The platform's RNG must be available before random_init() in order to be
useful for initial seeding, which in turn means that it needs to be
called from setup_arch(), rather than from an init call. Fortunately,
each platform already has a setup_arch function pointer, which means we
can wire it up that way. Complicating things, however, is that POWER8
systems need some per-cpu state and kmalloc, which isn't available at
this stage. So we split things up into an early phase and a later
opportunistic phase. This commit also removes some noisy log messages
that don't add much.

Cc: stable@vger.kernel.org
Cc: Michael Ellerman <mpe@ellerman.id.au>
Reviewed-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Fixes: a4da0d50b2a0 ("powerpc: Implement arch_get_random_long/int() for powernv")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
---
 arch/powerpc/platforms/powernv/powernv.h |  2 +
 arch/powerpc/platforms/powernv/rng.c     | 47 ++++++++++++++++--------
 arch/powerpc/platforms/powernv/setup.c   |  2 +
 3 files changed, 35 insertions(+), 16 deletions(-)

diff --git a/arch/powerpc/platforms/powernv/powernv.h b/arch/powerpc/platforms/powernv/powernv.h
index e297bf4abfcb..fd3f5e1eb10b 100644
--- a/arch/powerpc/platforms/powernv/powernv.h
+++ b/arch/powerpc/platforms/powernv/powernv.h
@@ -42,4 +42,6 @@ ssize_t memcons_copy(struct memcons *mc, char *to, loff_t pos, size_t count);
 u32 __init memcons_get_size(struct memcons *mc);
 struct memcons *__init memcons_init(struct device_node *node, const char *mc_prop_name);
 
+void powernv_rng_init(void);
+
 #endif /* _POWERNV_H */
diff --git a/arch/powerpc/platforms/powernv/rng.c b/arch/powerpc/platforms/powernv/rng.c
index e3d44b36ae98..e332e72e1857 100644
--- a/arch/powerpc/platforms/powernv/rng.c
+++ b/arch/powerpc/platforms/powernv/rng.c
@@ -17,6 +17,7 @@
 #include <asm/prom.h>
 #include <asm/machdep.h>
 #include <asm/smp.h>
+#include "powernv.h"
 
 #define DARN_ERR 0xFFFFFFFFFFFFFFFFul
 
@@ -28,7 +29,6 @@ struct powernv_rng {
 
 static DEFINE_PER_CPU(struct powernv_rng *, powernv_rng);
 
-
 int powernv_hwrng_present(void)
 {
 	struct powernv_rng *rng;
@@ -98,9 +98,6 @@ static int __init initialise_darn(void)
 			return 0;
 		}
 	}
-
-	pr_warn("Unable to use DARN for get_random_seed()\n");
-
 	return -EIO;
 }
 
@@ -163,32 +160,50 @@ static __init int rng_create(struct device_node *dn)
 
 	rng_init_per_cpu(rng, dn);
 
-	pr_info_once("Registering arch random hook.\n");
-
 	ppc_md.get_random_seed = powernv_get_random_long;
 
 	return 0;
 }
 
-static __init int rng_init(void)
+static int __init powernv_get_random_long_early(unsigned long *v)
 {
 	struct device_node *dn;
-	int rc;
+
+	if (!slab_is_available())
+		return 0;
+
+	if (cmpxchg(&ppc_md.get_random_seed, powernv_get_random_long_early,
+		    NULL) != powernv_get_random_long_early)
+		return 0;
 
 	for_each_compatible_node(dn, NULL, "ibm,power-rng") {
-		rc = rng_create(dn);
-		if (rc) {
-			pr_err("Failed creating rng for %pOF (%d).\n",
-				dn, rc);
+		if (rng_create(dn))
 			continue;
-		}
-
 		/* Create devices for hwrng driver */
 		of_platform_device_create(dn, NULL, NULL);
 	}
 
-	initialise_darn();
+	if (!ppc_md.get_random_seed)
+		return 0;
+	return ppc_md.get_random_seed(v);
+}
+
+void __init powernv_rng_init(void)
+{
+	/* Prefer darn over the rest. */
+	if (!initialise_darn())
+		return;
+
+	if (of_find_compatible_node(NULL, NULL, "ibm,power-rng"))
+		ppc_md.get_random_seed = powernv_get_random_long_early;
+}
 
+static int __init powernv_rng_late_init(void)
+{
+	unsigned long v;
+	/* In case it wasn't called during init for some other reason. */
+	if (ppc_md.get_random_seed == powernv_get_random_long_early)
+		powernv_get_random_long_early(&v);
 	return 0;
 }
-machine_subsys_initcall(powernv, rng_init);
+machine_subsys_initcall(powernv, powernv_rng_late_init);
diff --git a/arch/powerpc/platforms/powernv/setup.c b/arch/powerpc/platforms/powernv/setup.c
index 824c3ad7a0fa..a5fcb6796b22 100644
--- a/arch/powerpc/platforms/powernv/setup.c
+++ b/arch/powerpc/platforms/powernv/setup.c
@@ -203,6 +203,8 @@ static void __init pnv_setup_arch(void)
 	pnv_check_guarded_cores();
 
 	/* XXX PMCS */
+
+	powernv_rng_init();
 }
 
 static void __init pnv_init(void)
-- 
2.35.1

On Tue, 21 Jun 2022 16:08:49 +0200, Jason A. Donenfeld wrote:
> The platform's RNG must be available before random_init() in order to be
> useful for initial seeding, which in turn means that it needs to be
> called from setup_arch(), rather than from an init call. Fortunately,
> each platform already has a setup_arch function pointer, which means we
> can wire it up that way. Complicating things, however, is that POWER8
> systems need some per-cpu state and kmalloc, which isn't available at
> this stage. So we split things up into an early phase and a later
> opportunistic phase. This commit also removes some noisy log messages
> that don't add much.
> 
> [...]

