[U-Boot] [PATCH 0/5] sandbox: net: Fix sandbox eth drivers
It seems as sandbox moved to livetree these drivers were not updated. Heinrich suggested correctly that at this point we should be enumerating network interface names, not hard-coding eth0. Also, there were a few bugs that needed fixing.
Joe Hershberger (5): sandbox: eth-raw: Correct valid socket test in send/recv net: Correct comment in Kconfig net: sandbox: Convert sandbox mock eth driver to livetree net: sandbox-raw: Convert raw eth driver to livetree net: sandbox-raw: Allow interface to be specified by index
arch/sandbox/cpu/eth-raw-os.c | 32 ++++++++++++++++++++++++++++++-- arch/sandbox/dts/sandbox.dts | 2 +- arch/sandbox/include/asm/eth-raw-os.h | 4 +++- drivers/net/Kconfig | 2 +- drivers/net/sandbox-raw.c | 28 ++++++++++++++++++---------- drivers/net/sandbox.c | 15 +++++++++++---- 6 files changed, 64 insertions(+), 19 deletions(-)
In open, the socket is correctly checked to be -1 in the error case. In send and recv, we checked for 0, but that is a valid socket number.
Correct this by checking for -1 as a bad socket everywhere.
Signed-off-by: Joe Hershberger joe.hershberger@ni.com ---
arch/sandbox/cpu/eth-raw-os.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/arch/sandbox/cpu/eth-raw-os.c b/arch/sandbox/cpu/eth-raw-os.c index e054a0702a..61f23ed210 100644 --- a/arch/sandbox/cpu/eth-raw-os.c +++ b/arch/sandbox/cpu/eth-raw-os.c @@ -156,7 +156,7 @@ int sandbox_eth_raw_os_send(void *packet, int length, int retval; struct udphdr *udph = packet + sizeof(struct iphdr);
- if (!priv->sd || !priv->device) + if (priv->sd < 0 || !priv->device) return -EINVAL;
/* @@ -221,7 +221,7 @@ int sandbox_eth_raw_os_recv(void *packet, int *length, int retval; int saddr_size;
- if (!priv->sd || !priv->device) + if (priv->sd < 0 || !priv->device) return -EINVAL; saddr_size = sizeof(struct sockaddr); retval = recvfrom(priv->sd, packet, 1536, 0,
On 26 June 2018 at 14:19, Joe Hershberger joe.hershberger@ni.com wrote:
In open, the socket is correctly checked to be -1 in the error case. In send and recv, we checked for 0, but that is a valid socket number.
Correct this by checking for -1 as a bad socket everywhere.
Signed-off-by: Joe Hershberger joe.hershberger@ni.com
arch/sandbox/cpu/eth-raw-os.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
Reviewed-by: Simon Glass sjg@chromium.org
Signed-off-by: Joe Hershberger joe.hershberger@ni.com ---
drivers/net/Kconfig | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/net/Kconfig b/drivers/net/Kconfig index e88f056d84..5f26a0004a 100644 --- a/drivers/net/Kconfig +++ b/drivers/net/Kconfig @@ -8,7 +8,7 @@ config DM_ETH Enable driver model for Ethernet.
The eth_*() interface will be implemented by the UC_ETH class - This is currently implemented in net/eth.c + This is currently implemented in net/eth-uclass.c Look in include/net.h for details.
config DRIVER_TI_CPSW
On 26 June 2018 at 14:19, Joe Hershberger joe.hershberger@ni.com wrote:
Signed-off-by: Joe Hershberger joe.hershberger@ni.com
drivers/net/Kconfig | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Reviewed-by: Simon Glass sjg@chromium.org
Use the dev_ functions to access DT properties.
Also correct the reading of the fake MAC address. The format from the DT is u32s, so to accurately read the MAC from the DT, we need to cast each value to a u8.
Signed-off-by: Joe Hershberger joe.hershberger@ni.com ---
drivers/net/sandbox.c | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-)
diff --git a/drivers/net/sandbox.c b/drivers/net/sandbox.c index b34712bd06..d5a30c05b8 100644 --- a/drivers/net/sandbox.c +++ b/drivers/net/sandbox.c @@ -56,13 +56,20 @@ void sandbox_eth_skip_timeout(void) static int sb_eth_start(struct udevice *dev) { struct eth_sandbox_priv *priv = dev_get_priv(dev); + /* The DT integers are 32-bits */ + const u32 mac[ARP_HLEN];
debug("eth_sandbox: Start\n");
- fdtdec_get_byte_array(gd->fdt_blob, dev_of_offset(dev), - "fake-host-hwaddr", priv->fake_host_hwaddr, - ARP_HLEN); + if (dev_read_u32_array(dev, "fake-host-hwaddr", mac, ARP_HLEN)) { + printf("'fake-host-hwaddr' is missing from the DT\n"); + return -EINVAL; + } + for (int i = 0; i < ARP_HLEN; i++) + priv->fake_host_hwaddr[i] = (uint8_t)mac[i]; + priv->recv_packet_buffer = net_rx_packets[0]; + return 0; }
@@ -204,7 +211,7 @@ static int sb_eth_ofdata_to_platdata(struct udevice *dev) { struct eth_pdata *pdata = dev_get_platdata(dev);
- pdata->iobase = devfdt_get_addr(dev); + pdata->iobase = dev_read_addr(dev); return 0; }
Hi Joe,
On 26 June 2018 at 14:19, Joe Hershberger joe.hershberger@ni.com wrote:
Use the dev_ functions to access DT properties.
Also correct the reading of the fake MAC address. The format from the DT is u32s, so to accurately read the MAC from the DT, we need to cast each value to a u8.
Signed-off-by: Joe Hershberger joe.hershberger@ni.com
drivers/net/sandbox.c | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-)
diff --git a/drivers/net/sandbox.c b/drivers/net/sandbox.c index b34712bd06..d5a30c05b8 100644 --- a/drivers/net/sandbox.c +++ b/drivers/net/sandbox.c @@ -56,13 +56,20 @@ void sandbox_eth_skip_timeout(void) static int sb_eth_start(struct udevice *dev) { struct eth_sandbox_priv *priv = dev_get_priv(dev);
/* The DT integers are 32-bits */const u32 mac[ARP_HLEN]; debug("eth_sandbox: Start\n");
fdtdec_get_byte_array(gd->fdt_blob, dev_of_offset(dev),"fake-host-hwaddr", priv->fake_host_hwaddr,ARP_HLEN);
if (dev_read_u32_array(dev, "fake-host-hwaddr", mac, ARP_HLEN)) {printf("'fake-host-hwaddr' is missing from the DT\n");return -EINVAL;}
This is not equivalent - I think you need a dev_read_u8_array() or similar.
for (int i = 0; i < ARP_HLEN; i++)priv->fake_host_hwaddr[i] = (uint8_t)mac[i];priv->recv_packet_buffer = net_rx_packets[0];return 0;}
@@ -204,7 +211,7 @@ static int sb_eth_ofdata_to_platdata(struct udevice *dev) { struct eth_pdata *pdata = dev_get_platdata(dev);
pdata->iobase = devfdt_get_addr(dev);
pdata->iobase = dev_read_addr(dev); return 0;}
-- 2.11.0
Regards, Simon
Hi Simon,
On Tue, Jun 26, 2018 at 6:18 PM, Simon Glass sjg@chromium.org wrote:
Hi Joe,
On 26 June 2018 at 14:19, Joe Hershberger joe.hershberger@ni.com wrote:
Use the dev_ functions to access DT properties.
Also correct the reading of the fake MAC address. The format from the DT is u32s, so to accurately read the MAC from the DT, we need to cast each value to a u8.
