
> Calling pskb_pull() on new skb ensures that the offset will be 0
> everywhere in practice, and in particular for the parse function,
> unless the user modifies it themselves like tls does.
>
> This fixes a bug which can be exhibited by implementing a simple kcm
> parser that looks for the packet size in the first word of the packet,
> and sending two such packets in a single write() call on the other side:
> the second message will be cut at the length of the first message.
> Since this is a stream protocol, all the following messages will also
> be corrupt since it will start looking for the next offset at a wrong
> position.
>
> Signed-off-by: Dominique Martinet <asmadeus@codewreck.org>
> ---
There are a few issues with this patch. First, it seems like you're
trying to fix bugs in users of strparser by changing an implementation
detail of strparser. Second, this implementation change can add malloc's
and copies where there were none before.
If strparser users do not handle non-zero offset properly, then that
doesn't motivate changing the implementation of strparser to copy
around data to accomodate those buggy users.
Why not submit a patch that handles offset properly in the code you
pointed out?
^ permalink raw reply
* Re: [RFC PATCH net-next v2 10/17] ethtool: implement GET_SETTINGS message
From: Michal Kubecek @ 2018-08-21 14:52 UTC (permalink / raw)
To: Andrew Lunn
Cc: netdev, linux-kernel, Jiri Pirko, David Miller, Florian Fainelli,
Roopa Prabhu, Jakub Kicinski, John W. Linville
In-Reply-To: <20180821141022.GD2985@lunn.ch>
On Tue, Aug 21, 2018 at 04:10:22PM +0200, Andrew Lunn wrote:
> On Tue, Aug 21, 2018 at 11:32:46AM +0200, Michal Kubecek wrote:
> > I have prepared a patch converting 8390/etherh driver to use
> > {g,s}et_link_ksettings and I'm going to submit it when net-next opens.
> > Do you think we can then drop {g,s}et_settings callbacks completely
> > (i.e. also from ioctl() code and ethtool_ops)? Do we care about
> > unconverted out of tree drivers?
>
> We cannot break ethtool, the ABI it uses. But there is already code to
> use get_link_ksettings() and only fall back to get_settings if it does
> not exist. So we can clean up all the fallback code, remove the
> ethtool_ops, etc.
Yes, that's what I have in mind: keep the ethtool interface as is (both
ETHTOOL_{G,S}SET and ETHTOOL_{G,S}LINKSETTINGS) but drop get_settings
and set_settings from ethtool_ops and the code that falls back to them.
This way both old and new versions of ethtool would still work and the
only problem would be with out of tree drivers not converted to
{s,g}et_link_ksettings (which won't build so that it won't be silent
breakage).
Michal Kubecek
^ permalink raw reply
* Re: [PATCH v2 06/29] mtd: Add support for reading MTD devices via the nvmem API
From: Boris Brezillon @ 2018-08-21 14:26 UTC (permalink / raw)
To: Alban
Cc: Srinivas Kandagatla, Bartosz Golaszewski, Jonathan Corbet,
Sekhar Nori, Kevin Hilman, Russell King, Arnd Bergmann,
Greg Kroah-Hartman, David Woodhouse, Brian Norris, Marek Vasut,
Richard Weinberger, Grygorii Strashko, David S . Miller, Naren,
Mauro Carvalho Chehab, Andrew Morton, Lukas Wunner, Dan Carpenter
In-Reply-To: <20180821155706.2adf3b4c@eos>
On Tue, 21 Aug 2018 15:57:06 +0200
Alban <albeu@free.fr> wrote:
>
> That would only be needed if the NVMEM framework would do "forward"
> parsing, creating data structure for each NVMEM cell found under an
> NVMEM provider. However currently it doesn't do that and only goes
> "backward", starting by resolving a phandle pointing to a cell, then
> finding the provider that the cell belongs to.
Yes, I missed that when briefly looking at the code.
>
> This also has the side effect that nvmem cells defined in DT don't
> appear in sysfs, unlike those defined from board code.
Wow, that's not good. I guess we'll want to make that consistent at
some point.
> > > > > Furthermore xlate functions are more about converting
> > > > > from hardware parameters to internal kernel representation than
> > > > > to hide extra DT parsing.
> > > >
> > > > Hm, how is that different? ->of_xlate() is just a way for drivers
> > > > to have their own DT representation, which is exactly what we
> > > > want here.
