Linux CIFS filesystem development
 help / color / mirror / Atom feed
From: "Atte Heikkilä" <atteh.mailbox@gmail.com>
To: tom@talpey.com
Cc: atteh.mailbox@gmail.com, linkinjeon@kernel.org,
	linux-cifs@vger.kernel.org
Subject: Re: [PATCH v3] ksmbd: validate share name from share config response
Date: Sun,  2 Oct 2022 05:02:26 +0300	[thread overview]
Message-ID: <20221002020226.3917-1-atteh.mailbox@gmail.com> (raw)
In-Reply-To: <f7710758-996a-6bf5-4d7c-5c295577c965@talpey.com>

On Sat, 1 Oct 2022 20:38:06 -0400, Tom Talpey wrote:
> On 10/1/2022 8:25 PM, Atte Heikkilä wrote:
>> On Sat, 1 Oct 2022 19:20:52 -0400, Tom Talpey wrote:
>>> On 10/1/2022 6:53 PM, Atte Heikkilä wrote:
>>>> Share config response may contain the share name without casefolding as
>>>> it is known to the user space daemon. When it is present, casefold and
>>>> compare it to the share name the share config request was made with. If
>>>> they differ, we have a share config which is incompatible with the way
>>>> share config caching is done. This is the case when CONFIG_UNICODE is
>>>> not set, the share name contains non-ASCII characters, and those non-
>>>> ASCII characters do not match those in the share name known to user
>>>> space. In other words, when CONFIG_UNICODE is not set, UTF-8 share
>>>> names now work but are only case-insensitive in the ASCII range.
>>>>
>>>> Signed-off-by: Atte Heikkilä <atteh.mailbox@gmail.com>
>>>> ---
>>>>   v3:
>>>>     - removed initial strcmp() check since it could only save a call to
>>>>       ksmbd_casefold_sharename() for matching ASCII-only share names
>>>>
>>>>   v2:
>>>>     - no changes were made
>>>>
>>>>   fs/ksmbd/ksmbd_netlink.h     |  3 ++-
>>>>   fs/ksmbd/mgmt/share_config.c | 20 +++++++++++++++++---
>>>>   fs/ksmbd/mgmt/share_config.h |  4 +++-
>>>>   fs/ksmbd/mgmt/tree_connect.c |  4 ++--
>>>>   fs/ksmbd/misc.c              |  4 ++--
>>>>   fs/ksmbd/misc.h              |  1 +
>>>>   6 files changed, 27 insertions(+), 9 deletions(-)
>>>>
>>>> diff --git a/fs/ksmbd/ksmbd_netlink.h b/fs/ksmbd/ksmbd_netlink.h
>>>> index e0cbcfa98c7e..ff07c67f4565 100644
>>>> --- a/fs/ksmbd/ksmbd_netlink.h
>>>> +++ b/fs/ksmbd/ksmbd_netlink.h
>>>> @@ -163,7 +163,8 @@ struct ksmbd_share_config_response {
>>>>       __u16    force_directory_mode;
>>>>       __u16    force_uid;
>>>>       __u16    force_gid;
>>>> -    __u32    reserved[128];        /* Reserved room */
>>>> +    __s8    share_name[KSMBD_REQ_MAX_SHARE_NAME];
>>>> +    __u32    reserved[112];        /* Reserved room */
>>>
>>> Because this is a protocol/abi definition, it may be good practice
>>> to code as
>>>     __u32    reserved[128 - KSMBD_REQ_MAX_SHARE_NAME];
>>> just in case the sharename constant changes in future.
>>>
>> 
>> You are right. That would probably be better.
>> 
>>>>       __u32    veto_list_sz;
>>>>       __s8    ____payload[];
>>>>   };
>>>> diff --git a/fs/ksmbd/mgmt/share_config.c b/fs/ksmbd/mgmt/share_config.c
>>>> index 5d039704c23c..c62e87b67247 100644
>>>> --- a/fs/ksmbd/mgmt/share_config.c
>>>> +++ b/fs/ksmbd/mgmt/share_config.c
>>>> @@ -16,6 +16,7 @@
>>>>   #include "user_config.h"
>>>>   #include "user_session.h"
>>>>   #include "../transport_ipc.h"
>>>> +#include "../misc.h"
>>>>   #define SHARE_HASH_BITS        3
>>>>   static DEFINE_HASHTABLE(shares_table, SHARE_HASH_BITS);
>>>> @@ -119,7 +120,8 @@ static int parse_veto_list(struct 
>>>> ksmbd_share_config *share,
>>>>       return 0;
>>>>   }
>>>> -static struct ksmbd_share_config *share_config_request(const char 
>>>> *name)
>>>> +static struct ksmbd_share_config *share_config_request(struct 
>>>> unicode_map *um,
>>>> +                               const char *name)
>>>>   {
>>>>       struct ksmbd_share_config_response *resp;
>>>>       struct ksmbd_share_config *share = NULL;
>>>> @@ -133,6 +135,17 @@ static struct ksmbd_share_config 
>>>> *share_config_request(const char *name)
>>>>       if (resp->flags == KSMBD_SHARE_FLAG_INVALID)
>>>>           goto out;
>>>> +    if (*resp->share_name) {
>>>> +        char *cf_resp_name;
>>>> +        bool equal;
>>>> +
>>>> +        cf_resp_name = ksmbd_casefold_sharename(um, resp->share_name);
>>>> +        equal = cf_resp_name && !strcmp(cf_resp_name, name);
>>>
>>> This is a little bit hard to parse. So, if the casefolding returns
>>> a non-null string, and the result does not equal the original name?
>>> Maybe because it's evaluating a pointer as a bool, and using "!" to
>>> evaluate the strcpy result is the confusing part, to me.
>>>
>>>     equal = cf_resp_name != NULL && strcmp(cf_resp_name, name) 0= 0;
>>>
>> 
>> It should be `!IS_ERR(cf_resp_name) && !strcmp(cf_resp_name, name)'.
> 
> Aha - ok that's better (but I still have to pause after seeing "!strcmp")
> 

