[PATCH] fs/crypto: make crypto.c explicitly non-modular

[PATCH] fs/crypto: make crypto.c explicitly non-modular

[Paul Gortmaker]

As of commit 47134e6084f70fdf4381af75d4569cec6c7ebd50 ("fs crypto:
add Makefile and Kconfig") the compile of fs/crypto/crypto.c tripped
my local audit for non-modules using modular infrastructure vs. their
built in counterparts.

The Kconfig currently controlling compilation of this code is:

config FS_ENCRYPTION
        bool "FS Encryption (Per-file encryption)"

combined with:

obj-$(CONFIG_FS_ENCRYPTION)     += crypto.o policy.o keyinfo.o

...meaning that it currently is not being built as a module by anyone.

Lets remove the modular code that is essentially orphaned, so that
when reading the driver there is no doubt it is builtin-only.

Since module_init translates to device_initcall in the non-modular
case, the use of fs_initcall (which seems appropriate for fs code)
means the init comes slightly earlier.  However boot testing an
x86-64 defconfig didn't show this to be causing any issues.

We replace module.h with moduleparam.h since the file does declare
some module parameters, and leaving them as such is currently the
easiest way to remain compatible with existing boot arg use cases.

Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
---
 fs/crypto/crypto.c | 22 ++++++----------------
 1 file changed, 6 insertions(+), 16 deletions(-)

diff --git a/fs/crypto/crypto.c b/fs/crypto/crypto.c
index ed17895cda11..abca51e75572 100644
--- a/fs/crypto/crypto.c
+++ b/fs/crypto/crypto.c
@@ -23,7 +23,7 @@
 #include <linux/ecryptfs.h>
 #include <linux/pagemap.h>
 #include <linux/mempool.h>
-#include <linux/module.h>
+#include <linux/moduleparam.h>
 #include <linux/scatterlist.h>
 #include <linux/ratelimit.h>
 #include <linux/bio.h>
@@ -33,6 +33,10 @@
 static unsigned int num_prealloc_crypto_pages = 32;
 static unsigned int num_prealloc_crypto_ctxs = 128;
 
+/*
+ * Not really modular but using module_param is the easiest way to
+ * remain consistent with existing use cases for now.
+ */
 module_param(num_prealloc_crypto_pages, uint, 0444);
 MODULE_PARM_DESC(num_prealloc_crypto_pages,
 		"Number of crypto pages to preallocate");
@@ -537,18 +541,4 @@ fail_free_queue:
 fail:
 	return -ENOMEM;
 }
-module_init(fscrypt_init)
-
-/**
- * fscrypt_exit() - Shutdown the fs encryption system
- */
-void __exit fscrypt_exit(void)
-{
-	fscrypt_destroy();
-
-	if (fscrypt_read_workqueue)
-		destroy_workqueue(fscrypt_read_workqueue);
-	kmem_cache_destroy(fscrypt_ctx_cachep);
-	kmem_cache_destroy(fscrypt_info_cachep);
-}
-module_exit(fscrypt_exit);
+fs_initcall(fscrypt_init)
-- 
2.6.1

Re: [PATCH] fs/crypto: make crypto.c explicitly non-modular

[Arnd Bergmann]

On Saturday 27 February 2016 15:20:49 Paul Gortmaker wrote:
> As of commit 47134e6084f70fdf4381af75d4569cec6c7ebd50 ("fs crypto:
> add Makefile and Kconfig") the compile of fs/crypto/crypto.c tripped
> my local audit for non-modules using modular infrastructure vs. their
> built in counterparts.
> 
> The Kconfig currently controlling compilation of this code is:
> 
> config FS_ENCRYPTION
>         bool "FS Encryption (Per-file encryption)"
> 
> combined with:
> 
> obj-$(CONFIG_FS_ENCRYPTION)     += crypto.o policy.o keyinfo.o
> 
> ...meaning that it currently is not being built as a module by anyone.
> 
> Lets remove the modular code that is essentially orphaned, so that
> when reading the driver there is no doubt it is builtin-only.
> 
> Since module_init translates to device_initcall in the non-modular
> case, the use of fs_initcall (which seems appropriate for fs code)
> means the init comes slightly earlier.  However boot testing an
> x86-64 defconfig didn't show this to be causing any issues.
> 
> We replace module.h with moduleparam.h since the file does declare
> some module parameters, and leaving them as such is currently the
> easiest way to remain compatible with existing boot arg use cases.
> 

So why not make the option a 'tristate' instead? It looks like
that was intended here, and we should always try to make all
code loadable as modules if possible.

	Arnd

Re: [PATCH] fs/crypto: make crypto.c explicitly non-modular

[Paul Gortmaker]

[Re: [PATCH] fs/crypto: make crypto.c explicitly non-modular] On 29/02/2016 (Mon 10:56) Arnd Bergmann wrote:

> On Saturday 27 February 2016 15:20:49 Paul Gortmaker wrote:
> > As of commit 47134e6084f70fdf4381af75d4569cec6c7ebd50 ("fs crypto:
> > add Makefile and Kconfig") the compile of fs/crypto/crypto.c tripped
> > my local audit for non-modules using modular infrastructure vs. their
> > built in counterparts.
> > 
> > The Kconfig currently controlling compilation of this code is:
> > 
> > config FS_ENCRYPTION
> >         bool "FS Encryption (Per-file encryption)"
> > 
> > combined with:
> > 
> > obj-$(CONFIG_FS_ENCRYPTION)     += crypto.o policy.o keyinfo.o
> > 
> > ...meaning that it currently is not being built as a module by anyone.
> > 
> > Lets remove the modular code that is essentially orphaned, so that
> > when reading the driver there is no doubt it is builtin-only.
> > 
> > Since module_init translates to device_initcall in the non-modular
> > case, the use of fs_initcall (which seems appropriate for fs code)
> > means the init comes slightly earlier.  However boot testing an
> > x86-64 defconfig didn't show this to be causing any issues.
> > 
> > We replace module.h with moduleparam.h since the file does declare
> > some module parameters, and leaving them as such is currently the
> > easiest way to remain compatible with existing boot arg use cases.
> > 
> 
> So why not make the option a 'tristate' instead? It looks like
> that was intended here, and we should always try to make all
> code loadable as modules if possible.

It wasn't that obvious to me, and at the risk of repeating myself, I'll
always default to making the code consistent with existing functionality
vs. making assumptions about what was intended and implicitly adding new
functionality that may be invalid for reasons I'd never spot.

Of course authors, and maintainers are welcome to chime in and steer it
one way or another as per the PCI patches.

Paul.
--

> 
> 	Arnd

Re: [PATCH] fs/crypto: make crypto.c explicitly non-modular

[Jaegeuk Kim]

Hi Paul,

On Mon, Feb 29, 2016 at 09:31:59AM -0500, Paul Gortmaker wrote:
> [Re: [PATCH] fs/crypto: make crypto.c explicitly non-modular] On 29/02/2016 (Mon 10:56) Arnd Bergmann wrote:
> 
> > On Saturday 27 February 2016 15:20:49 Paul Gortmaker wrote:
> > > As of commit 47134e6084f70fdf4381af75d4569cec6c7ebd50 ("fs crypto:
> > > add Makefile and Kconfig") the compile of fs/crypto/crypto.c tripped
> > > my local audit for non-modules using modular infrastructure vs. their
> > > built in counterparts.
> > > 
> > > The Kconfig currently controlling compilation of this code is:
> > > 
> > > config FS_ENCRYPTION
> > >         bool "FS Encryption (Per-file encryption)"
> > > 
> > > combined with:
> > > 
> > > obj-$(CONFIG_FS_ENCRYPTION)     += crypto.o policy.o keyinfo.o
> > > 
> > > ...meaning that it currently is not being built as a module by anyone.
> > > 
> > > Lets remove the modular code that is essentially orphaned, so that
> > > when reading the driver there is no doubt it is builtin-only.
> > > 
> > > Since module_init translates to device_initcall in the non-modular
> > > case, the use of fs_initcall (which seems appropriate for fs code)
> > > means the init comes slightly earlier.  However boot testing an
> > > x86-64 defconfig didn't show this to be causing any issues.
> > > 
> > > We replace module.h with moduleparam.h since the file does declare
> > > some module parameters, and leaving them as such is currently the
> > > easiest way to remain compatible with existing boot arg use cases.
> > > 
> > 
> > So why not make the option a 'tristate' instead? It looks like
> > that was intended here, and we should always try to make all
> > code loadable as modules if possible.
> 
> It wasn't that obvious to me, and at the risk of repeating myself, I'll
> always default to making the code consistent with existing functionality
> vs. making assumptions about what was intended and implicitly adding new
> functionality that may be invalid for reasons I'd never spot.
> 
> Of course authors, and maintainers are welcome to chime in and steer it
> one way or another as per the PCI patches.

Meanwhile, I've modified and tested fscrypt as a module, as Arnd commented.
I'll submit v3 to resolve this inconsistency.
Thank you for the point that I missed.
> 
> Paul.
> --
> 
> > 
> > 	Arnd