All of lore.kernel.org
 help / color / mirror / Atom feed
From: Michael Ellerman <mpe@ellerman.id.au>
To: "Michal Suchánek" <msuchanek@suse.de>
Cc: "Jason A. Donenfeld" <Jason@zx2c4.com>,
	Herbert Xu <herbert@gondor.apana.org.au>,
	netdev@vger.kernel.org, dtsen@linux.ibm.com,
	linuxppc-dev@lists.ozlabs.org, wireguard@lists.zx2c4.com
Subject: Re: Cannot load wireguard module
Date: Wed, 20 Mar 2024 23:41:32 +1100	[thread overview]
Message-ID: <87le6dyt1f.fsf@mail.lhotse> (raw)
In-Reply-To: <20240319124742.GM20665@kitsune.suse.cz>

Michal Suchánek <msuchanek@suse.de> writes:
> On Mon, Mar 18, 2024 at 06:08:55PM +0100, Michal Suchánek wrote:
>> On Mon, Mar 18, 2024 at 10:50:49PM +1100, Michael Ellerman wrote:
>> > Michael Ellerman <mpe@ellerman.id.au> writes:
>> > > Michal Suchánek <msuchanek@suse.de> writes:
>> > >> Hello,
>> > >>
>> > >> I cannot load the wireguard module.
>> > >>
>> > >> Loading the module provides no diagnostic other than 'No such device'.
>> > >>
>> > >> Please provide maningful diagnostics for loading software-only driver,
>> > >> clearly there is no particular device needed.
>> > >
>> > > Presumably it's just bubbling up an -ENODEV from somewhere.
>> > >
>> > > Can you get a trace of it?
>> > >
>> > > Something like:
>> > >
>> > >   # trace-cmd record -p function_graph -F modprobe wireguard
>
> Attached.

Sorry :/, you need to also trace children of modprobe, with -c.

But, I was able to reproduce the same issue here.

On a P9, a kernel with CONFIG_CRYPTO_CHACHA20_P10=n everything works:

  $ modprobe -v wireguard
  insmod /lib/modules/6.8.0/kernel/net/ipv4/udp_tunnel.ko
  insmod /lib/modules/6.8.0/kernel/net/ipv6/ip6_udp_tunnel.ko
  insmod /lib/modules/6.8.0/kernel/lib/crypto/libchacha.ko
  insmod /lib/modules/6.8.0/kernel/lib/crypto/libchacha20poly1305.ko
  insmod /lib/modules/6.8.0/kernel/drivers/net/wireguard/wireguard.ko
  [   19.180564][  T692] wireguard: allowedips self-tests: pass
  [   19.185080][  T692] wireguard: nonce counter self-tests: pass
  [   19.310438][  T692] wireguard: ratelimiter self-tests: pass
  [   19.310639][  T692] wireguard: WireGuard 1.0.0 loaded. See www.wireguard.com for information.
  [   19.310746][  T692] wireguard: Copyright (C) 2015-2019 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved.


If I build CONFIG_CRYPTO_CHACHA20_P10 as a module then it breaks:

  $ modprobe -v wireguard
  insmod /lib/modules/6.8.0/kernel/net/ipv4/udp_tunnel.ko
  insmod /lib/modules/6.8.0/kernel/net/ipv6/ip6_udp_tunnel.ko
  insmod /lib/modules/6.8.0/kernel/lib/crypto/libchacha.ko
  insmod /lib/modules/6.8.0/kernel/arch/powerpc/crypto/chacha-p10-crypto.ko
  modprobe: ERROR: could not insert 'wireguard': No such device


The ENODEV is coming from module_cpu_feature_match(), which blocks the
driver from loading on non-p10.

Looking at other arches (arm64 at least) it seems like the driver should
instead be loading but disabling the p10 path. Which then allows
chacha_crypt_arch() to exist, and it has a fallback to use
chacha_crypt_generic().

I don't see how module_cpu_feature_match() can co-exist with the driver
also providing a fallback. Hopefully someone who knows crypto better
than me can explain it.

This diff fixes it for me:

diff --git a/arch/powerpc/crypto/chacha-p10-glue.c b/arch/powerpc/crypto/chacha-p10-glue.c
index 74fb86b0d209..9d2c30b0904c 100644
--- a/arch/powerpc/crypto/chacha-p10-glue.c
+++ b/arch/powerpc/crypto/chacha-p10-glue.c
@@ -197,6 +197,9 @@ static struct skcipher_alg algs[] = {
 
 static int __init chacha_p10_init(void)
 {
+	if (!cpu_has_feature(PPC_FEATURE2_ARCH_3_1))
+		return 0;
+
 	static_branch_enable(&have_p10);
 
 	return crypto_register_skciphers(algs, ARRAY_SIZE(algs));
@@ -207,7 +210,7 @@ static void __exit chacha_p10_exit(void)
 	crypto_unregister_skciphers(algs, ARRAY_SIZE(algs));
 }
 