Signed-off-by: Joe Hershberger joe.hershberger@ni.com
drivers/net/sandbox.c | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-)
diff --git a/drivers/net/sandbox.c b/drivers/net/sandbox.c index b34712bd06..d5a30c05b8 100644 --- a/drivers/net/sandbox.c +++ b/drivers/net/sandbox.c @@ -56,13 +56,20 @@ void sandbox_eth_skip_timeout(void) static int sb_eth_start(struct udevice *dev) { struct eth_sandbox_priv *priv = dev_get_priv(dev);
/* The DT integers are 32-bits */const u32 mac[ARP_HLEN]; debug("eth_sandbox: Start\n");
fdtdec_get_byte_array(gd->fdt_blob, dev_of_offset(dev),"fake-host-hwaddr", priv->fake_host_hwaddr,ARP_HLEN);
if (dev_read_u32_array(dev, "fake-host-hwaddr", mac, ARP_HLEN)) {printf("'fake-host-hwaddr' is missing from the DT\n");return -EINVAL;}This is not equivalent - I think you need a dev_read_u8_array() or similar.
Yes, I know it's not equivalent... I noted that in the commit log. This is fixing a bug too.
for (int i = 0; i < ARP_HLEN; i++)priv->fake_host_hwaddr[i] = (uint8_t)mac[i];priv->recv_packet_buffer = net_rx_packets[0];return 0;}
@@ -204,7 +211,7 @@ static int sb_eth_ofdata_to_platdata(struct udevice *dev) { struct eth_pdata *pdata = dev_get_platdata(dev);
pdata->iobase = devfdt_get_addr(dev);
pdata->iobase = dev_read_addr(dev); return 0;}
-- 2.11.0
Regards, Simon _______________________________________________ U-Boot mailing list U-Boot@lists.denx.de https://lists.denx.de/listinfo/u-boot
Hi Joe,
On 26 June 2018 at 16:25, Joe Hershberger joe.hershberger@ni.com wrote:
Hi Simon,
On Tue, Jun 26, 2018 at 6:18 PM, Simon Glass sjg@chromium.org wrote:
Hi Joe,
On 26 June 2018 at 14:19, Joe Hershberger joe.hershberger@ni.com wrote:
Use the dev_ functions to access DT properties.
Also correct the reading of the fake MAC address. The format from the DT is u32s, so to accurately read the MAC from the DT, we need to cast each value to a u8.
Signed-off-by: Joe Hershberger joe.hershberger@ni.com
drivers/net/sandbox.c | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-)
diff --git a/drivers/net/sandbox.c b/drivers/net/sandbox.c index b34712bd06..d5a30c05b8 100644 --- a/drivers/net/sandbox.c +++ b/drivers/net/sandbox.c @@ -56,13 +56,20 @@ void sandbox_eth_skip_timeout(void) static int sb_eth_start(struct udevice *dev) { struct eth_sandbox_priv *priv = dev_get_priv(dev);
/* The DT integers are 32-bits */const u32 mac[ARP_HLEN]; debug("eth_sandbox: Start\n");
fdtdec_get_byte_array(gd->fdt_blob, dev_of_offset(dev),"fake-host-hwaddr", priv->fake_host_hwaddr,ARP_HLEN);
if (dev_read_u32_array(dev, "fake-host-hwaddr", mac, ARP_HLEN)) {printf("'fake-host-hwaddr' is missing from the DT\n");return -EINVAL;}This is not equivalent - I think you need a dev_read_u8_array() or similar.
Yes, I know it's not equivalent... I noted that in the commit log. This is fixing a bug too.
Yes I see that, but I didn't realise there was a bug in the code.
Reviewed-by: Simon Glass sjg@chromium.org
Regards, Simon
On Tue, Jun 26, 2018 at 6:28 PM, Simon Glass sjg@chromium.org wrote:
Hi Joe,
On 26 June 2018 at 16:25, Joe Hershberger joe.hershberger@ni.com wrote:
Hi Simon,
On Tue, Jun 26, 2018 at 6:18 PM, Simon Glass sjg@chromium.org wrote:
Hi Joe,
On 26 June 2018 at 14:19, Joe Hershberger joe.hershberger@ni.com wrote:
Use the dev_ functions to access DT properties.
Also correct the reading of the fake MAC address. The format from the DT is u32s, so to accurately read the MAC from the DT, we need to cast each value to a u8.
Signed-off-by: Joe Hershberger joe.hershberger@ni.com
drivers/net/sandbox.c | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-)
diff --git a/drivers/net/sandbox.c b/drivers/net/sandbox.c index b34712bd06..d5a30c05b8 100644 --- a/drivers/net/sandbox.c +++ b/drivers/net/sandbox.c @@ -56,13 +56,20 @@ void sandbox_eth_skip_timeout(void) static int sb_eth_start(struct udevice *dev) { struct eth_sandbox_priv *priv = dev_get_priv(dev);
/* The DT integers are 32-bits */const u32 mac[ARP_HLEN]; debug("eth_sandbox: Start\n");
fdtdec_get_byte_array(gd->fdt_blob, dev_of_offset(dev),"fake-host-hwaddr", priv->fake_host_hwaddr,ARP_HLEN);
if (dev_read_u32_array(dev, "fake-host-hwaddr", mac, ARP_HLEN)) {printf("'fake-host-hwaddr' is missing from the DT\n");return -EINVAL;}This is not equivalent - I think you need a dev_read_u8_array() or similar.
Yes, I know it's not equivalent... I noted that in the commit log. This is fixing a bug too.
Yes I see that, but I didn't realise there was a bug in the code.
OK, turns out this is the result of an inconsistency in the dts files, and should actually be able to be u8s.
I'll correct it for v2.
Reviewed-by: Simon Glass sjg@chromium.org
Regards, Simon _______________________________________________ U-Boot mailing list U-Boot@lists.denx.de https://lists.denx.de/listinfo/u-boot
Signed-off-by: Joe Hershberger joe.hershberger@ni.com ---
drivers/net/sandbox-raw.c | 24 +++++++++++++----------- 1 file changed, 13 insertions(+), 11 deletions(-)
diff --git a/drivers/net/sandbox-raw.c b/drivers/net/sandbox-raw.c index 3f8020f629..f835a6a7f3 100644 --- a/drivers/net/sandbox-raw.c +++ b/drivers/net/sandbox-raw.c @@ -25,17 +25,19 @@ static int sb_eth_raw_start(struct udevice *dev)
debug("eth_sandbox_raw: Start\n");
- interface = fdt_getprop(gd->fdt_blob, dev_of_offset(dev), - "host-raw-interface", NULL); - if (interface == NULL) - return -EINVAL; - - if (strcmp(interface, "lo") == 0) { - priv->local = 1; - env_set("ipaddr", "127.0.0.1"); - env_set("serverip", "127.0.0.1"); + interface = dev_read_prop(dev, "host-raw-interface", NULL); + if (interface) { + printf("eth_sandbox_raw: Using %s from DT\n", interface); + if (strcmp(interface, "lo") == 0) { + priv->local = 1; + env_set("ipaddr", "127.0.0.1"); + env_set("serverip", "127.0.0.1"); + } + return sandbox_eth_raw_os_start(interface, pdata->enetaddr, + priv); } - return sandbox_eth_raw_os_start(interface, pdata->enetaddr, priv); + + return -EINVAL; }
static int sb_eth_raw_send(struct udevice *dev, void *packet, int length) @@ -144,7 +146,7 @@ static int sb_eth_raw_ofdata_to_platdata(struct udevice *dev) { struct eth_pdata *pdata = dev_get_platdata(dev);
- pdata->iobase = devfdt_get_addr(dev); + pdata->iobase = dev_read_addr(dev); return 0; }
Hi Joe,
On 26 June 2018 at 14:19, Joe Hershberger joe.hershberger@ni.com wrote:
Signed-off-by: Joe Hershberger joe.hershberger@ni.com
drivers/net/sandbox-raw.c | 24 +++++++++++++----------- 1 file changed, 13 insertions(+), 11 deletions(-)
Reviewed-by: Simon Glass sjg@chromium.org
diff --git a/drivers/net/sandbox-raw.c b/drivers/net/sandbox-raw.c index 3f8020f629..f835a6a7f3 100644 --- a/drivers/net/sandbox-raw.c +++ b/drivers/net/sandbox-raw.c @@ -25,17 +25,19 @@ static int sb_eth_raw_start(struct udevice *dev)
debug("eth_sandbox_raw: Start\n");
interface = fdt_getprop(gd->fdt_blob, dev_of_offset(dev),"host-raw-interface", NULL);if (interface == NULL)return -EINVAL;if (strcmp(interface, "lo") == 0) {priv->local = 1;env_set("ipaddr", "127.0.0.1");env_set("serverip", "127.0.0.1");
interface = dev_read_prop(dev, "host-raw-interface", NULL);
dev_read_string() ?
if (interface) {
Can you do:
if (!interface) return -EINVAL;
printf("eth_sandbox_raw: Using %s from DT\n", interface);if (strcmp(interface, "lo") == 0) {priv->local = 1;env_set("ipaddr", "127.0.0.1");env_set("serverip", "127.0.0.1");}return sandbox_eth_raw_os_start(interface, pdata->enetaddr,priv); }
return sandbox_eth_raw_os_start(interface, pdata->enetaddr, priv);
return -EINVAL;}
static int sb_eth_raw_send(struct udevice *dev, void *packet, int length) @@ -144,7 +146,7 @@ static int sb_eth_raw_ofdata_to_platdata(struct udevice *dev) { struct eth_pdata *pdata = dev_get_platdata(dev);
pdata->iobase = devfdt_get_addr(dev);
pdata->iobase = dev_read_addr(dev); return 0;}
-- 2.11.0
Regards, Simon
On Tue, Jun 26, 2018 at 6:18 PM, Simon Glass sjg@chromium.org wrote:
Hi Joe,
On 26 June 2018 at 14:19, Joe Hershberger joe.hershberger@ni.com wrote:
Signed-off-by: Joe Hershberger joe.hershberger@ni.com
drivers/net/sandbox-raw.c | 24 +++++++++++++----------- 1 file changed, 13 insertions(+), 11 deletions(-)
Reviewed-by: Simon Glass sjg@chromium.org
diff --git a/drivers/net/sandbox-raw.c b/drivers/net/sandbox-raw.c index 3f8020f629..f835a6a7f3 100644 --- a/drivers/net/sandbox-raw.c +++ b/drivers/net/sandbox-raw.c @@ -25,17 +25,19 @@ static int sb_eth_raw_start(struct udevice *dev)
debug("eth_sandbox_raw: Start\n");
interface = fdt_getprop(gd->fdt_blob, dev_of_offset(dev),"host-raw-interface", NULL);if (interface == NULL)return -EINVAL;if (strcmp(interface, "lo") == 0) {priv->local = 1;env_set("ipaddr", "127.0.0.1");env_set("serverip", "127.0.0.1");
interface = dev_read_prop(dev, "host-raw-interface", NULL);dev_read_string() ?
OK.
if (interface) {Can you do:
if (!interface) return -EINVAL;
I realize it looks a bit awkward in this patch, but it's laid out this way to accommodate patch 5/5.
printf("eth_sandbox_raw: Using %s from DT\n", interface);if (strcmp(interface, "lo") == 0) {priv->local = 1;env_set("ipaddr", "127.0.0.1");env_set("serverip", "127.0.0.1");}return sandbox_eth_raw_os_start(interface, pdata->enetaddr,priv); }
return sandbox_eth_raw_os_start(interface, pdata->enetaddr, priv);
return -EINVAL;}
static int sb_eth_raw_send(struct udevice *dev, void *packet, int length) @@ -144,7 +146,7 @@ static int sb_eth_raw_ofdata_to_platdata(struct udevice *dev) { struct eth_pdata *pdata = dev_get_platdata(dev);
pdata->iobase = devfdt_get_addr(dev);
pdata->iobase = dev_read_addr(dev); return 0;}
-- 2.11.0
Regards, Simon _______________________________________________ U-Boot mailing list U-Boot@lists.denx.de https://lists.denx.de/listinfo/u-boot
With systemd stable interface names, eth0 will almost never exist. Instead of using that name in the sandbox.dts, use index 2 (the first interface after "lo"). Enumerate the interfaces on the system to choose a valid interace name.
Signed-off-by: Joe Hershberger joe.hershberger@ni.com ---
arch/sandbox/cpu/eth-raw-os.c | 28 ++++++++++++++++++++++++++++ arch/sandbox/dts/sandbox.dts | 2 +- arch/sandbox/include/asm/eth-raw-os.h | 4 +++- drivers/net/sandbox-raw.c | 8 +++++++- 4 files changed, 39 insertions(+), 3 deletions(-)
diff --git a/arch/sandbox/cpu/eth-raw-os.c b/arch/sandbox/cpu/eth-raw-os.c index 61f23ed210..11b5271f31 100644 --- a/arch/sandbox/cpu/eth-raw-os.c +++ b/arch/sandbox/cpu/eth-raw-os.c @@ -150,6 +150,34 @@ int sandbox_eth_raw_os_start(const char *ifname, unsigned char *ethmac, return _raw_packet_start(ifname, ethmac, priv); }
+int sandbox_eth_raw_os_start_idx(unsigned int ifindex, unsigned char *ethmac, + struct eth_sandbox_raw_priv *priv) +{ + int ret; + struct if_nameindex *ni, *i; + const char *ifname = NULL; + + ni = if_nameindex(); + if (!ni) + return -EINVAL; + + for (i = ni; !(i->if_index == 0 && !i->if_name); i++) { + if (ifindex == i->if_index) { + ifname = i->if_name; + printf("(%s)\n", ifname); + break; + } + } + /* Don't bother supporting 'lo' by index. Just use the name in DTS */ + if (ifname) + ret = _raw_packet_start(ifname, ethmac, priv); + else + ret = -EINVAL; + + if_freenameindex(ni); + return ret; +} + int sandbox_eth_raw_os_send(void *packet, int length, struct eth_sandbox_raw_priv *priv) { diff --git a/arch/sandbox/dts/sandbox.dts b/arch/sandbox/dts/sandbox.dts index 0ea2452742..e7f6c194d0 100644 --- a/arch/sandbox/dts/sandbox.dts +++ b/arch/sandbox/dts/sandbox.dts @@ -56,7 +56,7 @@ eth@80000000 { compatible = "sandbox,eth-raw"; reg = <0x80000000 0x1000>; - host-raw-interface = "eth0"; + host-raw-interface-idx = <2>; };
eth@90000000 { diff --git a/arch/sandbox/include/asm/eth-raw-os.h b/arch/sandbox/include/asm/eth-raw-os.h index f986d3142d..70b29b4508 100644 --- a/arch/sandbox/include/asm/eth-raw-os.h +++ b/arch/sandbox/include/asm/eth-raw-os.h @@ -29,7 +29,9 @@ struct eth_sandbox_raw_priv { };
int sandbox_eth_raw_os_start(const char *ifname, unsigned char *ethmac, - struct eth_sandbox_raw_priv *priv); + struct eth_sandbox_raw_priv *priv); +int sandbox_eth_raw_os_start_idx(unsigned int ifindex, unsigned char *ethmac, + struct eth_sandbox_raw_priv *priv); int sandbox_eth_raw_os_send(void *packet, int length, struct eth_sandbox_raw_priv *priv); int sandbox_eth_raw_os_recv(void *packet, int *length, diff --git a/drivers/net/sandbox-raw.c b/drivers/net/sandbox-raw.c index f835a6a7f3..3bde8d84ce 100644 --- a/drivers/net/sandbox-raw.c +++ b/drivers/net/sandbox-raw.c @@ -22,6 +22,7 @@ static int sb_eth_raw_start(struct udevice *dev) struct eth_sandbox_raw_priv *priv = dev_get_priv(dev); struct eth_pdata *pdata = dev_get_platdata(dev); const char *interface; + u32 index;
debug("eth_sandbox_raw: Start\n");
@@ -37,7 +38,12 @@ static int sb_eth_raw_start(struct udevice *dev) priv); }
- return -EINVAL; + if (dev_read_u32(dev, "host-raw-interface-idx", &index) < 0) + return -EINVAL; + + printf("eth_sandbox_raw: Using interface index %d from DT ", index); + + return sandbox_eth_raw_os_start_idx(index, pdata->enetaddr, priv); }
static int sb_eth_raw_send(struct udevice *dev, void *packet, int length)
Hi Joe,
On 26 June 2018 at 14:19, Joe Hershberger joe.hershberger@ni.com wrote:
With systemd stable interface names, eth0 will almost never exist. Instead of using that name in the sandbox.dts, use index 2 (the first interface after "lo"). Enumerate the interfaces on the system to choose a valid interace name.
