[PATCH] efi_loader: Let networking support depend on NETDEVICES

[PATCH] efi_loader: Let networking support depend on NETDEVICES

[Jan Kiszka]

From: Jan Kiszka <jan.kiszka@siemens.com>

CONFIG_NET does not imply that there are actually network devices
available, only CONFIG_NETDEVICES does. Changing to this dependency
obsoletes the check in Kconfig because NETDEVICES means DM_ETH.

Suggested-by: Tom Rini <trini@konsulko.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
---
 lib/efi_loader/Kconfig           | 1 -
 lib/efi_loader/Makefile          | 2 +-
 lib/efi_loader/efi_device_path.c | 6 +++---
 lib/efi_loader/efi_setup.c       | 2 +-
 lib/efi_selftest/Makefile        | 2 +-
 5 files changed, 6 insertions(+), 7 deletions(-)

diff --git a/lib/efi_loader/Kconfig b/lib/efi_loader/Kconfig
index 41756ea5396..68e6c2531e1 100644
--- a/lib/efi_loader/Kconfig
+++ b/lib/efi_loader/Kconfig
@@ -11,7 +11,6 @@ config EFI_LOADER
 	# We need EFI_STUB_32BIT to be set on x86_32 with EFI_STUB
 	depends on !EFI_STUB || !X86 || X86_64 || EFI_STUB_32BIT
 	depends on BLK
-	depends on DM_ETH || !NET
 	depends on !EFI_APP
 	default y if !ARM || SYS_CPU = armv7 || SYS_CPU = armv8
 	select CHARSET
diff --git a/lib/efi_loader/Makefile b/lib/efi_loader/Makefile
index e187d2a914f..418db296b20 100644
--- a/lib/efi_loader/Makefile
+++ b/lib/efi_loader/Makefile
@@ -68,7 +68,7 @@ obj-y += efi_watchdog.o
 obj-$(CONFIG_EFI_ESRT) += efi_esrt.o
 obj-$(CONFIG_DM_VIDEO) += efi_gop.o
 obj-$(CONFIG_BLK) += efi_disk.o
-obj-$(CONFIG_NET) += efi_net.o
+obj-$(CONFIG_NETDEVICES) += efi_net.o
 obj-$(CONFIG_GENERATE_ACPI_TABLE) += efi_acpi.o
 obj-$(CONFIG_GENERATE_SMBIOS_TABLE) += efi_smbios.o
 obj-$(CONFIG_EFI_RNG_PROTOCOL) += efi_rng.o
diff --git a/lib/efi_loader/efi_device_path.c b/lib/efi_loader/efi_device_path.c
index acae007f26f..d45985a7601 100644
--- a/lib/efi_loader/efi_device_path.c
+++ b/lib/efi_loader/efi_device_path.c
@@ -613,7 +613,7 @@ __maybe_unused static void *dp_fill(void *buf, struct udevice *dev)
 		*vdp = ROOT;
 		return &vdp[1];
 	}
-#ifdef CONFIG_NET
+#ifdef CONFIG_NETDEVICES
 	case UCLASS_ETH: {
 		struct efi_device_path_mac_addr *dp =
 			dp_fill(buf, dev->parent);
@@ -1052,7 +1052,7 @@ struct efi_device_path *efi_dp_from_uart(void)
 	return buf;
 }
 
-#ifdef CONFIG_NET
+#ifdef CONFIG_NETDEVICES
 struct efi_device_path *efi_dp_from_eth(void)
 {
 	void *buf, *start;
@@ -1169,7 +1169,7 @@ efi_status_t efi_dp_from_name(const char *dev, const char *devnr,
 		return EFI_INVALID_PARAMETER;
 
 	if (!strcmp(dev, "Net")) {
-#ifdef CONFIG_NET
+#ifdef CONFIG_NETDEVICES
 		if (device)
 			*device = efi_dp_from_eth();
 #endif
diff --git a/lib/efi_loader/efi_setup.c b/lib/efi_loader/efi_setup.c
index 9d7189336dc..212ce34b31c 100644
--- a/lib/efi_loader/efi_setup.c
+++ b/lib/efi_loader/efi_setup.c
@@ -331,7 +331,7 @@ efi_status_t efi_init_obj_list(void)
 	if (ret != EFI_SUCCESS)
 		goto out;
 #endif
-#ifdef CONFIG_NET
+#ifdef CONFIG_NETDEVICES
 	ret = efi_net_register();
 	if (ret != EFI_SUCCESS)
 		goto out;
diff --git a/lib/efi_selftest/Makefile b/lib/efi_selftest/Makefile
index daac6c39682..e4d75420bff 100644
--- a/lib/efi_selftest/Makefile
+++ b/lib/efi_selftest/Makefile
@@ -50,7 +50,7 @@ efi_selftest_variables_runtime.o \
 efi_selftest_watchdog.o
 
 obj-$(CONFIG_EFI_ECPT) += efi_selftest_ecpt.o
-obj-$(CONFIG_NET) += efi_selftest_snp.o
+obj-$(CONFIG_NETDEVICES) += efi_selftest_snp.o
 
 obj-$(CONFIG_EFI_DEVICE_PATH_TO_TEXT) += efi_selftest_devicepath.o
 obj-$(CONFIG_EFI_UNICODE_COLLATION_PROTOCOL2) += \
-- 
2.35.3

Re: [PATCH] efi_loader: Let networking support depend on NETDEVICES

[Heinrich Schuchardt]

On 10/14/22 18:10, Jan Kiszka wrote:
> From: Jan Kiszka <jan.kiszka@siemens.com>
>
> CONFIG_NET does not imply that there are actually network devices
> available, only CONFIG_NETDEVICES does. Changing to this dependenc
CONFIG_NETDEVICES=y does not imply that any network driver is selected
it is just a tighter prerequisite than CONFIG_NET.

