[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <e8c2ce11ec8b6193d891288ce6b3caa0@suse.de>
Date: Wed, 13 Oct 2021 11:05:39 +0300
From: "Ivan T. Ivanov" <iivanov@...e.de>
To: Ard Biesheuvel <ardb@...nel.org>
Cc: Dominik Brodowski <linux@...inikbrodowski.net>,
linux-efi <linux-efi@...r.kernel.org>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [RESEND] Re: [PATCH] Revert "efi/random: Treat EFI_RNG_PROTOCOL
output as bootloader randomness"
On 2021-10-13 10:50, Ard Biesheuvel wrote:
> On Wed, 13 Oct 2021 at 09:30, Ivan T. Ivanov <iivanov@...e.de> wrote:
>>
>> Hi,
>>
>> Quoting Dominik Brodowski (2021-10-12 11:40:34)
>> > Am Tue, Oct 12, 2021 at 11:27:08AM +0300 schrieb Ivan T. Ivanov:
>> > > This reverts commit 18b915ac6b0ac5ba7ded03156860f60a9f16df2b.
>> > >
>> > > When CONFIG_RANDOM_TRUST_BOOTLOADER is enabled add_bootloader_randomness()
>> > > calls add_hwgenerator_randomness() which might sleep,
>> >
>> > Wouldn't it be better to fix add_bootloader_randomness(), considering
>> > that
>> > calls to that function are likely to happen quite early during kernel
>> > initialization? Especially as it seems to have worked beforehand?
>>
>> I have tried. I made wait_event_interruptible() optional, but then
>> crng_reseed() segfault badly. And I don't think crng_reseed() is
>> something that I could fix easily. Suggestions are welcomed ;-)
>>
>
> How about
>
> diff --git a/drivers/char/random.c b/drivers/char/random.c
> index 605969ed0f96..1828dc691ebf 100644
> --- a/drivers/char/random.c
> +++ b/drivers/char/random.c
> @@ -2297,9 +2297,8 @@ EXPORT_SYMBOL_GPL(add_hwgenerator_randomness);
> */
> void add_bootloader_randomness(const void *buf, unsigned int size)
> {
> + add_device_randomness(buf, size);
> if (IS_ENABLED(CONFIG_RANDOM_TRUST_BOOTLOADER))
> - add_hwgenerator_randomness(buf, size, size * 8);
> - else
> - add_device_randomness(buf, size);
> + credit_entropy(&input_pool, size * 8);
> }
> EXPORT_SYMBOL_GPL(add_bootloader_randomness);
This is more or less what I have done. credit_entropy() calls
crng_reseed() which crash badly.
Will check again, but I am sure it will give me same result.
Regards,
Ivan
Powered by blists - more mailing lists