new floating point bugs in 2.4.5-ac4
Here are two new uses of floating point that popped up in the 2.4.5-ac4 kernel. While the expressions that use FP are trivial, at least my version of gcc does not calculate them at compile time.
As a bonus, two old, but still existing FP uses are also included.
Dawson
MC linux bug database: http://hands.stanford.edu/linux
############################################################
# New errors.
#
---------------------------------------------------------
[BUG] DMFE_TX_TIMEOUT is HZ * 1.5 which seems easy to fix.
/u2/engler/mc/oses/linux/2.4.5-ac4/drivers/net/dmfe.c:1112:dmfe_timer: ERROR:FLOAT: cannot use fp in kernel
if ( db->tx_packet_cnt &&
((jiffies - dev->trans_start) > DMFE_TX_KICK) ) {
outl(0x1, dev->base_addr + DCR1); /* Tx polling again */
/* TX Timeout */Error --->
if ( (jiffies - dev->trans_start) > DMFE_TX_TIMEOUT ) {---------------------------------------------------------
[BUG] DMFE_TX_KICK is (HZ * 0.5) which gcc does as floating point. Fix is
trivial: just divide by 2 instead.
/u2/engler/mc/oses/linux/2.4.5-ac4/drivers/net/dmfe.c:1108:dmfe_timer: ERROR:FLOAT: cannot use fp in kernel
} db->interval_rx_cnt = 0; /* TX polling kick monitor */ if ( db->tx_packet_cnt &&
Error --->
((jiffies - dev->trans_start) > DMFE_TX_KICK) ) {############################################################
# Existing, unfixed errors
#
---------------------------------------------------------
[BUG]
/u2/engler/mc/oses/linux/2.4.5-ac4/drivers/video/sgivwfb.c:731:sgivwfb_set_var: ERROR:FLOAT: cannot use fp in kernel
var->green.msb_right = 0; var->blue.msb_right = 0; var->transp.msb_right = 0;
/* set video timing information */
Error --->
var->pixclock = (__u32)(1.0e+9/(float)timing->cfreq);
---------------------------------------------------------
[BUG]
/u2/engler/mc/oses/linux/2.4.5-ac4/drivers/video/sgivwfb.c:664:sgivwfb_set_var: ERROR:FLOAT: cannot use fp in kernel
return -EINVAL; /* Resolution to high */
/* XXX FIXME - should try to pick best refresh rate */ /* for now, pick closest dot-clock within 3MHz*/
#error "Floating point not allowed in kernel"
Error --->
req_dot = (int)((1.0e3/1.0e6) / (1.0e-12 * (float)var->pixclock));
############################################################
# Old fixed
#
[FIXED] sis_main.c:crtc_to_var:FLOAT: cannot use fp in kerne /u2/engler/mc/oses/linux/2.4.4/drivers/video/sis/sis_main.c:588:crtc_to_var: ERROR:FLOAT: cannot use fp in kernel
Quoted 6 lines · In article <[email protected]> you wrote:
Here are two new uses of floating point that popped up in the 2.4.5-ac4 kernel. While the expressions that use FP are trivial, at least my version of gcc does not calculate them at compile time.
[BUG]
DMFE_TX_KICKis (HZ * 0.5) which gcc does as floating point. Fix is
trivial: just divide by 2 instead.
Fixed in 2.4.5-ac5 already