[PATCH v2 2/2] modules: add CONFIG_MODPROBE_PATH
Allow the developer to specifiy the initial value of the modprobe_path[] string. This can be used to set it to the empty string initially, thus effectively disabling request_module() during early boot until userspace writes a new value via the /proc/sys/kernel/modprobe interface. [1] When building a custom kernel (often for an embedded target), it's normal to build everything into the kernel that is needed for booting, and indeed the initramfs often contains no modules at all, so every such request_module() done before userspace init has mounted the real rootfs is a waste of time. This is particularly useful when combined with the previous patch, which allowed the initramfs to be unpacked asynchronously - for that to work, it had to make any usermodehelper call wait for the unpacking to finish before attempting to invoke the userspace helper. By eliminating all such (known-to-be-futile) calls of usermodehelper, the initramfs unpacking and the {device,late}_initcalls can proceed in parallel for much longer. For a relatively slow ppc board I'm working on, the two patches combined lead to 0.2s faster boot - but more importantly, the fact that the initramfs unpacking proceeds completely in the background while devices get probed means I get to handle the gpio watchdog in time without getting reset. [1] __request_module() already has an early -ENOENT return when modprobe_path is the empty string. Signed-off-by: Rasmus Villemoes --- init/Kconfig | 12 kernel/kmod.c | 2 +- 2 files changed, 13 insertions(+), 1 deletion(-) diff --git a/init/Kconfig b/init/Kconfig index 22946fe5ded9..18b4ec7346d4 100644 --- a/init/Kconfig +++ b/init/Kconfig @@ -2264,6 +2264,18 @@ config MODULE_ALLOW_MISSING_NAMESPACE_IMPORTS If unsure, say N. +config MODPROBE_PATH + string "Path to modprobe binary" + default "/sbin/modprobe" + help + When kernel code requests a module, it does so by calling + the "modprobe" userspace utility. This option allows you to + set the path where that binary is found. This can be changed + at runtime via the sysctl file + /proc/sys/kernel/modprobe. Setting this to the empty string + removes the kernel's ability to request modules (but + userspace can still load modules explicitly). + config TRIM_UNUSED_KSYMS bool "Trim unused exported kernel symbols" if EXPERT depends on !COMPILE_TEST diff --git a/kernel/kmod.c b/kernel/kmod.c index 3cd075ce2a1e..b717134ebe17 100644 --- a/kernel/kmod.c +++ b/kernel/kmod.c @@ -58,7 +58,7 @@ static DECLARE_WAIT_QUEUE_HEAD(kmod_wq); /* modprobe_path is set via /proc/sys. */ -char modprobe_path[KMOD_PATH_LEN] = "/sbin/modprobe"; +char modprobe_path[KMOD_PATH_LEN] = CONFIG_MODPROBE_PATH; static void free_modprobe_argv(struct subprocess_info *info) { -- 2.29.2
Re: [PATCH v2 2/2] modules: add CONFIG_MODPROBE_PATH
On Tue, Mar 09, 2021 at 10:17:00PM +0100, Rasmus Villemoes wrote: > Allow the developer to specifiy the initial value of the > modprobe_path[] string. This can be used to set it to the empty string > initially, thus effectively disabling request_module() during early > boot until userspace writes a new value via the > /proc/sys/kernel/modprobe interface. [1] > > When building a custom kernel (often for an embedded target), it's > normal to build everything into the kernel that is needed for booting, > and indeed the initramfs often contains no modules at all, so every > such request_module() done before userspace init has mounted the real > rootfs is a waste of time. > > This is particularly useful when combined with the previous patch, > which allowed the initramfs to be unpacked asynchronously - for that > to work, it had to make any usermodehelper call wait for the unpacking > to finish before attempting to invoke the userspace helper. By > eliminating all such (known-to-be-futile) calls of usermodehelper, the > initramfs unpacking and the {device,late}_initcalls can proceed in > parallel for much longer. > > For a relatively slow ppc board I'm working on, the two patches > combined lead to 0.2s faster boot - but more importantly, the fact > that the initramfs unpacking proceeds completely in the background > while devices get probed means I get to handle the gpio watchdog in > time without getting reset. > > [1] __request_module() already has an early -ENOENT return when > modprobe_path is the empty string. > > Signed-off-by: Rasmus Villemoes > --- > init/Kconfig | 12 > kernel/kmod.c | 2 +- > 2 files changed, 13 insertions(+), 1 deletion(-) Looks sane to me, odd we didn't think of doing this before. Reviewed-by: Greg Kroah-Hartman
Re: [PATCH v2 2/2] modules: add CONFIG_MODPROBE_PATH
+++ Rasmus Villemoes [09/03/21 22:17 +0100]: Allow the developer to specifiy the initial value of the modprobe_path[] string. This can be used to set it to the empty string initially, thus effectively disabling request_module() during early boot until userspace writes a new value via the /proc/sys/kernel/modprobe interface. [1] When building a custom kernel (often for an embedded target), it's normal to build everything into the kernel that is needed for booting, and indeed the initramfs often contains no modules at all, so every such request_module() done before userspace init has mounted the real rootfs is a waste of time. This is particularly useful when combined with the previous patch, which allowed the initramfs to be unpacked asynchronously - for that to work, it had to make any usermodehelper call wait for the unpacking to finish before attempting to invoke the userspace helper. By eliminating all such (known-to-be-futile) calls of usermodehelper, the initramfs unpacking and the {device,late}_initcalls can proceed in parallel for much longer. For a relatively slow ppc board I'm working on, the two patches combined lead to 0.2s faster boot - but more importantly, the fact that the initramfs unpacking proceeds completely in the background while devices get probed means I get to handle the gpio watchdog in time without getting reset. [1] __request_module() already has an early -ENOENT return when modprobe_path is the empty string. Signed-off-by: Rasmus Villemoes Acked-by: Jessica Yu Thanks!