lkml.org 
[lkml]   [2005]   [Nov]   [21]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 4/12: eCryptfs] Main module functions
On 11/19/05, Pekka Enberg <penberg@cs.helsinki.fi> wrote:
> On 11/19/05, Phillip Hellewell <phillip@hellewell.homeip.net> wrote:
> > +int ecryptfs_verbosity = 1;
> > +
> > +module_param(ecryptfs_verbosity, int, 0);
> > +MODULE_PARM_DESC(ecryptfs_verbosity,
> > + "Initial verbosity level (0 or 1; defaults to "
> > + "0, which is Quiet)");
> > +
> > +void __ecryptfs_printk(int verb, const char *fmt, ...)
> > +{
> > + va_list args;
> > + va_start(args, fmt);
> > + if ((ecryptfs_verbosity >= verb) && printk_ratelimit())
> > + vprintk(fmt, args);
> > + va_end(args);
> > +}
> > +
>
> [snip]
>
> > +int ecryptfs_interpose(struct dentry *lower_dentry, struct dentry *dentry,
> > + struct super_block *sb, int flag)
> > +{
> > + struct inode *lower_inode;
> > + int err = 0;
> > + struct inode *inode;
> > +
> > + ecryptfs_printk(1, KERN_NOTICE, "Enter; lower_dentry = [%p], "
> > + "lower_dentry->d_name.name = [%s], dentry = "
> > + "[%p], dentry->d_name.name = [%s], sb = [%p]; "
> > + "flag = [%.4x]; lower_dentry->d_count "
> > + "= [%d]; dentry->d_count = [%d]\n", lower_dentry,
> > + lower_dentry->d_name.name, dentry, dentry->d_name.name,
> > + sb, flag, atomic_read(&lower_dentry->d_count),
> > + atomic_read(&dentry->d_count));
>
> Could you use KERN_DEBUG instead and drop ecryptfs_printk()?

I don't follow with what you mean. We use ecryptfs_printk for 2 reasons:
1) Debugging. There is (or atleast at one time was) a large amount of
ecryptfs_printk's scattered around during the development process.
Obviously, much has been removed because we don't need them much
anymore. However, the normal printk functionality doesn't, afaik,
provide function & line number by default. ecryptfs_printk inserts it
automatically for easy of use.

2) Verbosity switch. The intent, eventually atleast, is to be able to
toggle the value of the verbosity level of ecryptfs at run time (this
isn't implemented at the moment)... or atleast that's my plan :P Its
not my project so I have to get "approval" etc. The 0th argument is
the verbosity in which this message applies (1 being just informative,
0 being critical).


>
> > + if (NULL == ECRYPTFS_INODE_TO_LOWER(inode)) {
> > + ECRYPTFS_INODE_TO_LOWER(inode) = igrab(lower_inode);
> > + /* If we are still NULL at this point, igrab failed.
> > + * We are _NOT_ supposed to be failing here */
> > + if (NULL == ECRYPTFS_INODE_TO_LOWER(inode)) {
>
> If you're worried about assignment, why not use the normal idiom:
>
> if (!p) {
> ...
> }

No good reason. Is there a reason to change this? (other than
consistancy, which I agree with, so I'll change it)

>
> > + BUG();
> > + err = -EINVAL;
> > + goto out;
>
> Why do you want to BUG() and then handle the situation?

Simply, we don't. Atleast, not in production code. This was in there
to be a clear catch if we were ever to fall into this case. A remnant
of development I am afraid.

>
> > +kmem_cache_t *ecryptfs_sb_info_cache;
>
> We have struct kmem_cache now so please consider using it instead.

Noted. Changed.

>
> > +/* This provides a means of backing out cache creations out of the kernel
> > + * so that we can elegantly fail should we run out of memory.
> > + */
> > +#define ECRYPTFS_AUTH_TOK_LIST_ITEM_CACHE 0x0001
> > +#define ECRYPTFS_AUTH_TOK_PKT_SET_CACHE 0x0002
> > +#define ECRYPTFS_AUTH_TOK_REQUEST_CACHE 0x0004
> > +#define ECRYPTFS_AUTH_TOK_REQUEST_BLOB_CACHE 0x0008
> > +#define ECRYPTFS_FILE_INFO_CACHE 0x0010
> > +#define ECRYPTFS_DENTRY_INFO_CACHE 0x0020
> > +#define ECRYPTFS_INODE_INFO_CACHE 0x0040
> > +#define ECRYPTFS_SB_INFO_CACHE 0x0080
> > +#define ECRYPTFS_HEADER_CACHE_0 0x0100
> > +#define ECRYPTFS_HEADER_CACHE_1 0x0200
> > +#define ECRYPTFS_HEADER_CACHE_2 0x0400
> > +#define ECRYPTFS_LOWER_PAGE_CACHE 0x0800
> > +#define ECRYPTFS_CACHE_CREATION_SUCCESS 0x0FF1
>
> A better way would be to either (1) make ecryptfs_init_kmem_caches()
> clean up after itself on error using gotos or (2) keep caches NULL
> before initialization and check for that in
> ecryptfs_free_kmem_caches().

Only problem with cleanp gotos is it would make the function
needlessly larger than it already is. I do like the idea of the NULL
thing better, will have a go at it.

>
> > +
> > +static short ecryptfs_allocated_caches;
> > +
> > +/**
> > + * @return Zero on success; non-zero otherwise
> > + *
> > + * Sets ecryptfs_allocated_caches with flags so that we can
> > + * free created caches should we run out of memory during
> > + * creation period.
> > + *
> > + * The overhead for doing this is offset by the fact that we
> > + * only do this once, and that should there be insufficient
> > + * memory, then we can elegantly fail, and not leave extra
> > + * caches around, or worse, panic the kernel trying to free
> > + * something that's not there.
> > + */
> > +static int ecryptfs_init_kmem_caches(void)
> > +{
> > + int rc = 0;
> > +
> > + ecryptfs_auth_tok_list_item_cache =
> > + kmem_cache_create("ecryptfs_auth_tok_list_item",
> > + sizeof(struct ecryptfs_auth_tok_list_item),
> > + 0, SLAB_HWCACHE_ALIGN, NULL, NULL);
> > + if (ecryptfs_auth_tok_list_item_cache)
> > + rc |= ECRYPTFS_AUTH_TOK_LIST_ITEM_CACHE;
> > + else
> > + ecryptfs_printk(0, KERN_WARNING, "ecryptfs_auth_tok_list_item "
> > + "kmem_cache_create failed\n");
> > +
> > + ecryptfs_file_info_cache =
> > + kmem_cache_create("ecryptfs_file_cache",
> > + sizeof(struct ecryptfs_file_info),
> > + 0, SLAB_HWCACHE_ALIGN, NULL, NULL);
> > + if (ecryptfs_file_info_cache)
> > + rc |= ECRYPTFS_FILE_INFO_CACHE;
> > + else
> > + ecryptfs_printk(0, KERN_WARNING, "ecryptfs_file_cache "
> > + "kmem_cache_create failed\n");
> > +
> > + ecryptfs_dentry_info_cache =
> > + kmem_cache_create("ecryptfs_dentry_cache",
> > + sizeof(struct ecryptfs_dentry_info),
> > + 0, SLAB_HWCACHE_ALIGN, NULL, NULL);
> > + if (ecryptfs_dentry_info_cache)
> > + rc |= ECRYPTFS_DENTRY_INFO_CACHE;
> > + else
> > + ecryptfs_printk(0, KERN_WARNING, "ecryptfs_dentry_cache "
> > + "kmem_cache_create failed\n");
> > +
> > + ecryptfs_inode_info_cache =
> > + kmem_cache_create("ecryptfs_inode_cache",
> > + sizeof(struct ecryptfs_inode_info), 0,
> > + SLAB_HWCACHE_ALIGN, inode_info_init_once, NULL);
> > + if (ecryptfs_inode_info_cache)
> > + rc |= ECRYPTFS_INODE_INFO_CACHE;
> > + else
> > + ecryptfs_printk(0, KERN_WARNING, "ecryptfs_inode_cache "
> > + "kmem_cache_create failed\n");
> > +
> > + ecryptfs_sb_info_cache =
> > + kmem_cache_create("ecryptfs_sb_cache",
> > + sizeof(struct ecryptfs_sb_info),
> > + 0, SLAB_HWCACHE_ALIGN, NULL, NULL);
> > + if (ecryptfs_sb_info_cache)
> > + rc |= ECRYPTFS_SB_INFO_CACHE;
> > + else
> > + ecryptfs_printk(0, KERN_WARNING, "ecryptfs_sb_cache "
> > + "kmem_cache_create failed\n");
> > +
> > + ecryptfs_header_cache_0 =
> > + kmem_cache_create("ecryptfs_headers_0", PAGE_CACHE_SIZE,
> > + 0, SLAB_HWCACHE_ALIGN, NULL, NULL);
> > + if (ecryptfs_header_cache_0)
> > + rc |= ECRYPTFS_HEADER_CACHE_0;
> > + else
> > + ecryptfs_printk(0, KERN_WARNING, "ecryptfs_headers_0 "
> > + "kmem_cache_create failed\n");
> > +
> > + ecryptfs_header_cache_1 =
> > + kmem_cache_create("ecryptfs_headers_1", PAGE_CACHE_SIZE,
> > + 0, SLAB_HWCACHE_ALIGN, NULL, NULL);
> > + if (ecryptfs_header_cache_1)
> > + rc |= ECRYPTFS_HEADER_CACHE_1;
> > + else
> > + ecryptfs_printk(0, KERN_WARNING, "ecryptfs_headers_1 "
> > + "kmem_cache_create failed\n");
> > +
> > + ecryptfs_header_cache_2 =
> > + kmem_cache_create("ecryptfs_headers_2", PAGE_CACHE_SIZE,
> > + 0, SLAB_HWCACHE_ALIGN, NULL, NULL);
> > + if (ecryptfs_header_cache_2)
> > + rc |= ECRYPTFS_HEADER_CACHE_2;
> > + else
> > + ecryptfs_printk(0, KERN_WARNING, "ecryptfs_headers_2 "
> > + "kmem_cache_create failed\n");
> > +
> > + ecryptfs_lower_page_cache =
> > + kmem_cache_create("ecryptfs_lower_page_cache", PAGE_CACHE_SIZE,
> > + 0, SLAB_HWCACHE_ALIGN, NULL, NULL);
> > + if (ecryptfs_lower_page_cache)
> > + rc |= ECRYPTFS_LOWER_PAGE_CACHE;
> > + else
> > + ecryptfs_printk(0, KERN_WARNING, "ecryptfs_lower_page_cache "