> > >
> > > There is a big difference. DT represent the hardware and the
> > > relationship between the devices in an OS independent format. We
> > > don't add extra stuff in there just to map back internal Linux API
> > > details.
> >
> > And I'm not talking about adding SW information in the DT, I'm talking
> > about HW specific description. We have the same solution for pinctrl
> > configs (it's HW/driver specific).
>
> For pinctrl I do understand, these beast can be very different from SoC
> to SoC, having a single biding for all doesn't make much sense.
>
> However here we are talking about a simple linear storage, nothing
> special at all. I could see the need for an xlate to for example
> support a device with several partitions, but not to just allow each
> driver to have slightly incompatible bindings.
Maybe, but I guess that's up to the subsystem maintainer to decide what
he prefers.
> >
> > No because partitions defined the old way (as direct subnodes of the
> > MTD node) will be considered as NVMEM cells by the NVMEM framework,
> > and I don't want that.
>
> As I explained above that is not currently the case. If the NVMEM,
> framework is ever changed to explicitly parse NVMEM cells in advance
> we can first update the few existing users to add the compatible string.
We're supposed to be backward compatible (compatible with old DTs), so
that's not an option, though we could add a way to check the compat
string afterwards.
>
> > Plus, I don't want people to start defining their NVMEM cells and
> > forget the compat string (which would work just fine because the
> > NVMEM framework doesn't care).
>
> A review of a new DTS should check that it use each binding correctly,
> AFAIK the DT people do that. We could also add a warning when there is
> no compatible string, that would also help pushing people to update
> their DTS.
Yes, but I'd still prefer if we were preventing people from referencing
mtd-nvmem cells if the node does not have an "nvmem-cell" compat.
>
> > >
> > > > What forces people to add this compatible in their
> > > > DT? Nothing. I'll tell you what will happen: people will start
> > > > defining their nvmem cells directly under the MTD node because
> > > > that *works*, and even if the binding is not documented and we
> > > > consider it invalid, we'll be stuck supporting it forever.
> > >
> > > Do note that undocumented bindings are not allowed. DTS that use
> > > undocumented bindings (normally) just get rejected.
> >
> > Except that's just in theory. In practice, if people can do something
> > wrong, they'll complain if you later fix the bug and break their
> > setup. So no, if we go for the "nvmem cells have an 'nvmem-cell'
> > compat", then I'd like the NVMEM framework to enforce that somehow.
>
> That should be trivial to implement.
Exactly, and that's why I'm insisting on this point.
^ permalink raw reply
* Re: [RFC PATCH net-next v2 10/17] ethtool: implement GET_SETTINGS message
From: Andrew Lunn @ 2018-08-21 14:10 UTC (permalink / raw)
To: Michal Kubecek
Cc: netdev, linux-kernel, Jiri Pirko, David Miller, Florian Fainelli,
Roopa Prabhu, Jakub Kicinski, John W. Linville
In-Reply-To: <20180821093245.ktattosaduf6jvya@unicorn.suse.cz>
On Tue, Aug 21, 2018 at 11:32:46AM +0200, Michal Kubecek wrote:
> On Mon, Jul 30, 2018 at 08:54:55PM +0200, Andrew Lunn wrote:
> > > +/* Internal kernel helper to query a device ethtool_link_settings.
> > > + *
> > > + * Backward compatibility note: for compatibility with legacy drivers
> > > + * that implement only the ethtool_cmd API, this has to work with both
> > > + * drivers implementing get_link_ksettings API and drivers
> > > + * implementing get_settings API. When drivers implement get_settings
> > > + * and report ethtool_cmd deprecated fields
> > > + * (transceiver/maxrxpkt/maxtxpkt), these fields are silently ignored
> > > + * because the resulting struct ethtool_link_settings does not report them.
> >
> > ~/linux/drivers$ grep -r [.]get_settings *
> > net/ethernet/8390/etherh.c: .get_settings = etherh_get_settings,
> >
> > I don't think it is worth adding support for .get_settings for just
> > one driver. It is better to just convert that driver to the new API.
>
> I have prepared a patch converting 8390/etherh driver to use
> {g,s}et_link_ksettings and I'm going to submit it when net-next opens.
> Do you think we can then drop {g,s}et_settings callbacks completely
> (i.e. also from ioctl() code and ethtool_ops)? Do we care about
> unconverted out of tree drivers?