It looks like `!strcmp()' sees use over `strcmp() == 0' in ksmbd.

>> The share name in the share config response is the actual share name
>> which is known to user space. It is then casefolded according to the
>> kernel's capabilities, which is either utf8_casefold() or tolower().
>> If it differs from the casefolded form of the share name used by the
>> client, then the client-used share name is incompatible with the way
>> the kernel does share config caching.
>> 
>>> Also, strcmp() gets a lot of attention in the kernel these days. Is
>>> this guaranteed safe?
>> 
>> Both are always null-terminated if that is what you are asking.
> 
> Ok, but are they guaranteed to be ascii? Isn't strcmp doing an octet
> based comparison?
> 

They are both casefolded UTF-8 or the lowercase ASCII subset of it.
utf8_casefold() returns `nlen' only when the last written byte to the
output `dest' buffer is the null terminator. Otherwise, when it returns
-EINVAL, or when CONFIG_UNICODE is not set, ksmbd_casefold_sharename()
falls back to tolower() guarded by isascii() so as to not touch the
UTF-8-incompatible ISO8859-1 range. As I understand it, octet-wise
comparison terminated by null byte is just fine here.

> Ned Pyle has a hilarious demo where he sets a servername to the Bacon
> emoji, and a share name to Eggs-emoji, or something. I'm not sure if
> he creates a Fork-emoji file, but it's something like that. He then
> proceeds to show how it all works with Windows. I just don't want to
> see us running over the end of crazy unterminated or Unicode expansion
> strings.
> 
>> Tom, thank you very much for taking a look. I would not have noticed my
>> careless mistake otherwise.
>> 
>>>
>>>> +        kfree(cf_resp_name);
>>>> +        if (!equal)
>>>> +            goto out;
>>>> +    }
>>>> +
>>>>       share = kzalloc(sizeof(struct ksmbd_share_config), GFP_KERNEL);
>>>>       if (!share)
>>>>           goto out;
>>>> @@ -190,7 +203,8 @@ static struct ksmbd_share_config 
>>>> *share_config_request(const char *name)
>>>>       return share;
>>>>   }
>>>> -struct ksmbd_share_config *ksmbd_share_config_get(const char *name)
>>>> +struct ksmbd_share_config *ksmbd_share_config_get(struct unicode_map 
>>>> *um,
>>>> +                          const char *name)
>>>>   {
>>>>       struct ksmbd_share_config *share;
>>>> @@ -202,7 +216,7 @@ struct ksmbd_share_config 
>>>> *ksmbd_share_config_get(const char *name)
>>>>       if (share)
>>>>           return share;
>>>> -    return share_config_request(name);
>>>> +    return share_config_request(um, name);
>>>>   }
>>>>   bool ksmbd_share_veto_filename(struct ksmbd_share_config *share,
>>>> diff --git a/fs/ksmbd/mgmt/share_config.h b/fs/ksmbd/mgmt/share_config.h
>>>> index 7f7e89ecfe61..3fd338293942 100644
>>>> --- a/fs/ksmbd/mgmt/share_config.h
>>>> +++ b/fs/ksmbd/mgmt/share_config.h
>>>> @@ -9,6 +9,7 @@
>>>>   #include <linux/workqueue.h>
>>>>   #include <linux/hashtable.h>
>>>>   #include <linux/path.h>
>>>> +#include <linux/unicode.h>
>>>>   struct ksmbd_share_config {
>>>>       char            *name;
>>>> @@ -74,7 +75,8 @@ static inline void ksmbd_share_config_put(struct 
>>>> ksmbd_share_config *share)
>>>>       __ksmbd_share_config_put(share);
>>>>   }
>>>> -struct ksmbd_share_config *ksmbd_share_config_get(const char *name);
>>>> +struct ksmbd_share_config *ksmbd_share_config_get(struct unicode_map 