Signed-off-by: Joe Hershberger joe.hershberger@ni.com
arch/sandbox/cpu/eth-raw-os.c | 28 ++++++++++++++++++++++++++++ arch/sandbox/dts/sandbox.dts | 2 +- arch/sandbox/include/asm/eth-raw-os.h | 4 +++- drivers/net/sandbox-raw.c | 8 +++++++- 4 files changed, 39 insertions(+), 3 deletions(-)
This seems brittle too. A name seems better to me. I might be missing something though. Can we ask the OS for a name, etc?
Regards, Simon
On Tue, Jun 26, 2018 at 6:18 PM, Simon Glass sjg@chromium.org wrote:
Hi Joe,
On 26 June 2018 at 14:19, Joe Hershberger joe.hershberger@ni.com wrote:
With systemd stable interface names, eth0 will almost never exist. Instead of using that name in the sandbox.dts, use index 2 (the first interface after "lo"). Enumerate the interfaces on the system to choose a valid interace name.
Signed-off-by: Joe Hershberger joe.hershberger@ni.com
arch/sandbox/cpu/eth-raw-os.c | 28 ++++++++++++++++++++++++++++ arch/sandbox/dts/sandbox.dts | 2 +- arch/sandbox/include/asm/eth-raw-os.h | 4 +++- drivers/net/sandbox-raw.c | 8 +++++++- 4 files changed, 39 insertions(+), 3 deletions(-)
This seems brittle too. A name seems better to me. I might be missing something though. Can we ask the OS for a name, etc?
This asks for the equivalent of "eth0", but asks the OS to tell us the actual name. I don't know that we can possibly choose the correct network interface to use across any machine configuration. Choosing the first one, whatever it's name is, seems like a pretty reliable way to get a good interface in most cases.
Regards, Simon _______________________________________________ U-Boot mailing list U-Boot@lists.denx.de https://lists.denx.de/listinfo/u-boot
Hi Joe,
On 26 June 2018 at 16:31, Joe Hershberger joe.hershberger@ni.com wrote:
On Tue, Jun 26, 2018 at 6:18 PM, Simon Glass sjg@chromium.org wrote:
Hi Joe,
On 26 June 2018 at 14:19, Joe Hershberger joe.hershberger@ni.com wrote:
With systemd stable interface names, eth0 will almost never exist. Instead of using that name in the sandbox.dts, use index 2 (the first interface after "lo"). Enumerate the interfaces on the system to choose a valid interace name.
Signed-off-by: Joe Hershberger joe.hershberger@ni.com
arch/sandbox/cpu/eth-raw-os.c | 28 ++++++++++++++++++++++++++++ arch/sandbox/dts/sandbox.dts | 2 +- arch/sandbox/include/asm/eth-raw-os.h | 4 +++- drivers/net/sandbox-raw.c | 8 +++++++- 4 files changed, 39 insertions(+), 3 deletions(-)
This seems brittle too. A name seems better to me. I might be missing something though. Can we ask the OS for a name, etc?
This asks for the equivalent of "eth0", but asks the OS to tell us the actual name. I don't know that we can possibly choose the correct network interface to use across any machine configuration. Choosing the first one, whatever it's name is, seems like a pretty reliable way to get a good interface in most cases.
OK well let's see how it works out.
Reviewed-by: Simon Glass sjg@chromium.org
Regards, Simon
On 06/26/2018 11:19 PM, Joe Hershberger wrote:
With systemd stable interface names, eth0 will almost never exist. Instead of using that name in the sandbox.dts, use index 2 (the first interface after "lo"). Enumerate the interfaces on the system to choose a valid interace name.
Signed-off-by: Joe Hershberger joe.hershberger@ni.com
Hello Joe,
thanks for addressing this issue.
On my laptop the raw interfaces are currently called: 1 - lo - loopback 2 - enp0s25 - wired ethernet 3 - wlp12s0 - WLAN
But this list is not static:
It is allowable to define multiple loopback interfaces. Additional bridges, tap and tun interfaces may be defined. A computer may also have multiple wired and wireless interfaces. And of cause I may choose to plug in an additional ethernet adapter while my computer is running.
The interface index does not tell which interface is actually connected to the network. My laptop uses wpl12s0 (index = 3) for sending out this mail. So, please, do not replace one invalid assumption (eth0) by another (index = 2).
I doubt there is a guarantee that index 1 is a loopback interface on POSIX systems. If you need a special treatment of loopback devices you could analyze the properties of each interface to detect which one is a loopback interface.
There is no need for using the device tree for the specification of interfaces. Instead a raw ethernet device can be set up for each interface. Let the user choose at runtime which device he wants to use for testing.
Best regards
Heinrich
arch/sandbox/cpu/eth-raw-os.c | 28 ++++++++++++++++++++++++++++ arch/sandbox/dts/sandbox.dts | 2 +- arch/sandbox/include/asm/eth-raw-os.h | 4 +++- drivers/net/sandbox-raw.c | 8 +++++++- 4 files changed, 39 insertions(+), 3 deletions(-)
diff --git a/arch/sandbox/cpu/eth-raw-os.c b/arch/sandbox/cpu/eth-raw-os.c index 61f23ed210..11b5271f31 100644 --- a/arch/sandbox/cpu/eth-raw-os.c +++ b/arch/sandbox/cpu/eth-raw-os.c @@ -150,6 +150,34 @@ int sandbox_eth_raw_os_start(const char *ifname, unsigned char *ethmac, return _raw_packet_start(ifname, ethmac, priv); }
+int sandbox_eth_raw_os_start_idx(unsigned int ifindex, unsigned char *ethmac,
struct eth_sandbox_raw_priv *priv)+{
- int ret;
- struct if_nameindex *ni, *i;
- const char *ifname = NULL;
- ni = if_nameindex();
- if (!ni)
return -EINVAL;- for (i = ni; !(i->if_index == 0 && !i->if_name); i++) {
if (ifindex == i->if_index) {ifname = i->if_name;printf("(%s)\n", ifname);break;}- }
- /* Don't bother supporting 'lo' by index. Just use the name in DTS */
- if (ifname)
ret = _raw_packet_start(ifname, ethmac, priv);- else
ret = -EINVAL;- if_freenameindex(ni);
- return ret;
+}
int sandbox_eth_raw_os_send(void *packet, int length, struct eth_sandbox_raw_priv *priv) { diff --git a/arch/sandbox/dts/sandbox.dts b/arch/sandbox/dts/sandbox.dts index 0ea2452742..e7f6c194d0 100644 --- a/arch/sandbox/dts/sandbox.dts +++ b/arch/sandbox/dts/sandbox.dts @@ -56,7 +56,7 @@ eth@80000000 { compatible = "sandbox,eth-raw"; reg = <0x80000000 0x1000>;
host-raw-interface = "eth0";
host-raw-interface-idx = <2>;};
eth@90000000 {
diff --git a/arch/sandbox/include/asm/eth-raw-os.h b/arch/sandbox/include/asm/eth-raw-os.h index f986d3142d..70b29b4508 100644 --- a/arch/sandbox/include/asm/eth-raw-os.h +++ b/arch/sandbox/include/asm/eth-raw-os.h @@ -29,7 +29,9 @@ struct eth_sandbox_raw_priv { };
int sandbox_eth_raw_os_start(const char *ifname, unsigned char *ethmac,
struct eth_sandbox_raw_priv *priv);
struct eth_sandbox_raw_priv *priv);+int sandbox_eth_raw_os_start_idx(unsigned int ifindex, unsigned char *ethmac,
struct eth_sandbox_raw_priv *priv);int sandbox_eth_raw_os_send(void *packet, int length, struct eth_sandbox_raw_priv *priv); int sandbox_eth_raw_os_recv(void *packet, int *length, diff --git a/drivers/net/sandbox-raw.c b/drivers/net/sandbox-raw.c index f835a6a7f3..3bde8d84ce 100644 --- a/drivers/net/sandbox-raw.c +++ b/drivers/net/sandbox-raw.c @@ -22,6 +22,7 @@ static int sb_eth_raw_start(struct udevice *dev) struct eth_sandbox_raw_priv *priv = dev_get_priv(dev); struct eth_pdata *pdata = dev_get_platdata(dev); const char *interface;
u32 index;
debug("eth_sandbox_raw: Start\n");
@@ -37,7 +38,12 @@ static int sb_eth_raw_start(struct udevice *dev) priv); }
- return -EINVAL;
- if (dev_read_u32(dev, "host-raw-interface-idx", &index) < 0)
return -EINVAL;- printf("eth_sandbox_raw: Using interface index %d from DT ", index);
- return sandbox_eth_raw_os_start_idx(index, pdata->enetaddr, priv);
}
static int sb_eth_raw_send(struct udevice *dev, void *packet, int length)
Hi Heinrich,
On Tue, Jun 26, 2018 at 8:15 PM, Heinrich Schuchardt xypron.glpk@gmx.de wrote:
On 06/26/2018 11:19 PM, Joe Hershberger wrote:
With systemd stable interface names, eth0 will almost never exist. Instead of using that name in the sandbox.dts, use index 2 (the first interface after "lo"). Enumerate the interfaces on the system to choose a valid interace name.
Signed-off-by: Joe Hershberger joe.hershberger@ni.com
Hello Joe,
thanks for addressing this issue.
On my laptop the raw interfaces are currently called: 1 - lo - loopback 2 - enp0s25 - wired ethernet 3 - wlp12s0 - WLAN
But this list is not static:
It is allowable to define multiple loopback interfaces. Additional bridges, tap and tun interfaces may be defined. A computer may also have multiple wired and wireless interfaces. And of cause I may choose to plug in an additional ethernet adapter while my computer is running.
The interface index does not tell which interface is actually connected to the network. My laptop uses wpl12s0 (index = 3) for sending out this mail. So, please, do not replace one invalid assumption (eth0) by another (index = 2).
While index=2 is not an ideal assumption, I don't think invalid is a fair description. It is far better than eth0, which on modern systems is very unlikely. The driver is probed based on DTS entries. This is simply for manual testing, so if the common situation doesn't apply to a person, they can easily modify the DTS to specify the name of their interface, or they can specify a different index. I don't think the fact that it doesn't ideally fit every possible situation warrants changing how these are probed to somehow be dynamically created based on the OS.
I doubt there is a guarantee that index 1 is a loopback interface on POSIX systems. If you need a special treatment of loopback devices you could analyze the properties of each interface to detect which one is a loopback interface.
The only special treatment for local interfaces is that they don't work with raw sockets. You can see that in the code. At the same time, the local interface is pretty crippled for testing the network stack anyway (no ICMP support) so if there is more than one (why would you?) then that one not named "lo" will not work.
There is no need for using the device tree for the specification of interfaces. Instead a raw ethernet device can be set up for each interface. Let the user choose at runtime which device he wants to use for testing.
You say there is no need, but that's the way it works in the sandbox. There is no "OS" subsystem that is there to enumerate new interfaces dynamically, the way something like PCI on a real board would. I don't think we would want it completely automatic anyway, because things like the test.dts have only fake sandbox eth interfaces, not any raw, since we want the unit test to run consistently regardless of the arrangement of network interfaces on the test system it runs on.
Cheers, -Joe
Best regards
Heinrich
arch/sandbox/cpu/eth-raw-os.c | 28 ++++++++++++++++++++++++++++ arch/sandbox/dts/sandbox.dts | 2 +- arch/sandbox/include/asm/eth-raw-os.h | 4 +++- drivers/net/sandbox-raw.c | 8 +++++++- 4 files changed, 39 insertions(+), 3 deletions(-)
diff --git a/arch/sandbox/cpu/eth-raw-os.c b/arch/sandbox/cpu/eth-raw-os.c index 61f23ed210..11b5271f31 100644 --- a/arch/sandbox/cpu/eth-raw-os.c +++ b/arch/sandbox/cpu/eth-raw-os.c @@ -150,6 +150,34 @@ int sandbox_eth_raw_os_start(const char *ifname, unsigned char *ethmac, return _raw_packet_start(ifname, ethmac, priv); }
+int sandbox_eth_raw_os_start_idx(unsigned int ifindex, unsigned char *ethmac,
struct eth_sandbox_raw_priv *priv)+{
int ret;struct if_nameindex *ni, *i;const char *ifname = NULL;ni = if_nameindex();if (!ni)return -EINVAL;for (i = ni; !(i->if_index == 0 && !i->if_name); i++) {if (ifindex == i->if_index) {ifname = i->if_name;printf("(%s)\n", ifname);break;}}/* Don't bother supporting 'lo' by index. Just use the name in DTS */if (ifname)ret = _raw_packet_start(ifname, ethmac, priv);elseret = -EINVAL;if_freenameindex(ni);return ret;+}
int sandbox_eth_raw_os_send(void *packet, int length, struct eth_sandbox_raw_priv *priv) { diff --git a/arch/sandbox/dts/sandbox.dts b/arch/sandbox/dts/sandbox.dts index 0ea2452742..e7f6c194d0 100644 --- a/arch/sandbox/dts/sandbox.dts +++ b/arch/sandbox/dts/sandbox.dts @@ -56,7 +56,7 @@ eth@80000000 { compatible = "sandbox,eth-raw"; reg = <0x80000000 0x1000>;
host-raw-interface = "eth0";
host-raw-interface-idx = <2>; }; eth@90000000 {diff --git a/arch/sandbox/include/asm/eth-raw-os.h b/arch/sandbox/include/asm/eth-raw-os.h index f986d3142d..70b29b4508 100644 --- a/arch/sandbox/include/asm/eth-raw-os.h +++ b/arch/sandbox/include/asm/eth-raw-os.h @@ -29,7 +29,9 @@ struct eth_sandbox_raw_priv { };
int sandbox_eth_raw_os_start(const char *ifname, unsigned char *ethmac,
struct eth_sandbox_raw_priv *priv);
struct eth_sandbox_raw_priv *priv);+int sandbox_eth_raw_os_start_idx(unsigned int ifindex, unsigned char *ethmac,
struct eth_sandbox_raw_priv *priv);int sandbox_eth_raw_os_send(void *packet, int length, struct eth_sandbox_raw_priv *priv); int sandbox_eth_raw_os_recv(void *packet, int *length, diff --git a/drivers/net/sandbox-raw.c b/drivers/net/sandbox-raw.c index f835a6a7f3..3bde8d84ce 100644 --- a/drivers/net/sandbox-raw.c +++ b/drivers/net/sandbox-raw.c @@ -22,6 +22,7 @@ static int sb_eth_raw_start(struct udevice *dev) struct eth_sandbox_raw_priv *priv = dev_get_priv(dev); struct eth_pdata *pdata = dev_get_platdata(dev); const char *interface;
u32 index; debug("eth_sandbox_raw: Start\n");@@ -37,7 +38,12 @@ static int sb_eth_raw_start(struct udevice *dev) priv); }
return -EINVAL;
if (dev_read_u32(dev, "host-raw-interface-idx", &index) < 0)return -EINVAL;printf("eth_sandbox_raw: Using interface index %d from DT ", index);return sandbox_eth_raw_os_start_idx(index, pdata->enetaddr, priv);}
static int sb_eth_raw_send(struct udevice *dev, void *packet, int length)
U-Boot mailing list U-Boot@lists.denx.de https://lists.denx.de/listinfo/u-boot
On 06/27/2018 06:58 AM, Joe Hershberger wrote:
Hi Heinrich,
On Tue, Jun 26, 2018 at 8:15 PM, Heinrich Schuchardt xypron.glpk@gmx.de wrote:
On 06/26/2018 11:19 PM, Joe Hershberger wrote:
With systemd stable interface names, eth0 will almost never exist. Instead of using that name in the sandbox.dts, use index 2 (the first interface after "lo"). Enumerate the interfaces on the system to choose a valid interace name.