-module_cpu_feature_match(PPC_MODULE_FEATURE_P10, chacha_p10_init);
+module_init(chacha_p10_init);
 module_exit(chacha_p10_exit);
 
 MODULE_DESCRIPTION("ChaCha and XChaCha stream ciphers (P10 accelerated)");


Giving me:

  $ modprobe -v wireguard
  insmod /lib/modules/6.8.0-dirty/kernel/net/ipv4/udp_tunnel.ko
  insmod /lib/modules/6.8.0-dirty/kernel/net/ipv6/ip6_udp_tunnel.ko
  insmod /lib/modules/6.8.0-dirty/kernel/lib/crypto/libchacha.ko
  insmod /lib/modules/6.8.0-dirty/kernel/arch/powerpc/crypto/chacha-p10-crypto.ko
  insmod /lib/modules/6.8.0-dirty/kernel/lib/crypto/libchacha20poly1305.ko
  insmod /lib/modules/6.8.0-dirty/kernel/drivers/net/wireguard/wireguard.ko
  [   19.657941][  T718] wireguard: allowedips self-tests: pass
  [   19.662501][  T718] wireguard: nonce counter self-tests: pass
  [   19.782933][  T718] wireguard: ratelimiter self-tests: pass
  [   19.783114][  T718] wireguard: WireGuard 1.0.0 loaded. See www.wireguard.com for information.
  [   19.783223][  T718] wireguard: Copyright (C) 2015-2019 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved.
  

cheers

WARNING: multiple messages have this Message-ID (diff)
From: Michael Ellerman <mpe@ellerman.id.au>
To: "Michal Suchánek" <msuchanek@suse.de>
Cc: netdev@vger.kernel.org, "Jason A. Donenfeld" <Jason@zx2c4.com>,
	linuxppc-dev@lists.ozlabs.org, wireguard@lists.zx2c4.com,
	dtsen@linux.ibm.com, Herbert Xu <herbert@gondor.apana.org.au>
Subject: Re: Cannot load wireguard module
Date: Wed, 20 Mar 2024 23:41:32 +1100	[thread overview]
Message-ID: <87le6dyt1f.fsf@mail.lhotse> (raw)
In-Reply-To: <20240319124742.GM20665@kitsune.suse.cz>

Michal Suchánek <msuchanek@suse.de> writes:
> On Mon, Mar 18, 2024 at 06:08:55PM +0100, Michal Suchánek wrote:
>> On Mon, Mar 18, 2024 at 10:50:49PM +1100, Michael Ellerman wrote:
>> > Michael Ellerman <mpe@ellerman.id.au> writes:
>> > > Michal Suchánek <msuchanek@suse.de> writes:
>> > >> Hello,
>> > >>
>> > >> I cannot load the wireguard module.
>> > >>
>> > >> Loading the module provides no diagnostic other than 'No such device'.
>> > >>
>> > >> Please provide maningful diagnostics for loading software-only driver,
>> > >> clearly there is no particular device needed.
>> > >
>> > > Presumably it's just bubbling up an -ENODEV from somewhere.
>> > >
>> > > Can you get a trace of it?
>> > >
>> > > Something like:
>> > >
>> > >   # trace-cmd record -p function_graph -F modprobe wireguard
>
> Attached.

Sorry :/, you need to also trace children of modprobe, with -c.

But, I was able to reproduce the same issue here.

On a P9, a kernel with CONFIG_CRYPTO_CHACHA20_P10=n everything works:

  $ modprobe -v wireguard
  insmod /lib/modules/6.8.0/kernel/net/ipv4/udp_tunnel.ko
  insmod /lib/modules/6.8.0/kernel/net/ipv6/ip6_udp_tunnel.ko
  insmod /lib/modules/6.8.0/kernel/lib/crypto/libchacha.ko
  insmod /lib/modules/6.8.0/kernel/lib/crypto/libchacha20poly1305.ko
  insmod /lib/modules/6.8.0/kernel/drivers/net/wireguard/wireguard.ko
  [   19.180564][  T692] wireguard: allowedips self-tests: pass
  [   19.185080][  T692] wireguard: nonce counter self-tests: pass
  [   19.310438][  T692] wireguard: ratelimiter self-tests: pass
  [   19.310639][  T692] wireguard: WireGuard 1.0.0 loaded. See www.wireguard.com for information.
  [   19.310746][  T692] wireguard: Copyright (C) 2015-2019 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved.


If I build CONFIG_CRYPTO_CHACHA20_P10 as a module then it breaks:

  $ modprobe -v wireguard
  insmod /lib/modules/6.8.0/kernel/net/ipv4/udp_tunnel.ko
  insmod /lib/modules/6.8.0/kernel/net/ipv6/ip6_udp_tunnel.ko
  insmod /lib/modules/6.8.0/kernel/lib/crypto/libchacha.ko
  insmod /lib/modules/6.8.0/kernel/arch/powerpc/crypto/chacha-p10-crypto.ko
  modprobe: ERROR: could not insert 'wireguard': No such device


The ENODEV is coming from module_cpu_feature_match(), which blocks the
driver from loading on non-p10.

Looking at other arches (arm64 at least) it seems like the driver should
instead be loading but disabling the p10 path. Which then allows
chacha_crypt_arch() to exist, and it has a fallback to use
chacha_crypt_generic().

I don't see how module_cpu_feature_match() can co-exist with the driver
also providing a fallback. Hopefully someone who knows crypto better
than me can explain it.

This diff fixes it for me:

diff --git a/arch/powerpc/crypto/chacha-p10-glue.c b/arch/powerpc/crypto/chacha-p10-glue.c
index 74fb86b0d209..9d2c30b0904c 100644
--- a/arch/powerpc/crypto/chacha-p10-glue.c
+++ b/arch/powerpc/crypto/chacha-p10-glue.c
@@ -197,6 +197,9 @@ static struct skcipher_alg algs[] = {
 
 static int __init chacha_p10_init(void)
 {
+	if (!cpu_has_feature(PPC_FEATURE2_ARCH_3_1))
+		return 0;
+
 	static_branch_enable(&have_p10);
 
 	return crypto_register_skciphers(algs, ARRAY_SIZE(algs));
@@ -207,7 +210,7 @@ static void __exit chacha_p10_exit(void)
 	crypto_unregister_skciphers(algs, ARRAY_SIZE(algs));
 }
 
-module_cpu_feature_match(PPC_MODULE_FEATURE_P10, chacha_p10_init);
+module_init(chacha_p10_init);
 module_exit(chacha_p10_exit);
 
 MODULE_DESCRIPTION("ChaCha and XChaCha stream ciphers (P10 accelerated)");


Giving me:

  $ modprobe -v wireguard
  insmod /lib/modules/6.8.0-dirty/kernel/net/ipv4/udp_tunnel.ko
  insmod /lib/modules/6.8.0-dirty/kernel/net/ipv6/ip6_udp_tunnel.ko
  insmod /lib/modules/6.8.0-dirty/kernel/lib/crypto/libchacha.ko
  insmod /lib/modules/6.8.0-dirty/kernel/arch/powerpc/crypto/chacha-p10-crypto.ko
  insmod /lib/modules/6.8.0-dirty/kernel/lib/crypto/libchacha20poly1305.ko
  insmod /lib/modules/6.8.0-dirty/kernel/drivers/net/wireguard/wireguard.ko
  [   19.657941][  T718] wireguard: allowedips self-tests: pass
  [   19.662501][  T718] wireguard: nonce counter self-tests: pass
  [   19.782933][  T718] wireguard: ratelimiter self-tests: pass
  [   19.783114][  T718] wireguard: WireGuard 1.0.0 loaded. See www.wireguard.com for information.
  [   19.783223][  T718] wireguard: Copyright (C) 2015-2019 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved.
  

cheers

  reply	other threads:[~2024-03-20 12:42 UTC|newest]

Thread overview: 22+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-03-15 12:20 Cannot load wireguard module Michal Suchánek
2024-03-15 12:37 ` Christophe Leroy
2024-03-15 13:05   ` Michal Suchánek
2024-03-15 13:05     ` Michal Suchánek
2024-03-15 23:44 ` Michael Ellerman
2024-03-15 23:44   ` Michael Ellerman
2024-03-18 11:50   ` Michael Ellerman
2024-03-18 11:50     ` Michael Ellerman
2024-03-18 17:08     ` Michal Suchánek
2024-03-18 17:08       ` Michal Suchánek
2024-03-19 12:47       ` Michal Suchánek
2024-03-19 12:47         ` Michal Suchánek
2024-03-20 12:41         ` Michael Ellerman [this message]
2024-03-20 12:41           ` Michael Ellerman
2024-03-20 16:04           ` Michal Suchánek
2024-03-20 16:04             ` Michal Suchánek
2024-03-21 12:47             ` Michael Ellerman
2024-03-21 12:47               ` Michael Ellerman
2024-03-22 11:33           ` Herbert Xu
2024-03-22 11:33             ` Herbert Xu
2024-03-23  0:18             ` Michael Ellerman
2024-03-23  0:18               ` Michael Ellerman

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=87le6dyt1f.fsf@mail.lhotse \
    --to=mpe@ellerman.id.au \
    --cc=Jason@zx2c4.com \
    --cc=dtsen@linux.ibm.com \
    --cc=herbert@gondor.apana.org.au \
    --cc=linuxppc-dev@lists.ozlabs.org \
    --cc=msuchanek@suse.de \
    --cc=netdev@vger.kernel.org \
    --cc=wireguard@lists.zx2c4.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.