From: Vadim Lobanov This patch modifies the fdtable minimum allocation unit, setting it to a fixed size of 1024 bytes instead of basing it on the architecture page size. This leaves the algorithm heuristics unchanged for the common 4K page size, but does rebalance the "memory wasted versus number of reallocations" tradeoff for architectures with larger page sizes (like 64K). Signed-off-by: Vadim Lobanov Signed-off-by: Andrew Morton --- fs/file.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff -puN fs/file.c~fdtable-implement-new-pagesize-based-fdtable-allocator-bound-minimum-allocation-size fs/file.c --- a/fs/file.c~fdtable-implement-new-pagesize-based-fdtable-allocator-bound-minimum-allocation-size +++ a/fs/file.c @@ -139,12 +139,12 @@ static struct fdtable * alloc_fdtable(un * Figure out how many fds we actually want to support in this fdtable. * Allocation steps are keyed to the size of the fdarray, since it * grows far faster than any of the other dynamic data. We try to fit - * the fdarray into page-sized chunks: starting at a quarter of a page, + * the fdarray into comfortable page-tuned chunks: starting at 1024B * and growing in powers of two from there on. */ - nr /= (PAGE_SIZE / 4 / sizeof(struct file *)); + nr /= (1024 / sizeof(struct file *)); nr = roundup_pow_of_two(nr + 1); - nr *= (PAGE_SIZE / 4 / sizeof(struct file *)); + nr *= (1024 / sizeof(struct file *)); if (nr > NR_OPEN) nr = NR_OPEN; _