Signed-off-by: Joe Hershberger joe.hershberger@ni.com
Hello Joe,
thanks for addressing this issue.
On my laptop the raw interfaces are currently called: 1 - lo - loopback 2 - enp0s25 - wired ethernet 3 - wlp12s0 - WLAN
But this list is not static:
It is allowable to define multiple loopback interfaces. Additional bridges, tap and tun interfaces may be defined. A computer may also have multiple wired and wireless interfaces. And of cause I may choose to plug in an additional ethernet adapter while my computer is running.
The interface index does not tell which interface is actually connected to the network. My laptop uses wpl12s0 (index = 3) for sending out this mail. So, please, do not replace one invalid assumption (eth0) by another (index = 2).
While index=2 is not an ideal assumption, I don't think invalid is a fair description. It is far better than eth0, which on modern systems
Using a laptop over WLAN is not special. So many users will be using an interface with index=3.
So your assumption that all computers are created equal (using index=2) is invalid.
is very unlikely. The driver is probed based on DTS entries. This is simply for manual testing, so if the common situation doesn't apply to a person, they can easily modify the DTS to specify the name of their interface, or they can specify a different index. I don't think the fact that it doesn't ideally fit every possible situation warrants changing how these are probed to somehow be dynamically created based on the OS.
I doubt there is a guarantee that index 1 is a loopback interface on POSIX systems. If you need a special treatment of loopback devices you could analyze the properties of each interface to detect which one is a loopback interface.
The only special treatment for local interfaces is that they don't work with raw sockets. You can see that in the code. At the same time, the local interface is pretty crippled for testing the network stack anyway (no ICMP support) so if there is more than one (why would you?) then that one not named "lo" will not work.
There is no need for using the device tree for the specification of interfaces. Instead a raw ethernet device can be set up for each interface. Let the user choose at runtime which device he wants to use for testing.
You say there is no need, but that's the way it works in the sandbox. There is no "OS" subsystem that is there to enumerate new interfaces dynamically, the way something like PCI on a real board would. I don't think we would want it completely automatic anyway, because things like the test.dts have only fake sandbox eth interfaces, not any raw, since we want the unit test to run consistently regardless of the arrangement of network interfaces on the test system it runs on.
NAK until the Sandbox works in the same way both on my laptop (WLAN, index=3) and on my workstation (wired, index=2) without manually patching source code.
Best regards
Heinrich
Cheers, -Joe
Best regards
Heinrich
arch/sandbox/cpu/eth-raw-os.c | 28 ++++++++++++++++++++++++++++ arch/sandbox/dts/sandbox.dts | 2 +- arch/sandbox/include/asm/eth-raw-os.h | 4 +++- drivers/net/sandbox-raw.c | 8 +++++++- 4 files changed, 39 insertions(+), 3 deletions(-)
diff --git a/arch/sandbox/cpu/eth-raw-os.c b/arch/sandbox/cpu/eth-raw-os.c index 61f23ed210..11b5271f31 100644 --- a/arch/sandbox/cpu/eth-raw-os.c +++ b/arch/sandbox/cpu/eth-raw-os.c @@ -150,6 +150,34 @@ int sandbox_eth_raw_os_start(const char *ifname, unsigned char *ethmac, return _raw_packet_start(ifname, ethmac, priv); }
+int sandbox_eth_raw_os_start_idx(unsigned int ifindex, unsigned char *ethmac,
struct eth_sandbox_raw_priv *priv)+{
int ret;struct if_nameindex *ni, *i;const char *ifname = NULL;ni = if_nameindex();if (!ni)return -EINVAL;for (i = ni; !(i->if_index == 0 && !i->if_name); i++) {if (ifindex == i->if_index) {ifname = i->if_name;printf("(%s)\n", ifname);break;}}/* Don't bother supporting 'lo' by index. Just use the name in DTS */if (ifname)ret = _raw_packet_start(ifname, ethmac, priv);elseret = -EINVAL;if_freenameindex(ni);return ret;+}
int sandbox_eth_raw_os_send(void *packet, int length, struct eth_sandbox_raw_priv *priv) { diff --git a/arch/sandbox/dts/sandbox.dts b/arch/sandbox/dts/sandbox.dts index 0ea2452742..e7f6c194d0 100644 --- a/arch/sandbox/dts/sandbox.dts +++ b/arch/sandbox/dts/sandbox.dts @@ -56,7 +56,7 @@ eth@80000000 { compatible = "sandbox,eth-raw"; reg = <0x80000000 0x1000>;
host-raw-interface = "eth0";
host-raw-interface-idx = <2>; }; eth@90000000 {diff --git a/arch/sandbox/include/asm/eth-raw-os.h b/arch/sandbox/include/asm/eth-raw-os.h index f986d3142d..70b29b4508 100644 --- a/arch/sandbox/include/asm/eth-raw-os.h +++ b/arch/sandbox/include/asm/eth-raw-os.h @@ -29,7 +29,9 @@ struct eth_sandbox_raw_priv { };
int sandbox_eth_raw_os_start(const char *ifname, unsigned char *ethmac,
struct eth_sandbox_raw_priv *priv);
struct eth_sandbox_raw_priv *priv);+int sandbox_eth_raw_os_start_idx(unsigned int ifindex, unsigned char *ethmac,
struct eth_sandbox_raw_priv *priv);int sandbox_eth_raw_os_send(void *packet, int length, struct eth_sandbox_raw_priv *priv); int sandbox_eth_raw_os_recv(void *packet, int *length, diff --git a/drivers/net/sandbox-raw.c b/drivers/net/sandbox-raw.c index f835a6a7f3..3bde8d84ce 100644 --- a/drivers/net/sandbox-raw.c +++ b/drivers/net/sandbox-raw.c @@ -22,6 +22,7 @@ static int sb_eth_raw_start(struct udevice *dev) struct eth_sandbox_raw_priv *priv = dev_get_priv(dev); struct eth_pdata *pdata = dev_get_platdata(dev); const char *interface;
u32 index; debug("eth_sandbox_raw: Start\n");@@ -37,7 +38,12 @@ static int sb_eth_raw_start(struct udevice *dev) priv); }
return -EINVAL;
if (dev_read_u32(dev, "host-raw-interface-idx", &index) < 0)return -EINVAL;printf("eth_sandbox_raw: Using interface index %d from DT ", index);return sandbox_eth_raw_os_start_idx(index, pdata->enetaddr, priv);}
static int sb_eth_raw_send(struct udevice *dev, void *packet, int length)
U-Boot mailing list U-Boot@lists.denx.de https://lists.denx.de/listinfo/u-boot
On Wed, Jun 27, 2018 at 12:55 AM, Heinrich Schuchardt xypron.glpk@gmx.de wrote:
On 06/27/2018 06:58 AM, Joe Hershberger wrote:
Hi Heinrich,
On Tue, Jun 26, 2018 at 8:15 PM, Heinrich Schuchardt xypron.glpk@gmx.de wrote:
On 06/26/2018 11:19 PM, Joe Hershberger wrote:
With systemd stable interface names, eth0 will almost never exist. Instead of using that name in the sandbox.dts, use index 2 (the first interface after "lo"). Enumerate the interfaces on the system to choose a valid interace name.