Applied to powerpc/fixes.

[1/1] powerpc/powernv: wire up rng during setup_arch
      https://git.kernel.org/powerpc/c/f3eac426657d985b97c92fa5f7ae1d43f04721f3

cheers

Le 21/06/2022 à 16:08, Jason A. Donenfeld a écrit :
> The platform's RNG must be available before random_init() in order to be
> useful for initial seeding, which in turn means that it needs to be
> called from setup_arch(), rather than from an init call. Fortunately,
> each platform already has a setup_arch function pointer, which means we
> can wire it up that way. Complicating things, however, is that POWER8
> systems need some per-cpu state and kmalloc, which isn't available at
> this stage. So we split things up into an early phase and a later
> opportunistic phase. This commit also removes some noisy log messages
> that don't add much.

Regarding the kmalloc(), I have not looked at it in details, but usually 
you can use memblock_alloc() when kmalloc is not available yet.

Christophe

Hi Christophe,

On Tue, Jun 21, 2022 at 06:33:11PM +0000, Christophe Leroy wrote:
> Le 21/06/2022 à 16:08, Jason A. Donenfeld a écrit :
> > The platform's RNG must be available before random_init() in order to be
> > useful for initial seeding, which in turn means that it needs to be
> > called from setup_arch(), rather than from an init call. Fortunately,
> > each platform already has a setup_arch function pointer, which means we
> > can wire it up that way. Complicating things, however, is that POWER8
> > systems need some per-cpu state and kmalloc, which isn't available at
> > this stage. So we split things up into an early phase and a later
> > opportunistic phase. This commit also removes some noisy log messages
> > that don't add much.
> 
> Regarding the kmalloc(), I have not looked at it in details, but usually 
> you can use memblock_alloc() when kmalloc is not available yet.

That seems a bit excessive, especially as those allocations are long
lived. And we don't even *need* it that early, but just before
random_init(). Michael is running this v5 on the test rig overnight, so
we'll learn in the Australian morning whether this finally did the trick
(I hope).

Jason

Le 21/06/2022 à 20:47, Jason A. Donenfeld a écrit :
> Hi Christophe,
> 
> On Tue, Jun 21, 2022 at 06:33:11PM +0000, Christophe Leroy wrote:
>> Le 21/06/2022 à 16:08, Jason A. Donenfeld a écrit :
>>> The platform's RNG must be available before random_init() in order to be
>>> useful for initial seeding, which in turn means that it needs to be
>>> called from setup_arch(), rather than from an init call. Fortunately,
>>> each platform already has a setup_arch function pointer, which means we
>>> can wire it up that way. Complicating things, however, is that POWER8
>>> systems need some per-cpu state and kmalloc, which isn't available at
>>> this stage. So we split things up into an early phase and a later
>>> opportunistic phase. This commit also removes some noisy log messages
>>> that don't add much.
>>
>> Regarding the kmalloc(), I have not looked at it in details, but usually
>> you can use memblock_alloc() when kmalloc is not available yet.
> 
> That seems a bit excessive, especially as those allocations are long
> lived. And we don't even *need* it that early, but just before
> random_init(). Michael is running this v5 on the test rig overnight, so
> we'll learn in the Australian morning whether this finally did the trick
> (I hope).
> 

The fact that they are long lived make them a good candidate for 
memblock_alloc().

But fair enough, if they are not required that early then just do it later.

Christophe

Christophe Leroy <christophe.leroy@csgroup.eu> writes:
> Le 21/06/2022 à 20:47, Jason A. Donenfeld a écrit :
>> On Tue, Jun 21, 2022 at 06:33:11PM +0000, Christophe Leroy wrote:
>>> Le 21/06/2022 à 16:08, Jason A. Donenfeld a écrit :
>>>> The platform's RNG must be available before random_init() in order to be
>>>> useful for initial seeding, which in turn means that it needs to be
>>>> called from setup_arch(), rather than from an init call. Fortunately,
>>>> each platform already has a setup_arch function pointer, which means we
>>>> can wire it up that way. Complicating things, however, is that POWER8
>>>> systems need some per-cpu state and kmalloc, which isn't available at
>>>> this stage. So we split things up into an early phase and a later
>>>> opportunistic phase. This commit also removes some noisy log messages
>>>> that don't add much.
>>>
>>> Regarding the kmalloc(), I have not looked at it in details, but usually
>>> you can use memblock_alloc() when kmalloc is not available yet.
>> 
>> That seems a bit excessive, especially as those allocations are long
>> lived. And we don't even *need* it that early, but just before
>> random_init(). Michael is running this v5 on the test rig overnight, so
>> we'll learn in the Australian morning whether this finally did the trick
>> (I hope).
>
> The fact that they are long lived make them a good candidate for 
> memblock_alloc().
>
> But fair enough, if they are not required that early then just do it later.

memblock works but then we trip on ioremap vs early_ioremap.

Fixing that is a bit of a pain as we'd have to stop using of_iomap() and
we'd also need to switch the mappings to ioremap() later in boot.

We'd also have to defer the percpu initialisation.

So it's all just a bit of a pain when we actually only need to get the
hook ready before random_init() which is called much later in boot when
slab/ioremap/percpu are all ready.

cheers