I'm trying to reduce the time taken in the kernel in initcalls, with
my aim being to reduce the current ~700ms spent in initcalls before
userspace, down to something like 100ms. All times on my Broadwell-U
laptop, under virtualization. The purpose of this is to be able to
launch VMs around containers with minimal overhead, like Intel Clear
Containers, but using standard distro kernels and qemu.
Currently the kernel spends 28ms (on my laptop) running crypto
algorithm self-tests. Although it's possibe to disable these at
compile time, Fedora kernel maintainers want to maintain a single
kernel image for all uses.
So this commit adds a runtime flag which callers can set to skip the
self-tests in the fast container/virtualization case.
Rich.
Running self-tests for a short-lived KVM VM takes 28ms on my laptop.
This commit adds a flag 'cryptomgr.notests' which allows them to be
disabled.
Signed-off-by: Richard W.M. Jones <[email protected]>
---
Documentation/kernel-parameters.txt | 3 +++
crypto/testmgr.c | 9 +++++++++
2 files changed, 12 insertions(+)
diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt
index 0b3de80..d4d5fb7 100644
--- a/Documentation/kernel-parameters.txt
+++ b/Documentation/kernel-parameters.txt
@@ -826,6 +826,9 @@ bytes respectively. Such letter suffixes can also be entirely omitted.
It will be ignored when crashkernel=X,high is not used
or memory reserved is below 4G.
+ cryptomgr.notests
+ [KNL] Disable crypto self-tests
+
cs89x0_dma= [HW,NET]
Format: <dma>
diff --git a/crypto/testmgr.c b/crypto/testmgr.c
index b86883a..dc613f2 100644
--- a/crypto/testmgr.c
+++ b/crypto/testmgr.c
@@ -35,6 +35,10 @@
#include "internal.h"
+static bool notests;
+module_param(notests, bool, 0644);
+MODULE_PARM_DESC(notests, "disable crypto self-tests");
+
#ifdef CONFIG_CRYPTO_MANAGER_DISABLE_TESTS
/* a perfect nop */
@@ -3868,6 +3872,11 @@ int alg_test(const char *driver, const char *alg, u32 type, u32 mask)
int j;
int rc;
+ if (notests) {
+ pr_info("alg: self-tests disabled\n");
+ return 0;
+ }
+
alg_test_descs_check_order();
if ((type & CRYPTO_ALG_TYPE_MASK) == CRYPTO_ALG_TYPE_CIPHER) {
--
2.7.4
Am Freitag, 29. April 2016, 11:07:43 schrieb Richard W.M. Jones:
Hi Richard,
> Running self-tests for a short-lived KVM VM takes 28ms on my laptop.
> This commit adds a flag 'cryptomgr.notests' which allows them to be
> disabled.
>
> Signed-off-by: Richard W.M. Jones <[email protected]>
> ---
> Documentation/kernel-parameters.txt | 3 +++
> crypto/testmgr.c | 9 +++++++++
> 2 files changed, 12 insertions(+)
>
> diff --git a/Documentation/kernel-parameters.txt
> b/Documentation/kernel-parameters.txt index 0b3de80..d4d5fb7 100644
> --- a/Documentation/kernel-parameters.txt
> +++ b/Documentation/kernel-parameters.txt
> @@ -826,6 +826,9 @@ bytes respectively. Such letter suffixes can also be
> entirely omitted. It will be ignored when crashkernel=X,high is not used
> or memory reserved is below 4G.
>
> + cryptomgr.notests
> + [KNL] Disable crypto self-tests
> +
> cs89x0_dma= [HW,NET]
> Format: <dma>
>
> diff --git a/crypto/testmgr.c b/crypto/testmgr.c
> index b86883a..dc613f2 100644
> --- a/crypto/testmgr.c
> +++ b/crypto/testmgr.c
> @@ -35,6 +35,10 @@
>
> #include "internal.h"
>
> +static bool notests;
> +module_param(notests, bool, 0644);
> +MODULE_PARM_DESC(notests, "disable crypto self-tests");
> +
> #ifdef CONFIG_CRYPTO_MANAGER_DISABLE_TESTS
>
> /* a perfect nop */
> @@ -3868,6 +3872,11 @@ int alg_test(const char *driver, const char *alg, u32
> type, u32 mask) int j;
> int rc;
>
> + if (notests) {
What about if (!fips_enabled && notests) ?
I am not sure whether the kernel should prevent mistakes in user space. A
mistake would be when setting fips=1 and notests=1 as the FIPS mode mandates
the self tests.
> + pr_info("alg: self-tests disabled\n");
> + return 0;
> + }
> +
> alg_test_descs_check_order();
>
> if ((type & CRYPTO_ALG_TYPE_MASK) == CRYPTO_ALG_TYPE_CIPHER) {
Ciao
Stephan
--
| Nimm das Recht weg - |
| was ist dann der Staat noch anderes als eine gro?e R?uberbande? |
On Fri, Apr 29, 2016 at 12:59:57PM +0200, Stephan Mueller wrote:
> Am Freitag, 29. April 2016, 11:07:43 schrieb Richard W.M. Jones:
>
> Hi Richard,
[...]
> > + if (notests) {
>
> What about if (!fips_enabled && notests) ?
>
> I am not sure whether the kernel should prevent mistakes in user space. A
> mistake would be when setting fips=1 and notests=1 as the FIPS mode mandates
> the self tests.
(Sorry, I just posted v2 before I saw this message.) I saw the FIPS
stuff and thought about that. Should we prevent mistakes like that?
I really don't know.
Rich.
> > + pr_info("alg: self-tests disabled\n");
> > + return 0;
> > + }
> > +
> > alg_test_descs_check_order();
> >
> > if ((type & CRYPTO_ALG_TYPE_MASK) == CRYPTO_ALG_TYPE_CIPHER) {
>
>
> Ciao
> Stephan
> --
> | Nimm das Recht weg - |
> | was ist dann der Staat noch anderes als eine gro?e R?uberbande? |
--
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
Read my programming and virtualization blog: http://rwmj.wordpress.com
virt-df lists disk usage of guests without needing to install any
software inside the virtual machine. Supports Linux and Windows.
http://people.redhat.com/~rjones/virt-df/