cifs: clean up various nits in unicode routines (try #2)
authorJeff Layton <jlayton@redhat.com>
Tue, 5 Apr 2011 19:02:37 +0000 (15:02 -0400)
committerSteve French <sfrench@us.ibm.com>
Tue, 12 Apr 2011 00:57:12 +0000 (00:57 +0000)
Minor revision to the original patch. Don't abuse the __le16 variable
on the stack by casting it to wchar_t and handing it off to char2uni.
Declare an actual wchar_t on the stack instead. This fixes a valid
sparse warning.

Fix the spelling of UNI_ASTERISK. Eliminate the unneeded len_remaining
variable in cifsConvertToUCS.

Also, as David Howells points out. We were better off making
cifsConvertToUCS *not* use put_unaligned_le16 since it means that we
can't optimize the mapped characters at compile time. Switch them
instead to use cpu_to_le16, and simply use put_unaligned to set them
in the string.

Reported-and-acked-by: David Howells <dhowells@redhat.com>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Steve French <sfrench@us.ibm.com>
fs/cifs/cifs_unicode.c
fs/cifs/cifs_unicode.h

index fc0fd4fde306b989c1593ef11ce2611fd49dca54..23d43cde4306042407dcd3df76d78ab81b793db7 100644 (file)
@@ -90,7 +90,7 @@ cifs_mapchar(char *target, const __u16 src_char, const struct nls_table *cp,
        case UNI_COLON:
                *target = ':';
                break;
-       case UNI_ASTERIK:
+       case UNI_ASTERISK:
                *target = '*';
                break;
        case UNI_QUESTION:
@@ -264,40 +264,40 @@ cifs_strndup_from_ucs(const char *src, const int maxlen, const bool is_unicode,
  * names are little endian 16 bit Unicode on the wire
  */
 int
-cifsConvertToUCS(__le16 *target, const char *source, int maxlen,
+cifsConvertToUCS(__le16 *target, const char *source, int srclen,
                 const struct nls_table *cp, int mapChars)
 {
        int i, j, charlen;
-       int len_remaining = maxlen;
        char src_char;
-       __u16 temp;
+       __le16 dst_char;
+       wchar_t tmp;
 
        if (!mapChars)
                return cifs_strtoUCS(target, source, PATH_MAX, cp);
 
-       for (i = 0, j = 0; i < maxlen; j++) {
+       for (i = 0, j = 0; i < srclen; j++) {
                src_char = source[i];
                switch (src_char) {
                case 0:
-                       put_unaligned_le16(0, &target[j]);
+                       put_unaligned(0, &target[j]);
                        goto ctoUCS_out;
                case ':':
-                       temp = UNI_COLON;
+                       dst_char = cpu_to_le16(UNI_COLON);
                        break;
                case '*':
-                       temp = UNI_ASTERIK;
+                       dst_char = cpu_to_le16(UNI_ASTERISK);
                        break;
                case '?':
-                       temp = UNI_QUESTION;
+                       dst_char = cpu_to_le16(UNI_QUESTION);
                        break;
                case '<':
-                       temp = UNI_LESSTHAN;
+                       dst_char = cpu_to_le16(UNI_LESSTHAN);
                        break;
                case '>':
-                       temp = UNI_GRTRTHAN;
+                       dst_char = cpu_to_le16(UNI_GRTRTHAN);
                        break;
                case '|':
-                       temp = UNI_PIPE;
+                       dst_char = cpu_to_le16(UNI_PIPE);
                        break;
                /*
                 * FIXME: We can not handle remapping backslash (UNI_SLASH)
@@ -305,17 +305,17 @@ cifsConvertToUCS(__le16 *target, const char *source, int maxlen,
                 * as they use backslash as separator.
                 */
                default:
-                       charlen = cp->char2uni(source+i, len_remaining,
-                                               &temp);
+                       charlen = cp->char2uni(source + i, srclen - i, &tmp);
+                       dst_char = cpu_to_le16(tmp);
+
                        /*
                         * if no match, use question mark, which at least in
                         * some cases serves as wild card
                         */
                        if (charlen < 1) {
-                               temp = 0x003f;
+                               dst_char = cpu_to_le16(0x003f);
                                charlen = 1;
                        }
-                       len_remaining -= charlen;
                        /*
                         * character may take more than one byte in the source
                         * string, but will take exactly two bytes in the
@@ -324,9 +324,8 @@ cifsConvertToUCS(__le16 *target, const char *source, int maxlen,
                        i += charlen;
                        continue;
                }
-               put_unaligned_le16(temp, &target[j]);
+               put_unaligned(dst_char, &target[j]);
                i++; /* move to next char in source string */
-               len_remaining--;
        }
 
 ctoUCS_out:
index 7fe6b52df5076a2eb14a7ec563403795e2e4c041..644dd882a5604c5ed6f900aad2a47478bc31b49a 100644 (file)
@@ -44,7 +44,7 @@
  * reserved symbols (along with \ and /), otherwise illegal to store
  * in filenames in NTFS
  */
-#define UNI_ASTERI    (__u16) ('*' + 0xF000)
+#define UNI_ASTERISK    (__u16) ('*' + 0xF000)
 #define UNI_QUESTION    (__u16) ('?' + 0xF000)
 #define UNI_COLON       (__u16) (':' + 0xF000)
 #define UNI_GRTRTHAN    (__u16) ('>' + 0xF000)