Hi Michal
We cannot break ethtool, the ABI it uses. But there is already code to
use get_link_ksettings() and only fall back to get_settings if it does
not exist. So we can clean up all the fallback code, remove the
ethtool_ops, etc.
I personally don't care about out of tree drivers. They have had over
2 years to change to the new API.
Andrew
^ permalink raw reply
* Re: [PATCH v2 06/29] mtd: Add support for reading MTD devices via the nvmem API
From: Boris Brezillon @ 2018-08-21 10:43 UTC (permalink / raw)
To: Srinivas Kandagatla
Cc: Alban, Bartosz Golaszewski, Jonathan Corbet, Sekhar Nori,
Kevin Hilman, Russell King, Arnd Bergmann, Greg Kroah-Hartman,
David Woodhouse, Brian Norris, Marek Vasut, Richard Weinberger,
Grygorii Strashko, David S . Miller, Naren, Mauro Carvalho Chehab,
Andrew Morton, Lukas Wunner, Dan Carpenter, Florian Fainelli
In-Reply-To: <bd4dd892-3b6c-a77e-b2f3-386600c8c9bd@linaro.org>
On Tue, 21 Aug 2018 11:11:58 +0100
Srinivas Kandagatla <srinivas.kandagatla@linaro.org> wrote:
> On 21/08/18 10:56, Boris Brezillon wrote:
> > On Tue, 21 Aug 2018 10:50:07 +0100
> > Srinivas Kandagatla<srinivas.kandagatla@linaro.org> wrote:
> >
> >> On 20/08/18 19:20, Boris Brezillon wrote:
> >>> On Mon, 20 Aug 2018 11:43:34 +0100
> >>> Srinivas Kandagatla<srinivas.kandagatla@linaro.org> wrote:
> >>>
> >>>> Overall am still not able to clear visualize on how MTD bindings with
> >>>> nvmem cells would look in both partition and un-partition usecases?
> >>>> An example DT would be nice here!!
> >>> Something along those lines:
> >>>
> >> This looks good to me.
> >>> mtdnode {
> >>> nvmem-cells {
> >>> #address-cells = <1>;
> >>> #size-cells = <1>;
> >>>
> >>> cell@0 {
> >>> reg = <0x0 0x14>;
> >>> };
> >>> };
> >>>
> >>> partitions {
> >>> compatible = "fixed-partitions";
> >>> #address-cells = <1>;
> >>> #size-cells = <1>;
> >>>
> >>> partition@0 {
> >>> reg = <0x0 0x20000>;
> >>>
> >>> nvmem-cells {
> >>> #address-cells = <1>;
> >>> #size-cells = <1>;
> >>>
> >>> cell@0 {
> >>> reg = <0x0 0x10>;
> >>> };
> >>> };
> >>> };
> >>> };
> >>> }; >
> >> Just curious...Is there a reason why we can't do it like this?:
> >> Is this because of issue of #address-cells and #size-cells Or mtd
> >> bindings always prefer subnodes?
> >>
> >> mtdnode {
> >> reg = <0x0123000 0x40000>;
> >> #address-cells = <1>;
> >> #size-cells = <1>;
> >> cell@0 {
> >> compatible = "nvmem-cell";
> >> reg = <0x0 0x14>;
> >> };
> >>
> >> partitions {
> >> compatible = "fixed-partitions";
> >> #address-cells = <1>;
> >> #size-cells = <1>;
> >>
> >> partition@0 {
> >> reg = <0x0 0x20000>;
> >> cell@0 {
> >> compatible = "nvmem-cell";
> >> reg = <0x0 0x10>;
> >> };
> >> };
> >> };
> >> };
> > It's because partitions were initially directly defined under the mtd
> > node, so, if you have an old DT you might have something like:
> >
> > mtdnode {
> > reg = <0x0123000 0x40000>;
> > #address-cells = <1>;
> > #size-cells = <1>;
> >
> > partition@0 {
> > reg = <0x0 0x20000>;
> > ...
> > };
> > ...
> > };
> >
> > If we use such a DT with this patch applied, the NVMEM framework will
> > consider MTD partitions as nvmem cells, which is not what we want.
> Yep, I agree.
> TBH, I wanted to add compatible string to nvmem-cell at some point in
> time and it seems more natural update too. One of the reason we
> discussed this in the past was parsers. Looks like mtd can make use of this.
>
> We should be able to add this as an optional flag in nvmem_config to
> enforce this check in case providers wanted to.
>
> Do you think that would help mtd nvmem case?
