host_path_cntl = INREG(HOST_PATH_CNTL);
        rbbm_soft_reset = INREG(RBBM_SOFT_RESET);
 
-       if (rinfo->family == CHIP_FAMILY_R300 ||
-           rinfo->family == CHIP_FAMILY_R350 ||
-           rinfo->family == CHIP_FAMILY_RV350) {
+       if (IS_R300_VARIANT(rinfo)) {
                u32 tmp;
 
                OUTREG(RBBM_SOFT_RESET, (rbbm_soft_reset |
        INREG(HOST_PATH_CNTL);
        OUTREG(HOST_PATH_CNTL, host_path_cntl);
 
-       if (rinfo->family != CHIP_FAMILY_R300 &&
-           rinfo->family != CHIP_FAMILY_R350 &&
-           rinfo->family != CHIP_FAMILY_RV350)
+       if (!IS_R300_VARIANT(rinfo))
                OUTREG(RBBM_SOFT_RESET, rbbm_soft_reset);
 
        OUTREG(CLOCK_CNTL_INDEX, clock_cntl_index);
        radeonfb_engine_reset(rinfo);
 
        radeon_fifo_wait (1);
-       if ((rinfo->family != CHIP_FAMILY_R300) &&
-           (rinfo->family != CHIP_FAMILY_R350) &&
-           (rinfo->family != CHIP_FAMILY_RV350))
+       if (IS_R300_VARIANT(rinfo)) {
+               OUTREG(RB2D_DSTCACHE_MODE, INREG(RB2D_DSTCACHE_MODE) |
+                      RB2D_DC_AUTOFLUSH_ENABLE |
+                      RB2D_DC_DC_DISABLE_IGNORE_PE);
+       } else {
+               /* This needs to be double checked with ATI. Latest X driver
+                * completely "forgets" to set this register on < r3xx, and
+                * we used to just write 0 there... I'll keep the 0 and update
+                * that when we have sorted things out on X side.
+                */
                OUTREG(RB2D_DSTCACHE_MODE, 0);
+       }
 
        radeon_fifo_wait (3);
        /* We re-read MC_FB_LOCATION from card as it can have been
 
        radeon_pll_errata_after_data(rinfo);
 
        /* Set PPLL ref. div */
-       if (rinfo->family == CHIP_FAMILY_R300 ||
+       if (IS_R300_VARIANT(rinfo) ||
            rinfo->family == CHIP_FAMILY_RS300 ||
-           rinfo->family == CHIP_FAMILY_R350 ||
-           rinfo->family == CHIP_FAMILY_RV350 ||
-           rinfo->family == CHIP_FAMILY_RV380 ) {
+           rinfo->family == CHIP_FAMILY_RS400 ||
+           rinfo->family == CHIP_FAMILY_RS480) {
                if (mode->ppll_ref_div & R300_PPLL_REF_DIV_ACC_MASK) {
                        /* When restoring console mode, use saved PPLL_REF_DIV
                         * setting.
                /* Not all chip revs have the same format for this register,
                 * extract the source selection
                 */
-               if (rinfo->family == CHIP_FAMILY_R200 ||
-                   rinfo->family == CHIP_FAMILY_R300 ||
-                   rinfo->family == CHIP_FAMILY_R350 ||
-                   rinfo->family == CHIP_FAMILY_RV350) {
+               if (rinfo->family == CHIP_FAMILY_R200 || IS_R300_VARIANT(rinfo)) {
                        source = (fp2_gen_cntl >> 10) & 0x3;
                        /* sourced from transform unit, check for transform unit
                         * own source
             (rinfo->family == CHIP_FAMILY_RS200) ||
             (rinfo->family == CHIP_FAMILY_RS300) ||
             (rinfo->family == CHIP_FAMILY_RC410) ||
+            (rinfo->family == CHIP_FAMILY_RS400) ||
            (rinfo->family == CHIP_FAMILY_RS480) ) {
           u32 tom = INREG(NB_TOM);
           tmp = ((((tom >> 16) - (tom & 0xffff) + 1) << 6) * 1024);
 
        CHIP_FAMILY_RV380,    /* RV370/RV380/M22/M24 */
        CHIP_FAMILY_R420,     /* R420/R423/M18 */
        CHIP_FAMILY_RC410,
+       CHIP_FAMILY_RS400,
        CHIP_FAMILY_RS480,
        CHIP_FAMILY_LAST,
 };
 /*
  * 2D Engine helper routines
  */
-static inline void radeon_engine_flush (struct radeonfb_info *rinfo)
+
+static inline void _radeon_fifo_wait(struct radeonfb_info *rinfo, int entries)
 {
        int i;
 
-       /* initiate flush */
-       OUTREGP(RB2D_DSTCACHE_CTLSTAT, RB2D_DC_FLUSH_ALL,
-               ~RB2D_DC_FLUSH_ALL);
-
-       for (i=0; i < 2000000; i++) {
-               if (!(INREG(RB2D_DSTCACHE_CTLSTAT) & RB2D_DC_BUSY))
+       for (i=0; i<2000000; i++) {
+               if ((INREG(RBBM_STATUS) & 0x7f) >= entries)
                        return;
                udelay(1);
        }
-       printk(KERN_ERR "radeonfb: Flush Timeout !\n");
+       printk(KERN_ERR "radeonfb: FIFO Timeout !\n");
 }
 
-
-static inline void _radeon_fifo_wait(struct radeonfb_info *rinfo, int entries)
+static inline void radeon_engine_flush (struct radeonfb_info *rinfo)
 {
        int i;
 
-       for (i=0; i<2000000; i++) {
-               if ((INREG(RBBM_STATUS) & 0x7f) >= entries)
+       /* Initiate flush */
+       OUTREGP(DSTCACHE_CTLSTAT, RB2D_DC_FLUSH_ALL,
+               ~RB2D_DC_FLUSH_ALL);
+
+       /* Ensure FIFO is empty, ie, make sure the flush commands
+        * has reached the cache
+        */
+       _radeon_fifo_wait (rinfo, 64);
+
+       /* Wait for the flush to complete */
+       for (i=0; i < 2000000; i++) {
+               if (!(INREG(DSTCACHE_CTLSTAT) & RB2D_DC_BUSY))
                        return;
                udelay(1);
        }
-       printk(KERN_ERR "radeonfb: FIFO Timeout !\n");
+       printk(KERN_ERR "radeonfb: Flush Timeout !\n");
 }
 
 
 
 #define SC_BOTTOM_RIGHT                        0x16F0  
 #define SRC_SC_BOTTOM_RIGHT                    0x16F4  
 #define RB2D_DSTCACHE_MODE                    0x3428
-#define RB2D_DSTCACHE_CTLSTAT                 0x342C
+#define RB2D_DSTCACHE_CTLSTAT_broken          0x342C /* do not use */
 #define LVDS_GEN_CNTL                         0x02d0
 #define LVDS_PLL_CNTL                         0x02d4
 #define FP2_GEN_CNTL                           0x0288
 #define RB2D_DC_FLUSH_ALL                         (RB2D_DC_FLUSH_2D | RB2D_DC_FREE_2D)
 #define RB2D_DC_BUSY                              (1 << 31)
 
+/* DSTCACHE_MODE bits constants */
+#define RB2D_DC_AUTOFLUSH_ENABLE                   (1 << 8)
+#define RB2D_DC_DC_DISABLE_IGNORE_PE               (1 << 17)
 
 /* CRTC_GEN_CNTL bit constants */
 #define CRTC_DBL_SCAN_EN                           0x00000001