Signed-off-by: Joe Hershberger joe.hershberger@ni.com
Hello Joe,
thanks for addressing this issue.
On my laptop the raw interfaces are currently called: 1 - lo - loopback 2 - enp0s25 - wired ethernet 3 - wlp12s0 - WLAN
But this list is not static:
It is allowable to define multiple loopback interfaces. Additional bridges, tap and tun interfaces may be defined. A computer may also have multiple wired and wireless interfaces. And of cause I may choose to plug in an additional ethernet adapter while my computer is running.
The interface index does not tell which interface is actually connected to the network. My laptop uses wpl12s0 (index = 3) for sending out this mail. So, please, do not replace one invalid assumption (eth0) by another (index = 2).
While index=2 is not an ideal assumption, I don't think invalid is a fair description. It is far better than eth0, which on modern systems
Using a laptop over WLAN is not special. So many users will be using an interface with index=3.
So your assumption that all computers are created equal (using index=2) is invalid.
is very unlikely. The driver is probed based on DTS entries. This is simply for manual testing, so if the common situation doesn't apply to a person, they can easily modify the DTS to specify the name of their interface, or they can specify a different index. I don't think the fact that it doesn't ideally fit every possible situation warrants changing how these are probed to somehow be dynamically created based on the OS.
I doubt there is a guarantee that index 1 is a loopback interface on POSIX systems. If you need a special treatment of loopback devices you could analyze the properties of each interface to detect which one is a loopback interface.
The only special treatment for local interfaces is that they don't work with raw sockets. You can see that in the code. At the same time, the local interface is pretty crippled for testing the network stack anyway (no ICMP support) so if there is more than one (why would you?) then that one not named "lo" will not work.
There is no need for using the device tree for the specification of interfaces. Instead a raw ethernet device can be set up for each interface. Let the user choose at runtime which device he wants to use for testing.
You say there is no need, but that's the way it works in the sandbox. There is no "OS" subsystem that is there to enumerate new interfaces dynamically, the way something like PCI on a real board would. I don't think we would want it completely automatic anyway, because things like the test.dts have only fake sandbox eth interfaces, not any raw, since we want the unit test to run consistently regardless of the arrangement of network interfaces on the test system it runs on.
NAK until the Sandbox works in the same way both on my laptop (WLAN, index=3) and on my workstation (wired, index=2) without manually patching source code.
So it sounds like you'd be fine if there were simply 2 entries in the DTS... one that lists index=2 and another that lists index=3.
Best regards
Heinrich
Cheers, -Joe
Best regards
Heinrich
arch/sandbox/cpu/eth-raw-os.c | 28 ++++++++++++++++++++++++++++ arch/sandbox/dts/sandbox.dts | 2 +- arch/sandbox/include/asm/eth-raw-os.h | 4 +++- drivers/net/sandbox-raw.c | 8 +++++++- 4 files changed, 39 insertions(+), 3 deletions(-)
diff --git a/arch/sandbox/cpu/eth-raw-os.c b/arch/sandbox/cpu/eth-raw-os.c index 61f23ed210..11b5271f31 100644 --- a/arch/sandbox/cpu/eth-raw-os.c +++ b/arch/sandbox/cpu/eth-raw-os.c @@ -150,6 +150,34 @@ int sandbox_eth_raw_os_start(const char *ifname, unsigned char *ethmac, return _raw_packet_start(ifname, ethmac, priv); }
+int sandbox_eth_raw_os_start_idx(unsigned int ifindex, unsigned char *ethmac,
struct eth_sandbox_raw_priv *priv)+{
int ret;struct if_nameindex *ni, *i;const char *ifname = NULL;ni = if_nameindex();if (!ni)return -EINVAL;for (i = ni; !(i->if_index == 0 && !i->if_name); i++) {if (ifindex == i->if_index) {ifname = i->if_name;printf("(%s)\n", ifname);break;}}/* Don't bother supporting 'lo' by index. Just use the name in DTS */if (ifname)ret = _raw_packet_start(ifname, ethmac, priv);elseret = -EINVAL;if_freenameindex(ni);return ret;+}
int sandbox_eth_raw_os_send(void *packet, int length, struct eth_sandbox_raw_priv *priv) { diff --git a/arch/sandbox/dts/sandbox.dts b/arch/sandbox/dts/sandbox.dts index 0ea2452742..e7f6c194d0 100644 --- a/arch/sandbox/dts/sandbox.dts +++ b/arch/sandbox/dts/sandbox.dts @@ -56,7 +56,7 @@ eth@80000000 { compatible = "sandbox,eth-raw"; reg = <0x80000000 0x1000>;
host-raw-interface = "eth0";
host-raw-interface-idx = <2>; }; eth@90000000 {diff --git a/arch/sandbox/include/asm/eth-raw-os.h b/arch/sandbox/include/asm/eth-raw-os.h index f986d3142d..70b29b4508 100644 --- a/arch/sandbox/include/asm/eth-raw-os.h +++ b/arch/sandbox/include/asm/eth-raw-os.h @@ -29,7 +29,9 @@ struct eth_sandbox_raw_priv { };
int sandbox_eth_raw_os_start(const char *ifname, unsigned char *ethmac,
struct eth_sandbox_raw_priv *priv);
struct eth_sandbox_raw_priv *priv);+int sandbox_eth_raw_os_start_idx(unsigned int ifindex, unsigned char *ethmac,
struct eth_sandbox_raw_priv *priv);int sandbox_eth_raw_os_send(void *packet, int length, struct eth_sandbox_raw_priv *priv); int sandbox_eth_raw_os_recv(void *packet, int *length, diff --git a/drivers/net/sandbox-raw.c b/drivers/net/sandbox-raw.c index f835a6a7f3..3bde8d84ce 100644 --- a/drivers/net/sandbox-raw.c +++ b/drivers/net/sandbox-raw.c @@ -22,6 +22,7 @@ static int sb_eth_raw_start(struct udevice *dev) struct eth_sandbox_raw_priv *priv = dev_get_priv(dev); struct eth_pdata *pdata = dev_get_platdata(dev); const char *interface;
u32 index; debug("eth_sandbox_raw: Start\n");@@ -37,7 +38,12 @@ static int sb_eth_raw_start(struct udevice *dev) priv); }
return -EINVAL;
if (dev_read_u32(dev, "host-raw-interface-idx", &index) < 0)return -EINVAL;printf("eth_sandbox_raw: Using interface index %d from DT ", index);return sandbox_eth_raw_os_start_idx(index, pdata->enetaddr, priv);}
static int sb_eth_raw_send(struct udevice *dev, void *packet, int length)
U-Boot mailing list U-Boot@lists.denx.de https://lists.denx.de/listinfo/u-boot
U-Boot mailing list U-Boot@lists.denx.de https://lists.denx.de/listinfo/u-boot
On 06/27/2018 08:01 AM, Joe Hershberger wrote:
On Wed, Jun 27, 2018 at 12:55 AM, Heinrich Schuchardt xypron.glpk@gmx.de wrote:
On 06/27/2018 06:58 AM, Joe Hershberger wrote:
Hi Heinrich,
On Tue, Jun 26, 2018 at 8:15 PM, Heinrich Schuchardt xypron.glpk@gmx.de wrote:
On 06/26/2018 11:19 PM, Joe Hershberger wrote:
With systemd stable interface names, eth0 will almost never exist. Instead of using that name in the sandbox.dts, use index 2 (the first interface after "lo"). Enumerate the interfaces on the system to choose a valid interace name.