Yes, it should work if nvmem cells are defined directly under the mtd
node (or the partition they belong to).
> Also I felt like nvmem-cells subnode seems to be a bit heavy!
I still think grouping nvmem cells in a subnode is cleaner (just like
we do for partitions), but I won't object if all parties (you, Alban
and Rob) agree on this solution.
^ permalink raw reply
* Re: [PATCH v2 06/29] mtd: Add support for reading MTD devices via the nvmem API
From: Alban @ 2018-08-21 13:57 UTC (permalink / raw)
To: Boris Brezillon
Cc: Alban Bedel, Srinivas Kandagatla, Bartosz Golaszewski,
Jonathan Corbet, Sekhar Nori, Kevin Hilman, Russell King,
Arnd Bergmann, Greg Kroah-Hartman, David Woodhouse, Brian Norris,
Marek Vasut, Richard Weinberger, Grygorii Strashko,
David S . Miller, Naren, Mauro Carvalho Chehab, Andrew Morton,
Lukas Wunner, Dan
In-Reply-To: <20180821145725.3b385399@bbrezillon>
[-- Attachment #1: Type: text/plain, Size: 14073 bytes --]
On Tue, 21 Aug 2018 14:57:25 +0200
Boris Brezillon <boris.brezillon@bootlin.com> wrote:
> On Tue, 21 Aug 2018 14:27:16 +0200
> Alban <albeu@free.fr> wrote:
>
> > On Tue, 21 Aug 2018 07:44:04 +0200
> > Boris Brezillon <boris.brezillon@bootlin.com> wrote:
> >
> > > On Tue, 21 Aug 2018 00:53:27 +0200
> > > Alban <albeu@free.fr> wrote:
> > >
> > > > On Sun, 19 Aug 2018 18:46:09 +0200
> > > > Boris Brezillon <boris.brezillon@bootlin.com> wrote:
> > > >
> > > > > On Sun, 19 Aug 2018 13:31:06 +0200
> > > > > Alban <albeu@free.fr> wrote:
> > > > >
> > > > > > On Fri, 17 Aug 2018 18:27:20 +0200
> > > > > > Boris Brezillon <boris.brezillon@bootlin.com> wrote:
> > > > > >
> > > > > > > Hi Bartosz,
> > > > > > >
> > > > > > > On Fri, 10 Aug 2018 10:05:03 +0200
> > > > > > > Bartosz Golaszewski <brgl@bgdev.pl> wrote:
> > > > > > >
> > > > > > > > From: Alban Bedel <albeu@free.fr>
> > > > > > > >
> > > > > > > > Allow drivers that use the nvmem API to read data
> > > > > > > > stored on MTD devices. For this the mtd devices are
> > > > > > > > registered as read-only NVMEM providers.
> > > > > > > >
> > > > > > > > Signed-off-by: Alban Bedel <albeu@free.fr>
> > > > > > > > [Bartosz:
> > > > > > > > - use the managed variant of nvmem_register(),
> > > > > > > > - set the nvmem name]
> > > > > > > > Signed-off-by: Bartosz Golaszewski
> > > > > > > > <bgolaszewski@baylibre.com>
> > > > > > >
> > > > > > > What happened to the 2 other patches of Alban's series?
> > > > > > > I'd really like the DT case to be handled/agreed on in
> > > > > > > the same patchset, but IIRC, Alban and Srinivas disagreed
> > > > > > > on how this should be represented. I hope this time we'll
> > > > > > > come to an agreement, because the MTD <-> NVMEM glue has
> > > > > > > been floating around for quite some time...
> > > > > >
> > > > > > These other patches were to fix what I consider a
> > > > > > fundamental flaw in the generic NVMEM bindings, however we
> > > > > > couldn't agree on this point. Bartosz later contacted me to
> > > > > > take over this series and I suggested to just change the
> > > > > > MTD NVMEM binding to use a compatible string on the NVMEM
> > > > > > cells as an alternative solution to fix the clash with the
> > > > > > old style MTD partition.
> > > > > >
> > > > > > However all this has no impact on the code needed to add
> > > > > > NVMEM support to MTD, so the above patch didn't change at
> > > > > > all.
> > > > >
> > > > > It does have an impact on the supported binding though.