> > + "kmem_cache_create failed\n");
> > +
> > + ecryptfs_allocated_caches = rc;
> > + rc = ECRYPTFS_CACHE_CREATION_SUCCESS ^ rc;
> > + return rc;
> > +}
> > +
> > +/**
> > + * @return Zero on success; non-zero otherwise
> > + */
> > +static int ecryptfs_free_kmem_caches(void)
> > +{
> > + int rc = 0;
> > + int err;
> > +
> > + if (ecryptfs_allocated_caches & ECRYPTFS_AUTH_TOK_LIST_ITEM_CACHE) {
> > + rc = kmem_cache_destroy(ecryptfs_auth_tok_list_item_cache);
> > + if (rc)
> > + ecryptfs_printk(0, KERN_WARNING,
> > + "Not all ecryptfs_auth_tok_"
> > + "list_item_cache structures were "
> > + "freed\n");
> > + }
> > + if (ecryptfs_allocated_caches & ECRYPTFS_FILE_INFO_CACHE) {
> > + err = kmem_cache_destroy(ecryptfs_file_info_cache);
> > + if (err)
> > + ecryptfs_printk(0, KERN_WARNING,
> > + "Not all ecryptfs_file_info_"
> > + "cache regions were freed\n");
> > + rc |= err;
> > + }
> > + if (ecryptfs_allocated_caches & ECRYPTFS_DENTRY_INFO_CACHE) {
> > + err = kmem_cache_destroy(ecryptfs_dentry_info_cache);
> > + if (err)
> > + ecryptfs_printk(0, KERN_WARNING,
> > + "Not all ecryptfs_dentry_info_"
> > + "cache regions were freed\n");
> > + rc |= err;
> > + }
> > + if (ecryptfs_allocated_caches & ECRYPTFS_INODE_INFO_CACHE) {
> > + err = kmem_cache_destroy(ecryptfs_inode_info_cache);
> > + if (err)
> > + ecryptfs_printk(0, KERN_WARNING,
> > + "Not all ecryptfs_inode_info_"
> > + "cache regions were freed\n");
> > + rc |= err;
> > + }
> > + if (ecryptfs_allocated_caches & ECRYPTFS_SB_INFO_CACHE) {
> > + err = kmem_cache_destroy(ecryptfs_sb_info_cache);
> > + if (err)
> > + ecryptfs_printk(0, KERN_WARNING,
> > + "Not all ecryptfs_sb_info_"
> > + "cache regions were freed\n");
> > + rc |= err;
> > + }
> > + if (ecryptfs_allocated_caches & ECRYPTFS_HEADER_CACHE_0) {
> > + err = kmem_cache_destroy(ecryptfs_header_cache_0);
> > + if (err)
> > + ecryptfs_printk(0, KERN_WARNING, "Not all ecryptfs_"
> > + "header_cache_0 regions were freed\n");
> > + rc |= err;
> > + }
> > + if (ecryptfs_allocated_caches & ECRYPTFS_HEADER_CACHE_1) {
> > + err = kmem_cache_destroy(ecryptfs_header_cache_1);
> > + if (err)
> > + ecryptfs_printk(0, KERN_WARNING, "Not all ecryptfs_"
> > + "header_cache_1 regions were freed\n");
> > + rc |= err;
> > + }
> > + if (ecryptfs_allocated_caches & ECRYPTFS_HEADER_CACHE_2) {
> > + err = kmem_cache_destroy(ecryptfs_header_cache_2);
> > + if (err)
> > + ecryptfs_printk(0, KERN_WARNING, "Not all ecryptfs_"
> > + "header_cache_2 regions were freed\n");
> > + rc |= err;
> > + }
> > + if (ecryptfs_allocated_caches & ECRYPTFS_LOWER_PAGE_CACHE) {
> > + err = kmem_cache_destroy(ecryptfs_lower_page_cache);
> > + if (err)
> > + ecryptfs_printk(0, KERN_WARNING, "Not all ecryptfs_"
> > + "lower_page_cache regions were "
> > + "freed\n");
> > + rc |= err;
> > + }
> > + return rc;
> > +}
> > +
> > +static int __init init_ecryptfs_fs(void)
> > +{
> > + int rc;
> > +
> > + rc = ecryptfs_init_kmem_caches();
> > + if (rc) {
> > + ecryptfs_printk(0, KERN_EMERG, "Failure occured while "
> > + "attempting to create caches [CREATED: %x]."
> > + "Now freeing caches.\n",
> > + ecryptfs_allocated_caches);
> > + ecryptfs_free_kmem_caches();
> > + return -ENOMEM;
> > + }
> > + ecryptfs_printk(1, KERN_NOTICE, "Registering eCryptfs\n");
> > + return register_filesystem(&ecryptfs_fs_type);
>
> register_filesystem() can fail in which case youre leaking all the
> slab caches here.

I wasn't aware it _could_ fail, thanks for that.

>
> Pekka
> -
> To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>


--
Michael C. Thompson <mcthomps@us.ibm.com>
Software-Engineer, IBM LTC Security
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-11-21 17:13    [W:0.582 / U:0.032 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site