Sorry! These underflow flow check isn't released. I probably could though. It doesn't create too many false positives. See attached. drivers/clk/qcom/clk-rcg.c:618 clk_rcg_pixel_determine_rate() warn: 'request - delta' negative user limit promoted to high drivers/clk/qcom/clk-rcg2.c:617 clk_edp_pixel_determine_rate() warn: 'request - delta' negative user limit promoted to high drivers/clk/qcom/clk-rcg2.c:794 clk_pixel_determine_rate() warn: 'request - delta' negative user limit promoted to high drivers/mfd/tps6594-i2c.c:159 tps6594_i2c_write() warn: 'count - 2' negative user limit promoted to high drivers/gpu/drm/radeon/si.c:4594 si_vm_packet3_gfx_check() warn: 'pkt->count - 2' negative user limit promoted to high drivers/gpu/drm/radeon/si.c:4697 si_vm_packet3_compute_check() warn: 'pkt->count - 2' negative user limit promoted to high drivers/net/dsa/microchip/ksz8863_smi.c:66 ksz8863_mdio_write() warn: 'count - 4' negative user limit promoted to high net/wireless/chan.c:456 cfg80211_set_chans_dfs_state() warn: 'center_freq + bandwidth / 2 - 10' negative user limit promoted to high mm/page_isolation.c:558 start_isolate_page_range() warn: '(null)' negative user limit promoted to high I think the radeon and wireless warnings look like real bugs. The rest are false positives. I've looked at all these warnings before so they should have all been false positives... :/ regards, dan carpenter
The functionality is interesting. Pity not to have it in smatch. Maybe you could differentiate 2 levels: "warning" VS "check", similarly to checkpatch. :) Julien