> > > > > nvmem->dev.of_node is automatically assigned to
> > > > > mtd->dev.of_node, which means people will be able to define
> > > > > their NVMEM cells directly under the MTD device and reference
> > > > > them from other nodes (even if it's not documented), and as
> > > > > you said, it conflict with the old MTD partition bindings. So
> > > > > we'd better agree on this binding before merging this
> > > > > patch.
> > > >
> > > > Unless the nvmem cell node has a compatible string, then it
> > > > won't be considered as a partition by the MTD code. That is
> > > > were the clash is, both bindings allow free named child nodes
> > > > without a compatible string.
> > >
> > > Except the current nvmem cells parsing code does not enforce
> > > that, and existing DTs rely on this behavior, so we're screwed.
> > > Or are you suggesting to add a new "bool check_cells_compat;"
> > > field to nvmem_config?
> >
> > There is no nvmem cell parsing at the moment. The DT lookup just
> > resolve the phandle to the cell node, take the parent node and
> > search for the nvmem provider that has this OF node. So extending
> > it in case the node has a *new* compatible string would not break
> > users of the old binding, none of them has a compatible string.
>
> But we want to enforce the compat check on MTD devices, otherwise old
> MTD partitions (those defined under the MTD node) will be considered
> as NVMEM cells by the NVMEM framework. Hence the bool
> check_cells_compat field.
That would only be needed if the NVMEM framework would do "forward"
parsing, creating data structure for each NVMEM cell found under an
NVMEM provider. However currently it doesn't do that and only goes
"backward", starting by resolving a phandle pointing to a cell, then
finding the provider that the cell belongs to.
This also has the side effect that nvmem cells defined in DT don't
appear in sysfs, unlike those defined from board code.
> >
> > > >
> > > > > I see several options:
> > > > >
> > > > > 1/ provide a way to tell the NVMEM framework not to use
> > > > > parent->of_node even if it's != NULL. This way we really don't
> > > > > support defining NVMEM cells in the DT, and also don't support
> > > > > referencing the nvmem device using a phandle.
> > > >
> > > > I really don't get what the point of this would be. Make the
> > > > whole API useless?
> > >
> > > No, just allow Bartosz to get his changes merged without waiting
> > > for you and Srinivas to agree on how to handle the new binding.
> > > As I said earlier, this mtd <-> nvmem stuff has been around for
> > > quite some time, and instead of trying to find an approach that
> > > makes everyone happy, you decided to let the patchset die.
> >
> > As long as that wouldn't prevent using DT in the future I'm fine
> > with it.
> >
> > > >
> > > > > 2/ define a new binding where all nvmem-cells are placed in an
> > > > > "nvmem" subnode (just like we have this "partitions"
> > > > > subnode for partitions), and then add a config->of_node field
> > > > > so that the nvmem provider can explicitly specify the DT node
> > > > > representing the nvmem device. We'll also need to set this
> > > > > field to ERR_PTR(-ENOENT) in case this node does not exist so
> > > > > that the nvmem framework knows that it should not assign
> > > > > nvmem->dev.of_node to parent->of_node
> > > >
> > > > This is not good. First the NVMEM device is only a virtual
> > > > concept of the Linux kernel, it has no place in the DT.
> > >
> > > nvmem-cells is a virtual concept too, still, you define them in
> > > the DT.
> >
> > To be honest I also think that naming this concept "nvmem" in the
> > DT was a bad idea. Perhaps something like "driver-data" or
> > "data-cell" would have been better as that would make it clear what
> > this is about, nvmem is just the Linux implementation of this
> > concept.
>
> I'm fine using a different name.
>
> >
> > > > Secondly the NVMEM
> > > > provider (here the MTD device) then has to manually parse its DT
> > > > node to find this subnode, pass it to the NVMEM framework to
> > > > later again resolve it back to the MTD device.
> > >
> > > We don't resolve it back to the MTD device, because the MTD
> > > device is just the parent of the nvmem device.
> > >
> > > > Not very complex but still a lot of
> > > > useless code, just registering the MTD device is a lot simpler
> > > > and much more inline with most other kernel API that register a
> > > > "service" available from a device.
> > >
> > > I'm not a big fan of this option either, but I thought I had to
> > > propose it.
> > >
> > > >
> > > > > 3/ only declare partitions as nvmem providers. This would
> > > > > solve the problem we have with partitions defined in the DT
> > > > > since defining sub-partitions in the DT is not (yet?)
> > > > > supported and partition nodes are supposed to be leaf nodes.
> > > > > Still, I'm not a big fan of this solution because it will
> > > > > prevent us from supporting sub-partitions if we ever
> > > > > want/need to.
> > > >
> > > > That sound like a poor workaround.
> > >
> > > Yes, that's a workaround. And the reason I propose it, is, again,
> > > because I don't want to block Bartosz.
> > >
> > > > Remember that this problem could
> > > > appear with any device that has a binding that use child
> > > > nodes.
> > >
> > > I'm talking about partitions, and you're talking about mtd
> > > devices. Right now partitions don't have subnodes, and if we
> > > define that partition subnodes should describe nvmem-cells, then
> > > it becomes part of the official binding. So, no, the problem you
> > > mention does not (yet) exist.
> >
> > That would add another binding that allow free named child nodes
> > without compatible string although experience has repeatedly shown
> > that this was a bad idea.
>
> Yes, I agree. Just thought it was important to have this solution in
> the list, even if it's just to reject it.
>
> >
> > > >
> > > > > 4/ Add a ->of_xlate() hook that would be called if present by
> > > > > the framework instead of using the default parsing we have
> > > > > right now.
> > > >
> > > > That is a bit cleaner, but I don't think it would be worse the
> > > > complexity.
> > >
> > > But it's way more flexible than putting everything in the nvmem
> > > framework. BTW, did you notice that nvmem-cells parsing does not
> > > work with flashes bigger than 4GB, because the framework assumes
> > > #address-cells and #size-cells are always 1. That's probably
> > > something we'll have to fix for the MTD case.
> >
> > Yes, however that's just an implementation limitation which is
> > trivial to solve.
>
> Agree. I was just pointing it in case you hadn't noticed.
>
> >
> > > > Furthermore xlate functions are more about converting
> > > > from hardware parameters to internal kernel representation than
> > > > to hide extra DT parsing.
> > >
> > > Hm, how is that different? ->of_xlate() is just a way for drivers
> > > to have their own DT representation, which is exactly what we
> > > want here.
> >
> > There is a big difference. DT represent the hardware and the
> > relationship between the devices in an OS independent format. We
> > don't add extra stuff in there just to map back internal Linux API
> > details.
>
> And I'm not talking about adding SW information in the DT, I'm talking
> about HW specific description. We have the same solution for pinctrl
> configs (it's HW/driver specific).
For pinctrl I do understand, these beast can be very different from SoC
to SoC, having a single biding for all doesn't make much sense.
However here we are talking about a simple linear storage, nothing
special at all. I could see the need for an xlate to for example
support a device with several partitions, but not to just allow each
driver to have slightly incompatible bindings.
> >
> > > >
> > > > > 5/ Tell the nvmem framework the name of the subnode containing
> > > > > nvmem cell definitions (if NULL that means cells are directly
> > > > > defined under the nvmem provider node). We would set it to
> > > > > "nvmem-cells" (or whatever you like) for the MTD case.
> > > >
> > > > If so please match on compatible and not on the node name.
> > >
> > > If you like.
> > >
> > > >
> > > > 6/ Extend the current NVMEM cell lookup to check if the parent
> > > > node of the cell has a compatible string set to "nvmem-cells".
> > > > If it doesn't it mean we have the current binding and this node
> > > > is the NVMEM device. If it does the device node is just the
> > > > next parent. This is trivial to implement (literally 2 lines of
> > > > code) and cover all the cases currently known.
> > >
> > > Except Srinivas was not happy with this solution, and this
> > > stalled the discussion. I'm trying to find other options and you
> > > keep rejecting all of them to come back to this one.
> >
> > Well, I think this is the best solution :/
> >
> > > >
> > > > 7/ Just add a compatible string to the nvmem cell. No code
> > > > change is needed,
> > >
> > > That's not true!!!
> >
> > What is not true in this statement? The current nvmem lookup don't
> > care about compatible strings, so the cell lookup would just works
> > fine. The MTD partition parser won't consider them as a partition
> > because of the compatible string. Problem solved!
>
> No because partitions defined the old way (as direct subnodes of the
> MTD node) will be considered as NVMEM cells by the NVMEM framework,
> and I don't want that.
As I explained above that is not currently the case. If the NVMEM,
framework is ever changed to explicitly parse NVMEM cells in advance
we can first update the few existing users to add the compatible string.
> Plus, I don't want people to start defining their NVMEM cells and
> forget the compat string (which would work just fine because the
> NVMEM framework doesn't care).
A review of a new DTS should check that it use each binding correctly,
AFAIK the DT people do that. We could also add a warning when there is
no compatible string, that would also help pushing people to update
their DTS.
> >
> > > What forces people to add this compatible in their
> > > DT? Nothing. I'll tell you what will happen: people will start
> > > defining their nvmem cells directly under the MTD node because
> > > that *works*, and even if the binding is not documented and we
> > > consider it invalid, we'll be stuck supporting it forever.
> >
> > Do note that undocumented bindings are not allowed. DTS that use
> > undocumented bindings (normally) just get rejected.
>
> Except that's just in theory. In practice, if people can do something
> wrong, they'll complain if you later fix the bug and break their
> setup. So no, if we go for the "nvmem cells have an 'nvmem-cell'
> compat", then I'd like the NVMEM framework to enforce that somehow.
That should be trivial to implement.
Alban
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]
^ permalink raw reply
* Re: [PATCH v2 06/29] mtd: Add support for reading MTD devices via the nvmem API
From: Srinivas Kandagatla @ 2018-08-21 10:11 UTC (permalink / raw)
To: Boris Brezillon
Cc: Alban, Bartosz Golaszewski, Jonathan Corbet, Sekhar Nori,
Kevin Hilman, Russell King, Arnd Bergmann, Greg Kroah-Hartman,
David Woodhouse, Brian Norris, Marek Vasut, Richard Weinberger,
Grygorii Strashko, David S . Miller, Naren, Mauro Carvalho Chehab,
Andrew Morton, Lukas Wunner, Dan Carpenter, Florian Fainelli
In-Reply-To: <20180821115639.4894c1c9@bbrezillon>
On 21/08/18 10:56, Boris Brezillon wrote:
> On Tue, 21 Aug 2018 10:50:07 +0100
> Srinivas Kandagatla<srinivas.kandagatla@linaro.org> wrote:
>
>> On 20/08/18 19:20, Boris Brezillon wrote:
>>> On Mon, 20 Aug 2018 11:43:34 +0100
>>> Srinivas Kandagatla<srinivas.kandagatla@linaro.org> wrote:
>>>
>>>> Overall am still not able to clear visualize on how MTD bindings with
>>>> nvmem cells would look in both partition and un-partition usecases?
>>>> An example DT would be nice here!!
>>> Something along those lines:
>>>
>> This looks good to me.
>>> mtdnode {
>>> nvmem-cells {
>>> #address-cells = <1>;
>>> #size-cells = <1>;
>>>
>>> cell@0 {
>>> reg = <0x0 0x14>;
>>> };
>>> };
>>>
>>> partitions {
>>> compatible = "fixed-partitions";
>>> #address-cells = <1>;
>>> #size-cells = <1>;
>>>
>>> partition@0 {
>>> reg = <0x0 0x20000>;
>>>
>>> nvmem-cells {
>>> #address-cells = <1>;
>>> #size-cells = <1>;
>>>
>>> cell@0 {
>>> reg = <0x0 0x10>;
>>> };
>>> };
>>> };
>>> };
>>> }; >
>> Just curious...Is there a reason why we can't do it like this?:
>> Is this because of issue of #address-cells and #size-cells Or mtd
>> bindings always prefer subnodes?
>>
>> mtdnode {
>> reg = <0x0123000 0x40000>;
>> #address-cells = <1>;
>> #size-cells = <1>;
>> cell@0 {
>> compatible = "nvmem-cell";
>> reg = <0x0 0x14>;
>> };
>>
>> partitions {
>> compatible = "fixed-partitions";
>> #address-cells = <1>;
>> #size-cells = <1>;
>>
>> partition@0 {
>> reg = <0x0 0x20000>;
>> cell@0 {
>> compatible = "nvmem-cell";
>> reg = <0x0 0x10>;
>> };
>> };
>> };
>> };
> It's because partitions were initially directly defined under the mtd
> node, so, if you have an old DT you might have something like:
>
> mtdnode {
> reg = <0x0123000 0x40000>;
> #address-cells = <1>;
> #size-cells = <1>;
>
> partition@0 {
> reg = <0x0 0x20000>;
> ...
> };
> ...
> };
>
> If we use such a DT with this patch applied, the NVMEM framework will
> consider MTD partitions as nvmem cells, which is not what we want.
Yep, I agree.
TBH, I wanted to add compatible string to nvmem-cell at some point in
time and it seems more natural update too. One of the reason we
discussed this in the past was parsers. Looks like mtd can make use of this.
We should be able to add this as an optional flag in nvmem_config to
enforce this check in case providers wanted to.
Do you think that would help mtd nvmem case?
Also I felt like nvmem-cells subnode seems to be a bit heavy!
thanks,
srini
^ permalink raw reply
* Re: [PATCH v1 6/7] net: mvneta: Don't use GRO on Armada 3720
From: Marek Behún @ 2018-08-21 10:07 UTC (permalink / raw)
To: Jisheng Zhang
Cc: Thomas Petazzoni, Andrew Lunn, linux-arm-kernel, netdev,
Gregory CLEMENT, Tomas Hlavacek, Russell King - ARM Linux,
David S. Miller
In-Reply-To: <20180809200832.7de1fa9e@xhacker.debian>
On Thu, 9 Aug 2018 20:08:32 +0800
Jisheng Zhang <Jisheng.Zhang@synaptics.com> wrote:
> On Thu, 9 Aug 2018 19:27:55 +0800 Jisheng Zhang wrote:
>
> > Hi,
> >
> > On Thu, 9 Aug 2018 12:40:41 +0800 Jisheng Zhang wrote:
> >
> > > + more people
> > >
> > > On Wed, 8 Aug 2018 17:27:05 +0200 Marek Behún wrote:
> > >
> > > > For some reason on Armada 3720 boards (EspressoBin and Turris
> > > > Mox) the networking driver behaves weirdly when using
> > > > napi_gro_receive.
> > > >
> > > > For example downloading a big file from a local network (low
> > > > ping) is fast, but when downloading from a remote server
> > > > (higher ping), the download speed is at first high but drops
> > > > rapidly to almost nothing or absolutely nothing.
> > >
> > > We also met this issue on some berlin platforms. I tried to fix
> > > the bug, but no clue so far.
> > >
> > > >
> > > > This is fixed when using netif_receive_skb instead of
> > > > napi_gro_receive.
> > >
> > > This is a workaround. The good news is this workaround also fixes
> > > the issue we saw on berlin.
> > >
> > > >
> > > > Signed-off-by: Marek Behun <marek.behun@nic.cz>
> > > > Cc: Russell King - ARM Linux <linux@armlinux.org.uk>
> > > > Cc: netdev@vger.kernel.org
> > > >
> > > > diff --git a/drivers/net/ethernet/marvell/mvneta.c
> > > > b/drivers/net/ethernet/marvell/mvneta.c index
> > > > 0ad2f3f7da85..27f3017d94c5 100644 ---
> > > > a/drivers/net/ethernet/marvell/mvneta.c +++
> > > > b/drivers/net/ethernet/marvell/mvneta.c @@ -1959,7 +1959,10 @@
> > > > static int mvneta_rx_swbm(struct mvneta_port *pp, int rx_todo,
> > > > skb->protocol = eth_type_trans(skb,
> > > > dev); mvneta_rx_csum(pp, rx_status, skb);
> > > > - napi_gro_receive(&port->napi, skb);
> > > > + if (pp->neta_armada3700)
> > > > + netif_receive_skb(skb);
> > > > + else
> > > > + napi_gro_receive(&port->napi,
> > > > skb);
> >
> > I think I found the root cause, if neta_armada3700 is true, the
> > port got from this_cpu_ptr(pp->ports) is invalid, this is bug...
> > I'll cook a patch for this
>
> correct it as:
>
> the port's(port is got from this_cpu_ptr(pp->ports) napi is invalid.
>
> Patch is sent out. Could you please try?
>
> Per my test, it solves the issue we saw on berlin.
>
> >
> > Thanks
> >
> > > >
> > > > rcvd_pkts++;
> > > > rcvd_bytes += rx_bytes;
> > > > @@ -2001,7 +2004,10 @@ static int mvneta_rx_swbm(struct
> > > > mvneta_port *pp, int rx_todo,
> > > > mvneta_rx_csum(pp, rx_status, skb);
> > > >
> > > > - napi_gro_receive(&port->napi, skb);
> > > > + if (pp->neta_armada3700)
> > > > + netif_receive_skb(skb);
> > > > + else
> > > > + napi_gro_receive(&port->napi, skb);
> > > > }
> > > >
> > > > if (rcvd_pkts) {
> > >
> >
>
Jisheng, the issue is solved with your patch. Thanks :)
^ permalink raw reply
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox