2020-02-19 09:04:12

by Martin Haass

Subject: [PATCH] module support: during lockdown, log name of unsigned module

during lockdown loading of unsigned modules is restricted to signed
modules only. The old error message does not show which module misses
the signature, making it very difficult for a user to determine which
module is at fault.
This patch adds a line to the logs which additionally contains the
module name that caused the error message. The old message cannot
be replaced as it is generated by lockdown_is_locked_down
---
kernel/module.c | 10 ++++++++--
1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/kernel/module.c b/kernel/module.c
index 33569a01d6e..6dcb28139a0 100644
--- a/kernel/module.c
+++ b/kernel/module.c
@@ -2807,7 +2807,8 @@ static int module_sig_check(struct load_info *info,
int flags)
const unsigned long markerlen = sizeof(MODULE_SIG_STRING) - 1;
const char *reason;
const void *mod = info->hdr;
-
+ int is_locked = -EPERM;
+
/*
* Require flags == 0, as a module with version information
* removed is no longer the module that was signed
@@ -2843,7 +2844,12 @@ static int module_sig_check(struct load_info *info,
int flags)
return -EKEYREJECTED;
}

- return security_locked_down(LOCKDOWN_MODULE_SIGNATURE);
+ is_locked = security_locked_down(LOCKDOWN_MODULE_SIGNATURE);
+ if (is_locked == -EPERM) {
+ pr_notice("Lockdown: %s: rejected module '%s' cause: %s",
+ current->comm, info->name, reason);
+ }
+ return is_locked;

/* All other errors are fatal, including nomem, unparseable
* signatures and signature check failures - even if signatures
--
2.25.0

2020-02-26 17:59:26

by Jessica Yu

Subject: Re: [PATCH] module support: during lockdown, log name of unsigned module

+++ Martin Haass [19/02/20 10:02 +0100]:
>during lockdown loading of unsigned modules is restricted to signed
>modules only. The old error message does not show which module misses
>the signature, making it very difficult for a user to determine which
>module is at fault.
>This patch adds a line to the logs which additionally contains the
>module name that caused the error message. The old message cannot
>be replaced as it is generated by lockdown_is_locked_down
>---
> kernel/module.c | 10 ++++++++--
> 1 file changed, 8 insertions(+), 2 deletions(-)
>
>diff --git a/kernel/module.c b/kernel/module.c
>index 33569a01d6e..6dcb28139a0 100644
>--- a/kernel/module.c
>+++ b/kernel/module.c
>@@ -2807,7 +2807,8 @@ static int module_sig_check(struct load_info *info,
>int flags)
> const unsigned long markerlen = sizeof(MODULE_SIG_STRING) - 1;
> const char *reason;
> const void *mod = info->hdr;
>-
>+ int is_locked = -EPERM;
>+
> /*
> * Require flags == 0, as a module with version information
> * removed is no longer the module that was signed
>@@ -2843,7 +2844,12 @@ static int module_sig_check(struct load_info *info,
>int flags)
> return -EKEYREJECTED;
> }
>
>- return security_locked_down(LOCKDOWN_MODULE_SIGNATURE);
>+ is_locked = security_locked_down(LOCKDOWN_MODULE_SIGNATURE);
>+ if (is_locked == -EPERM) {
>+ pr_notice("Lockdown: %s: rejected module '%s' cause: %s",
>+ current->comm, info->name, reason);
>+ }
>+ return is_locked;

Hi!

Actually, I think we can just reuse the pr_notice() from the previous if
(is_module_sig_enforced()) block. It already logs the module name as well as
the reason. And we'd better leave the lockdown-specific messages to the LSM.
Something like this perhaps?

diff --git a/kernel/module.c b/kernel/module.c
index b88ec9cd2a7f..2c881e3b9d92 100644
--- a/kernel/module.c
+++ b/kernel/module.c
@@ -2838,12 +2838,13 @@ static int module_sig_check(struct load_info *info, int flags)
case -ENOKEY:
reason = "Loading of module with unavailable key";
decide:
- if (is_module_sig_enforced()) {
+ err = is_module_sig_enforced() ? \
+ -EKEYREJECTED : security_locked_down(LOCKDOWN_MODULE_SIGNATURE);
+
+ if (err)
pr_notice("%s: %s is rejected\n", info->name, reason);
- return -EKEYREJECTED;
- }

- return security_locked_down(LOCKDOWN_MODULE_SIGNATURE);
+ return err;

/* All other errors are fatal, including nomem, unparseable
* signatures and signature check failures - even if signatures