Make DMA bottom-half driven (v2)

The current DMA routines are driven by a call in main_loop_wait() after every select. This patch converts the DMA code to be driven by a constantly rescheduled bottom half. The advantage of using a scheduled bottom half is that we can stop scheduling the bottom half when there no DMA channels are runnable. This means we can potentially detect this case and sleep longer in the main loop. The only two architectures implementing DMA_run() are cris and i386. For cris, I converted it to a simple repeating bottom half. I've only compile tested this as cris does not seem to work on a 64-bit host. It should be functionally identical to the previous implementation so I expect it to work. For x86, I've made sure to only fire the DMA bottom half if there is a DMA channel that is runnable. The effect of this is that unless you're using sb16 or a floppy disk, the DMA bottom half never fires. You probably should test this malc. My own benchmarks actually show slight improvement by it's possible the change in timing could affect your demos. Since v1, I've changed the code to use a BH instead of a timer. cris at least seems to depend on faster than 10ms polling. Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@5573 c046a42c-6fe2-441c-8c8c-71466251a162

Make DMA bottom-half driven (v2)
The current DMA routines are driven by a call in main_loop_wait() after every select. This patch converts the DMA code to be driven by a constantly rescheduled bottom half. The advantage of using a scheduled bottom half is that we can stop scheduling the bottom half when there no DMA channels are runnable. This means we can potentially detect this case and sleep longer in the main loop. The only two architectures implementing DMA_run() are cris and i386. For cris, I converted it to a simple repeating bottom half. I've only compile tested this as cris does not seem to work on a 64-bit host. It should be functionally identical to the previous implementation so I expect it to work. For x86, I've made sure to only fire the DMA bottom half if there is a DMA channel that is runnable. The effect of this is that unless you're using sb16 or a floppy disk, the DMA bottom half never fires. You probably should test this malc. My own benchmarks actually show slight improvement by it's possible the change in timing could affect your demos. Since v1, I've changed the code to use a BH instead of a timer. cris at least seems to depend on faster than 10ms polling. Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@5573 c046a42c-6fe2-441c-8c8c-71466251a162
aliguori
1 parent 1b435b10
Showing 9 changed files with 40 additions and 26 deletions
hw/an5206.c
hw/dma.c
hw/etraxfs_dma.c
hw/integratorcp.c
hw/isa.h
hw/shix.c
hw/sun4m.c
hw/sun4u.c
vl.c
@@ -24,10 +24,6 @@ void irq_info(void)
 {
 }
-void DMA_run (void)
-{
-}
-
 /* Board init.  */
 static void an5206_init(ram_addr_t ram_size, int vga_ram_size,
@@ -78,6 +78,8 @@ enum {
 };
+static void DMA_run (void);
+
 static int channels[8] = {-1, 2, 3, 1, -1, -1, -1, 0};
 static void write_page (void *opaque, uint32_t nport, uint32_t data)
@@ -214,6 +216,7 @@ static void write_cont (void *opaque, uint32_t nport, uint32_t data)
             d->status &= ~(1 << (ichan + 4));
         }
         d->status &= ~(1 << ichan);
+        DMA_run();
         break;
     case 0x0a:                  /* single mask */
@@ -221,6 +224,7 @@ static void write_cont (void *opaque, uint32_t nport, uint32_t data)
             d->mask |= 1 << (data & 3);
         else
             d->mask &= ~(1 << (data & 3));
+        DMA_run();
         break;
     case 0x0b:                  /* mode */
@@ -255,10 +259,12 @@ static void write_cont (void *opaque, uint32_t nport, uint32_t data)
     case 0x0e:                  /* clear mask for all channels */
         d->mask = 0;
+        DMA_run();
         break;
     case 0x0f:                  /* write mask for all channels */
         d->mask = data;
+        DMA_run();
         break;
     default:
@@ -310,6 +316,7 @@ void DMA_hold_DREQ (int nchan)
     ichan = nchan & 3;
     linfo ("held cont=%d chan=%d\n", ncont, ichan);
     dma_controllers[ncont].status |= 1 << (ichan + 4);
+    DMA_run();
 }
 void DMA_release_DREQ (int nchan)
@@ -320,6 +327,7 @@ void DMA_release_DREQ (int nchan)
     ichan = nchan & 3;
     linfo ("released cont=%d chan=%d\n", ncont, ichan);
     dma_controllers[ncont].status &= ~(1 << (ichan + 4));
+    DMA_run();
 }
 static void channel_run (int ncont, int ichan)
@@ -347,10 +355,13 @@ static void channel_run (int ncont, int ichan)
     ldebug ("dma_pos %d size %d\n", n, (r->base[COUNT] + 1) << ncont);
 }
-void DMA_run (void)
+static QEMUBH *dma_bh;
+
+static void DMA_run (void)
 {
     struct dma_cont *d;
     int icont, ichan;
+    int rearm = 0;
     d = dma_controllers;
@@ -360,10 +371,20 @@ void DMA_run (void)
             mask = 1 << ichan;
-            if ((0 == (d->mask & mask)) && (0 != (d->status & (mask << 4))))
+            if ((0 == (d->mask & mask)) && (0 != (d->status & (mask << 4)))) {
                 channel_run (icont, ichan);
+                rearm = 1;
+            }
         }
     }
+
+    if (rearm)
+        qemu_bh_schedule_idle(dma_bh);
+}
+
+static void DMA_run_bh(void *unused)
+{
+    DMA_run();
 }
 void DMA_register_channel (int nchan,
@@ -534,6 +555,9 @@ static int dma_load (QEMUFile *f, void *opaque, int version_id)
         qemu_get_8s (f, &r->dack);
         qemu_get_8s (f, &r->eop);
     }
+
+    DMA_run();
+
     return 0;
 }
@@ -545,4 +569,6 @@ void DMA_init (int high_page_enable)
               high_page_enable ? 0x488 : -1);
     register_savevm ("dma", 0, 1, dma_save, dma_load, &dma_controllers[0]);
     register_savevm ("dma", 1, 1, dma_save, dma_load, &dma_controllers[1]);
+
+    dma_bh = qemu_bh_new(DMA_run_bh, NULL);
 }
@@ -24,6 +24,8 @@
 #include <stdio.h>
 #include <sys/time.h>
 #include "hw.h"
+#include "qemu-common.h"
+#include "sysemu.h"
 #include "etraxfs_dma.h"
@@ -190,6 +192,8 @@ struct fs_dma_ctrl
 	int nr_channels;
 	struct fs_dma_channel *channels;
+
+        QEMUBH *bh;
 };
 static inline uint32_t channel_reg(struct fs_dma_ctrl *ctrl, int c, int reg)
@@ -712,11 +716,12 @@ void etraxfs_dmac_connect_client(void *opaque, int c,
 }
-static void *etraxfs_dmac;
-void DMA_run(void)
+static void DMA_run(void *opaque)
 {
-	if (etraxfs_dmac)
-		etraxfs_dmac_run(etraxfs_dmac);
+    struct fs_dma_ctrl *etraxfs_dmac = opaque;
+    if (vm_running)
+        etraxfs_dmac_run(etraxfs_dmac);
+    qemu_bh_schedule_idle(etraxfs_dmac->bh);
 }
 void *etraxfs_dmac_init(CPUState *env, 
@@ -729,6 +734,9 @@ void *etraxfs_dmac_init(CPUState *env,
 	if (!ctrl)
 		return NULL;
+        ctrl->bh = qemu_bh_new(DMA_run, ctrl);
+        qemu_bh_schedule_idle(ctrl->bh);
+
 	ctrl->base = base;
 	ctrl->env = env;
 	ctrl->nr_channels = nr_channels;
@@ -747,8 +755,6 @@ void *etraxfs_dmac_init(CPUState *env,
 					      ctrl->channels[i].regmap);
 	}
-	/* Hax, we only support one DMA controller at a time.  */
-	etraxfs_dmac = ctrl;
 	return ctrl;
   err:
 	qemu_free(ctrl->channels);
@@ -15,10 +15,6 @@
 #include "arm-misc.h"
 #include "net.h"
-void DMA_run (void)
-{
-}
-
 typedef struct {
     uint32_t flash_offset;
     uint32_t cm_osc;
@@ -19,7 +19,6 @@ int DMA_write_memory (int nchan, void *buf, int pos, int size);
 void DMA_hold_DREQ (int nchan);
 void DMA_release_DREQ (int nchan);
 void DMA_schedule(int nchan);
-void DMA_run (void);
 void DMA_init (int high_page_enable);
 void DMA_register_channel (int nchan,
                            DMA_transfer_handler transfer_handler,
@@ -35,11 +35,6 @@
 #define BIOS_FILENAME "shix_bios.bin"
 #define BIOS_ADDRESS 0xA0000000
-void DMA_run(void)
-{
-    /* XXXXX */
-}
-
 void irq_info(void)
 {
     /* XXXXX */
@@ -164,7 +164,6 @@ int DMA_write_memory (int nchan, void *buf, int pos, int size)
 void DMA_hold_DREQ (int nchan) {}
 void DMA_release_DREQ (int nchan) {}
 void DMA_schedule(int nchan) {}
-void DMA_run (void) {}
 void DMA_init (int high_page_enable) {}
 void DMA_register_channel (int nchan,
                            DMA_transfer_handler transfer_handler,
@@ -79,7 +79,6 @@ int DMA_write_memory (int nchan, void *buf, int pos, int size)
 void DMA_hold_DREQ (int nchan) {}
 void DMA_release_DREQ (int nchan) {}
 void DMA_schedule(int nchan) {}
-void DMA_run (void) {}
 void DMA_init (int high_page_enable) {}
 void DMA_register_channel (int nchan,
                            DMA_transfer_handler transfer_handler,
@@ -7983,8 +7983,6 @@ void main_loop_wait(int timeout)
         if (likely(!(cur_cpu->singlestep_enabled & SSTEP_NOTIMER)))
         qemu_run_timers(&active_timers[QEMU_TIMER_VIRTUAL],
                         qemu_get_clock(vm_clock));
-        /* run dma transfers, if any */
-        DMA_run();
     }
     /* real time timers */