Signed-off-by: Joe Hershberger joe.hershberger@ni.com
Hello Joe,
thanks for addressing this issue.
On my laptop the raw interfaces are currently called: 1 - lo - loopback 2 - enp0s25 - wired ethernet 3 - wlp12s0 - WLAN
But this list is not static:
It is allowable to define multiple loopback interfaces. Additional bridges, tap and tun interfaces may be defined. A computer may also have multiple wired and wireless interfaces. And of cause I may choose to plug in an additional ethernet adapter while my computer is running.
The interface index does not tell which interface is actually connected to the network. My laptop uses wpl12s0 (index = 3) for sending out this mail. So, please, do not replace one invalid assumption (eth0) by another (index = 2).
While index=2 is not an ideal assumption, I don't think invalid is a fair description. It is far better than eth0, which on modern systems
Using a laptop over WLAN is not special. So many users will be using an interface with index=3.
So your assumption that all computers are created equal (using index=2) is invalid.
is very unlikely. The driver is probed based on DTS entries. This is simply for manual testing, so if the common situation doesn't apply to a person, they can easily modify the DTS to specify the name of their interface, or they can specify a different index. I don't think the fact that it doesn't ideally fit every possible situation warrants changing how these are probed to somehow be dynamically created based on the OS.
I doubt there is a guarantee that index 1 is a loopback interface on POSIX systems. If you need a special treatment of loopback devices you could analyze the properties of each interface to detect which one is a loopback interface.
The only special treatment for local interfaces is that they don't work with raw sockets. You can see that in the code. At the same time, the local interface is pretty crippled for testing the network stack anyway (no ICMP support) so if there is more than one (why would you?) then that one not named "lo" will not work.
There is no need for using the device tree for the specification of interfaces. Instead a raw ethernet device can be set up for each interface. Let the user choose at runtime which device he wants to use for testing.
You say there is no need, but that's the way it works in the sandbox. There is no "OS" subsystem that is there to enumerate new interfaces dynamically, the way something like PCI on a real board would. I don't think we would want it completely automatic anyway, because things like the test.dts have only fake sandbox eth interfaces, not any raw, since we want the unit test to run consistently regardless of the arrangement of network interfaces on the test system it runs on.
NAK until the Sandbox works in the same way both on my laptop (WLAN, index=3) and on my workstation (wired, index=2) without manually patching source code.
So it sounds like you'd be fine if there were simply 2 entries in the DTS... one that lists index=2 and another that lists index=3.
No, you didn't get the point. First there is no guarantee that the loopback is index 1. And there is no guarantee that index 2 or 3 represent the connection to the internet.
Would you accept a patch that will only work if the power connector is for 110V/60 Hz and volume is measured in imperial gallons?
Best regards
Heinrich
Best regards
Heinrich
Cheers, -Joe
Best regards
Heinrich
arch/sandbox/cpu/eth-raw-os.c | 28 ++++++++++++++++++++++++++++ arch/sandbox/dts/sandbox.dts | 2 +- arch/sandbox/include/asm/eth-raw-os.h | 4 +++- drivers/net/sandbox-raw.c | 8 +++++++- 4 files changed, 39 insertions(+), 3 deletions(-)
diff --git a/arch/sandbox/cpu/eth-raw-os.c b/arch/sandbox/cpu/eth-raw-os.c index 61f23ed210..11b5271f31 100644 --- a/arch/sandbox/cpu/eth-raw-os.c +++ b/arch/sandbox/cpu/eth-raw-os.c @@ -150,6 +150,34 @@ int sandbox_eth_raw_os_start(const char *ifname, unsigned char *ethmac, return _raw_packet_start(ifname, ethmac, priv); }
+int sandbox_eth_raw_os_start_idx(unsigned int ifindex, unsigned char *ethmac,
struct eth_sandbox_raw_priv *priv)+{
int ret;struct if_nameindex *ni, *i;const char *ifname = NULL;ni = if_nameindex();if (!ni)return -EINVAL;for (i = ni; !(i->if_index == 0 && !i->if_name); i++) {if (ifindex == i->if_index) {ifname = i->if_name;printf("(%s)\n", ifname);break;}}/* Don't bother supporting 'lo' by index. Just use the name in DTS */if (ifname)ret = _raw_packet_start(ifname, ethmac, priv);elseret = -EINVAL;if_freenameindex(ni);return ret;+}
- int sandbox_eth_raw_os_send(void *packet, int length, struct eth_sandbox_raw_priv *priv) {
diff --git a/arch/sandbox/dts/sandbox.dts b/arch/sandbox/dts/sandbox.dts index 0ea2452742..e7f6c194d0 100644 --- a/arch/sandbox/dts/sandbox.dts +++ b/arch/sandbox/dts/sandbox.dts @@ -56,7 +56,7 @@ eth@80000000 { compatible = "sandbox,eth-raw"; reg = <0x80000000 0x1000>;
host-raw-interface = "eth0";
host-raw-interface-idx = <2>; }; eth@90000000 {diff --git a/arch/sandbox/include/asm/eth-raw-os.h b/arch/sandbox/include/asm/eth-raw-os.h index f986d3142d..70b29b4508 100644 --- a/arch/sandbox/include/asm/eth-raw-os.h +++ b/arch/sandbox/include/asm/eth-raw-os.h @@ -29,7 +29,9 @@ struct eth_sandbox_raw_priv { };
int sandbox_eth_raw_os_start(const char *ifname, unsigned char *ethmac,
struct eth_sandbox_raw_priv *priv);
struct eth_sandbox_raw_priv *priv);+int sandbox_eth_raw_os_start_idx(unsigned int ifindex, unsigned char *ethmac,
struct eth_sandbox_raw_priv *priv);diff --git a/drivers/net/sandbox-raw.c b/drivers/net/sandbox-raw.c index f835a6a7f3..3bde8d84ce 100644 --- a/drivers/net/sandbox-raw.c +++ b/drivers/net/sandbox-raw.c @@ -22,6 +22,7 @@ static int sb_eth_raw_start(struct udevice *dev) struct eth_sandbox_raw_priv *priv = dev_get_priv(dev); struct eth_pdata *pdata = dev_get_platdata(dev); const char *interface;
u32 index; debug("eth_sandbox_raw: Start\n");@@ -37,7 +38,12 @@ static int sb_eth_raw_start(struct udevice *dev) priv); }
return -EINVAL;
if (dev_read_u32(dev, "host-raw-interface-idx", &index) < 0)return -EINVAL;printf("eth_sandbox_raw: Using interface index %d from DT ", index);return sandbox_eth_raw_os_start_idx(index, pdata->enetaddr, priv);}
static int sb_eth_raw_send(struct udevice *dev, void *packet, int length)
U-Boot mailing list U-Boot@lists.denx.de https://lists.denx.de/listinfo/u-boot
U-Boot mailing list U-Boot@lists.denx.de https://lists.denx.de/listinfo/u-boot
participants (3)
-
Heinrich Schuchardt -
Joe Hershberger -
Simon Glass