| Message ID | 1291233547-9437-1-git-send-email-sven.eckelmann@gmx.de |
|---|---|
| State | Superseded, archived |
| Headers | show |
On Wednesday 01 December 2010 20:59:07 Sven Eckelmann wrote: > set_bit and test_bit provide an efficient way to set and test bits of an > unsigned long. Since test_bit() / set_bit() expect 32Bit values we should at least add a comment next to TYPE_OF_WORD, so that nobody changes this define ?! Cheers, Marek
On Sunday 12 December 2010 20:29:46 Marek Lindner wrote: > On Wednesday 01 December 2010 20:59:07 Sven Eckelmann wrote: > > set_bit and test_bit provide an efficient way to set and test bits of an > > unsigned long. > > Since test_bit() / set_bit() expect 32Bit values we should at least add a > comment next to TYPE_OF_WORD, so that nobody changes this define ?! I applied the patch including a comment in revision 1889. Thanks, Marek
Marek Lindner wrote: > On Wednesday 01 December 2010 20:59:07 Sven Eckelmann wrote: > > set_bit and test_bit provide an efficient way to set and test bits of an > > unsigned long. > > Since test_bit() / set_bit() expect 32Bit values we should at least add a > comment next to TYPE_OF_WORD, so that nobody changes this define ?! unsigned long is not always 32 bit.... I will send a patch which completely replaces TYPE_OF_WORD Best regards, Sven
Marek Lindner wrote: > On Sunday 12 December 2010 20:29:46 Marek Lindner wrote: > > On Wednesday 01 December 2010 20:59:07 Sven Eckelmann wrote: > > > set_bit and test_bit provide an efficient way to set and test bits of > > > an unsigned long. > > > > Since test_bit() / set_bit() expect 32Bit values we should at least add a > > comment next to TYPE_OF_WORD, so that nobody changes this define ?! > > I applied the patch including a comment in revision 1889. Ok, since there were a race condition, please ignore the v2 version of the patches. v3 will be delivered in some minutes. Best regards, Sven
On Sunday 12 December 2010 21:46:58 Sven Eckelmann wrote: > "Use kernel facilities for bit operations" increased accidently the > character count on a line to more than 80 characters when counting a tab > as 8 characters. Applied in revision 1893. Thanks, Marek
diff --git a/batman-adv/bitarray.c b/batman-adv/bitarray.c index 814274f..14606fc 100644 --- a/batman-adv/bitarray.c +++ b/batman-adv/bitarray.c @@ -40,7 +40,7 @@ uint8_t get_bit_status(TYPE_OF_WORD *seq_bits, uint32_t last_seqno, /* which position in the selected word */ word_offset = (last_seqno - curr_seqno) % WORD_BIT_SIZE; - if (seq_bits[word_num] & 1 << word_offset) + if (test_bit(word_offset, &seq_bits[word_num])) return 1; else return 0; @@ -61,7 +61,7 @@ void bit_mark(TYPE_OF_WORD *seq_bits, int32_t n) /* which position in the selected word */ word_offset = n % WORD_BIT_SIZE; - seq_bits[word_num] |= 1 << word_offset; /* turn the position on */ + set_bit(word_offset, &seq_bits[word_num]); /* turn the position on */ } /* shift the packet array by n places. */
set_bit and test_bit provide an efficient way to set and test bits of an unsigned long. This also fixes the problem that a very old ogm got not recorded as received due to the missing constant definition "1" as unsigned long inside the bit_mark operation - also known as "1UL". Reported-by: David Miller <davem@davemloft.net> Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de> --- batman-adv/bitarray.c | 4 ++-- 1 files changed, 2 insertions(+), 2 deletions(-)