The `mult' parameter is negated if the user data begins with a '-' so that the final value has the appropriate sign. But `mult' is only used if the user data does not include a "units" suffix. In this case, `mult' is overridden with the numeric scale conveyed by the units suffix, but retains the sign of the original value. Having `mult' serving double-duty works but is confusing. Use a new local variable to store the sign of the user data instead. This also fixes a pitfall of passing 0 to `mult', expecting it to be ignored when a units suffix is specified, but having the effect of taking the absolute value of the user-provided data. Signed-off-by: Chris Rorvick <chris@xxxxxxxxxxx> Reviewed-by: Andreas Dilger <andreas.dilger@xxxxxxxxx> --- drivers/staging/lustre/lustre/obdclass/lprocfs_status.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/drivers/staging/lustre/lustre/obdclass/lprocfs_status.c b/drivers/staging/lustre/lustre/obdclass/lprocfs_status.c index a7b270e..22e13f0 100644 --- a/drivers/staging/lustre/lustre/obdclass/lprocfs_status.c +++ b/drivers/staging/lustre/lustre/obdclass/lprocfs_status.c @@ -1862,6 +1862,7 @@ int lprocfs_write_frac_u64_helper(const char *buffer, unsigned long count, char kernbuf[22], *end, *pbuf; __u64 whole, frac = 0, units; unsigned frac_d = 1; + int sign = 1; if (count > (sizeof(kernbuf) - 1)) return -EINVAL; @@ -1872,7 +1873,7 @@ int lprocfs_write_frac_u64_helper(const char *buffer, unsigned long count, kernbuf[count] = '\0'; pbuf = kernbuf; if (*pbuf == '-') { - mult = -mult; + sign = -1; pbuf++; } @@ -1909,11 +1910,11 @@ int lprocfs_write_frac_u64_helper(const char *buffer, unsigned long count, } /* Specified units override the multiplier */ if (units > 1) - mult = mult < 0 ? -units : units; + mult = units; frac *= mult; do_div(frac, frac_d); - *val = whole * mult + frac; + *val = sign * (whole * mult + frac); return 0; } EXPORT_SYMBOL(lprocfs_write_frac_u64_helper); -- 2.1.0 _______________________________________________ devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxx http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel