If the device is not ready to provide data the kernel will
be stuck indefinitely in the init function.
This is not a problem if the device is driven using a module,
but if the driver is linked directly into the kernel then the
kernel boot sequence hangs.
This can happen with virtio-rng device with rng-egd backend
with no data provider, for instance with QEMU command line parameters:
...
-chardev socket,id=charrng0,host=localhost,port=2345,server,nowait \
-object rng-egd,id=objrng0,chardev=charrng0 \
-device virtio-rng-pci,rng=objrng0,id=rng0
To avoid that, we can call rng_get_data() in non blocking mode because
the function already manages the case where byte_read is
0 (if the device is not already initialized).
See also commit d3cc7996473a
("hwrng: fetch randomness only after device init")
Signed-off-by: Laurent Vivier <[email protected]>
---
drivers/char/hw_random/core.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/char/hw_random/core.c b/drivers/char/hw_random/core.c
index 95be7228f327..3866d6b8017c 100644
--- a/drivers/char/hw_random/core.c
+++ b/drivers/char/hw_random/core.c
@@ -67,7 +67,7 @@ static void add_early_randomness(struct hwrng *rng)
size_t size = min_t(size_t, 16, rng_buffer_size());
mutex_lock(&reading_mutex);
- bytes_read = rng_get_data(rng, rng_buffer, size, 1);
+ bytes_read = rng_get_data(rng, rng_buffer, size, 0);
mutex_unlock(&reading_mutex);
if (bytes_read > 0)
add_device_randomness(rng_buffer, bytes_read);
--
2.20.1
On Fri, Mar 29, 2019 at 05:30:11PM +0100, Laurent Vivier wrote:
> If the device is not ready to provide data the kernel will
> be stuck indefinitely in the init function.
>
> This is not a problem if the device is driven using a module,
> but if the driver is linked directly into the kernel then the
> kernel boot sequence hangs.
>
> This can happen with virtio-rng device with rng-egd backend
> with no data provider, for instance with QEMU command line parameters:
>
> ...
> -chardev socket,id=charrng0,host=localhost,port=2345,server,nowait \
> -object rng-egd,id=objrng0,chardev=charrng0 \
> -device virtio-rng-pci,rng=objrng0,id=rng0
>
> To avoid that, we can call rng_get_data() in non blocking mode because
> the function already manages the case where byte_read is
> 0 (if the device is not already initialized).
>
> See also commit d3cc7996473a
> ("hwrng: fetch randomness only after device init")
>
> Signed-off-by: Laurent Vivier <[email protected]>
I think this is either not a bug (make sure that the user supplies
a RNG source) or it should be addressed in virtio-rng.
Cheers,
--
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
On 08/04/2019 08:31, Herbert Xu wrote:
> On Fri, Mar 29, 2019 at 05:30:11PM +0100, Laurent Vivier wrote:
>> If the device is not ready to provide data the kernel will
>> be stuck indefinitely in the init function.
>>
>> This is not a problem if the device is driven using a module,
>> but if the driver is linked directly into the kernel then the
>> kernel boot sequence hangs.
>>
>> This can happen with virtio-rng device with rng-egd backend
>> with no data provider, for instance with QEMU command line parameters:
>>
>> ...
>> -chardev socket,id=charrng0,host=localhost,port=2345,server,nowait \
>> -object rng-egd,id=objrng0,chardev=charrng0 \
>> -device virtio-rng-pci,rng=objrng0,id=rng0
>>
>> To avoid that, we can call rng_get_data() in non blocking mode because
>> the function already manages the case where byte_read is
>> 0 (if the device is not already initialized).
>>
>> See also commit d3cc7996473a
>> ("hwrng: fetch randomness only after device init")
>>
>> Signed-off-by: Laurent Vivier <[email protected]>
>
> I think this is either not a bug (make sure that the user supplies
> a RNG source) or it should be addressed in virtio-rng.
Yes, I did more investigation and I agree.
Thanks,
Laurent