> obsoletes the check in Kconfig because NETDEVICES means DM_ETH.
>
> Suggested-by: Tom Rini <trini@konsulko.com>
> Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>

Thanks for the patch.

It is good to get rid of the dependency for CONFIG_EFI_LOADER.

Reviewed-by: Heinrich Schuchardt <xypron.glpk@gmx.de>

Re: [PATCH] efi_loader: Let networking support depend on NETDEVICES

[Heinrich Schuchardt]

On 10/15/22 08:31, Heinrich Schuchardt wrote:
> On 10/14/22 18:10, Jan Kiszka wrote:
>> From: Jan Kiszka <jan.kiszka@siemens.com>
>>
>> CONFIG_NET does not imply that there are actually network devices
>> available, only CONFIG_NETDEVICES does. Changing to this dependenc
> CONFIG_NETDEVICES=y does not imply that any network driver is selected
> it is just a tighter prerequisite than CONFIG_NET.
>
>> obsoletes the check in Kconfig because NETDEVICES means DM_ETH.
>>
>> Suggested-by: Tom Rini <trini@konsulko.com>
>> Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
>
> Thanks for the patch.
>
> It is good to get rid of the dependency for CONFIG_EFI_LOADER.
>
> Reviewed-by: Heinrich Schuchardt <xypron.glpk@gmx.de>

For LicheePi_Zero_defconfig this patch results in:

arm-linux-gnueabi-ld.bfd: net/bootp.o: in function `dhcp_handler':
net/bootp.c:1080: undefined reference to `efi_net_set_dhcp_ack'

I guess we should add:

diff --git a/net/bootp.c b/net/bootp.c
index b260d5d09e..6c01e38153 100644
--- a/net/bootp.c
+++ b/net/bootp.c
@@ -1077,7 +1077,9 @@ static void dhcp_handler(uchar *pkt, unsigned
dest, struct in_addr sip,
                             strlen(CONFIG_SYS_BOOTFILE_PREFIX)) == 0) {
  #endif /* CONFIG_SYS_BOOTFILE_PREFIX */
                         dhcp_packet_process_options(bp);
-                       efi_net_set_dhcp_ack(pkt, len);
+                       if (CONFIG_IS_ENABLED(EFI_LOADER) &&
+                           CONFIG_IS_ENABLED(NET_DEVICES))
+                               efi_net_set_dhcp_ack(pkt, len);

As this was already incorrect before your patch:

Fixes: 0efe1bcf5c2c ("efi_loader: Add network access support")

Best regards

Heinrich

Re: [PATCH] efi_loader: Let networking support depend on NETDEVICES

[Jan Kiszka]

On 24.10.22 22:47, Heinrich Schuchardt wrote:
> On 10/15/22 08:31, Heinrich Schuchardt wrote:
>> On 10/14/22 18:10, Jan Kiszka wrote:
>>> From: Jan Kiszka <jan.kiszka@siemens.com>
>>>
>>> CONFIG_NET does not imply that there are actually network devices
>>> available, only CONFIG_NETDEVICES does. Changing to this dependenc
>> CONFIG_NETDEVICES=y does not imply that any network driver is selected
>> it is just a tighter prerequisite than CONFIG_NET.
>>
>>> obsoletes the check in Kconfig because NETDEVICES means DM_ETH.
>>>
>>> Suggested-by: Tom Rini <trini@konsulko.com>
>>> Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
>>
>> Thanks for the patch.
>>
>> It is good to get rid of the dependency for CONFIG_EFI_LOADER.
>>
>> Reviewed-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
> 
> For LicheePi_Zero_defconfig this patch results in:
> 
> arm-linux-gnueabi-ld.bfd: net/bootp.o: in function `dhcp_handler':
> net/bootp.c:1080: undefined reference to `efi_net_set_dhcp_ack'
> 
> I guess we should add:
> 
> diff --git a/net/bootp.c b/net/bootp.c
> index b260d5d09e..6c01e38153 100644
> --- a/net/bootp.c
> +++ b/net/bootp.c
> @@ -1077,7 +1077,9 @@ static void dhcp_handler(uchar *pkt, unsigned
> dest, struct in_addr sip,
>                             strlen(CONFIG_SYS_BOOTFILE_PREFIX)) == 0) {
>  #endif /* CONFIG_SYS_BOOTFILE_PREFIX */
>                         dhcp_packet_process_options(bp);
> -                       efi_net_set_dhcp_ack(pkt, len);
> +                       if (CONFIG_IS_ENABLED(EFI_LOADER) &&
> +                           CONFIG_IS_ENABLED(NET_DEVICES))
> +                               efi_net_set_dhcp_ack(pkt, len);
> 
> As this was already incorrect before your patch:
> 
> Fixes: 0efe1bcf5c2c ("efi_loader: Add network access support")
> 
> Best regards
> 
> Heinrich

Will you send this, or should I submit a v2 of my patch with this included?

Jan

-- 
Siemens AG, Technology
Competence Center Embedded Linux