>>>> *um,
>>>> +                          const char *name);
>>>>   bool ksmbd_share_veto_filename(struct ksmbd_share_config *share,
>>>>                      const char *filename);
>>>>   #endif /* __SHARE_CONFIG_MANAGEMENT_H__ */
>>>> diff --git a/fs/ksmbd/mgmt/tree_connect.c b/fs/ksmbd/mgmt/tree_connect.c
>>>> index 867c0286b901..8ce17b3fb8da 100644
>>>> --- a/fs/ksmbd/mgmt/tree_connect.c
>>>> +++ b/fs/ksmbd/mgmt/tree_connect.c
>>>> @@ -26,7 +26,7 @@ ksmbd_tree_conn_connect(struct ksmbd_conn *conn, 
>>>> struct ksmbd_session *sess,
>>>>       struct sockaddr *peer_addr;
>>>>       int ret;
>>>> -    sc = ksmbd_share_config_get(share_name);
>>>> +    sc = ksmbd_share_config_get(conn->um, share_name);
>>>>       if (!sc)
>>>
>>> Another pointer boolean evaluation, but I guess this one is existing.
>>>
>>>>           return status;
>>>> @@ -61,7 +61,7 @@ ksmbd_tree_conn_connect(struct ksmbd_conn *conn, 
>>>> struct ksmbd_session *sess,
>>>>           struct ksmbd_share_config *new_sc;
>>>>           ksmbd_share_config_del(sc);
>>>> -        new_sc = ksmbd_share_config_get(share_name);
>>>> +        new_sc = ksmbd_share_config_get(conn->um, share_name);
>>>>           if (!new_sc) {
>>>
>>> Ditto
>>>>               pr_err("Failed to update stale share config\n");
>>>>               status.ret = -ESTALE;
>>>> diff --git a/fs/ksmbd/misc.c b/fs/ksmbd/misc.c
>>>> index 28459b1efaa8..9e8afaa686e3 100644
>>>> --- a/fs/ksmbd/misc.c
>>>> +++ b/fs/ksmbd/misc.c
>>>> @@ -227,7 +227,7 @@ void ksmbd_conv_path_to_windows(char *path)
>>>>       strreplace(path, '/', '\\');
>>>>   }
>>>> -static char *casefold_sharename(struct unicode_map *um, const char 
>>>> *name)
>>>> +char *ksmbd_casefold_sharename(struct unicode_map *um, const char 
>>>> *name)
>>>>   {
>>>>       char *cf_name;
>>>>       int cf_len;
>>>> @@ -273,7 +273,7 @@ char *ksmbd_extract_sharename(struct unicode_map 
>>>> *um, const char *treename)
>>>>           name = (pos + 1);
>>>>       /* caller has to free the memory */
>>>> -    return casefold_sharename(um, name);
>>>> +    return ksmbd_casefold_sharename(um, name);
>>>>   }
>>>>   /**
>>>> diff --git a/fs/ksmbd/misc.h b/fs/ksmbd/misc.h
>>>> index cc72f4e6baf2..1facfcd21200 100644
>>>> --- a/fs/ksmbd/misc.h
>>>> +++ b/fs/ksmbd/misc.h
>>>> @@ -20,6 +20,7 @@ int get_nlink(struct kstat *st);
>>>>   void ksmbd_conv_path_to_unix(char *path);
>>>>   void ksmbd_strip_last_slash(char *path);
>>>>   void ksmbd_conv_path_to_windows(char *path);
>>>> +char *ksmbd_casefold_sharename(struct unicode_map *um, const char 
>>>> *name);
>>>>   char *ksmbd_extract_sharename(struct unicode_map *um, const char 
>>>> *treename);
>>>>   char *convert_to_unix_name(struct ksmbd_share_config *share, const 
>>>> char *name);
>>>
>> 
> 

      reply	other threads:[~2022-10-02  2:02 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-10-01 22:53 [PATCH v3] ksmbd: validate share name from share config response Atte Heikkilä
2022-10-01 23:20 ` Tom Talpey
2022-10-02  0:25   ` Atte Heikkilä
2022-10-02  0:38     ` Tom Talpey
2022-10-02  2:02       ` Atte Heikkilä [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20221002020226.3917-1-atteh.mailbox@gmail.com \
    --to=atteh.mailbox@gmail.com \
    --cc=linkinjeon@kernel.org \
    --cc=linux-cifs@vger.kernel.org \
    --cc=tom@talpey.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox