[Qemu-devel] [PATCH RFC] dma_rw.h (was Re: [PATCH 0/7] AMD IOMMU emulation patchset v4)

[Michael S. Tsirkin]

So I think the following will give the idea of what an API
might look like that will let us avoid the scary hacks in
e.g. the ide layer and other generic layers that need to do DMA,
without either binding us to pci, adding more complexity with
callbacks, or losing type safety with casts and void*.

Basically we have DMADevice that we can use container_of on
to get a PCIDevice from, and DMAMmu that will get instanciated
in a specific MMU.

This is not complete code - just a header - I might complete
this later if/when there's interest or hopefully someone interested
in iommu emulation will.

Notes:
the IOMMU_PERM_RW code seem unused, so I replaced
this with plain is_write. Is it ever useful?

It seems that invalidate callback should be able to
get away with just a device, so I switched to that
from a void pointer for type safety.
Seems enough for the users I saw.

I saw devices do stl_le_phys and such, these
might need to be wrapped as well.

Signed-off-by: Michael S. Tsirkin <address@hidden>

---

diff --git a/hw/dma_rw.h b/hw/dma_rw.h
new file mode 100644
index 0000000..d63fd17
--- /dev/null
+++ b/hw/dma_rw.h
@@ -0,0 +1,122 @@
+#ifndef DMA_RW_H
+#define DMA_RW_H
+
+#include "qemu-common.h"
+
+/* We currently only have pci mmus, but using
+   a generic type makes it possible to use this
+   e.g. from the generic ide code without callbacks. */
+typedef uint64_t dma_addr_t;
+
+typedef struct DMAMmu DMAMmu;
+typedef struct DMADevice DMADevice;
+
+typedef int DMATranslateFunc(DMAMmu *mmu,
+                             DMADevice *dev,
+                             dma_addr_t addr,
+                             dma_addr_t *paddr,
+                             dma_addr_t *len,
+                             int is_write);
+
+typedef int DMAInvalidateMapFunc(DMADevice *);
+struct DMAMmu {
+       /* invalidate, etc. */
+       DmaTranslateFunc *translate;
+};
+
+struct DMADevice {
+       DMAMmu *mmu;
+       DMAInvalidateMapFunc *invalidate;
+};
+
+void dma_device_init(DMADevice *, DMAMmu *, DMAInvalidateMapFunc *);
+
+static inline void dma_memory_rw(DMADevice *dev,
+                                dma_addr_t addr,
+                                void *buf,
+                                uint32_t len,
+                                int is_write)
+{
+    uint32_t plen;
+    /* Fast-path non-iommu.
+     * More importantly, makes it obvious what this function does. */
+    if (!dev->mmu) {
+       cpu_physical_memory_rw(paddr, buf, plen, is_write);
+       return;
+    }
+    while (len) {
+        err = dev->mmu->translate(iommu, dev, addr, &paddr, &plen, is_write);
+        if (err) {
+            return;
+        }
+                                      
+        /* The translation might be valid for larger regions. */
+        if (plen > len) {
+            plen = len;
+        }
+    
+        cpu_physical_memory_rw(paddr, buf, plen, is_write);
+
+        len -= plen;
+        addr += plen;
+        buf += plen;
+    }
+}
+
+void *dma_memory_map(DMADevice *dev,
+                            dma_addr_t addr,
+                            uint32_t *len,
+                            int is_write);
+void dma_memory_unmap(DMADevice *dev,
+                     void *buffer,
+                     uint32_t len,
+                     int is_write,
+                     uint32_t access_len);
+
+
++#define DEFINE_DMA_LD(suffix, size)                                       \
++uint##size##_t dma_ld##suffix(DMADevice *dev, dma_addr_t addr)            \
++{                                                                         \
++    int err;                                                              \
++    target_phys_addr_t paddr, plen;                                       \
++    if (!dev->mmu) {                                                      \
++        return ld##suffix##_phys(addr, val);                              \
++    }                                                                     \
++                                                                          \
++    err = dev->mmu->translate(dev->bus->iommu, dev,                       \
++                              addr, &paddr, &plen, IOMMU_PERM_READ);      \
++    if (err || (plen < size / 8))                                         \
++        return 0;                                                         \
++                                                                          \
++    return ld##suffix##_phys(paddr);                                      \
++}
++
++#define DEFINE_DMA_ST(suffix, size)                                       \
++void dma_st##suffix(DMADevice *dev, dma_addr_t addr, uint##size##_t val)  \
++{                                                                         \
++    int err;                                                              \
++    target_phys_addr_t paddr, plen;                                       \
++                                                                          \
++    if (!dev->mmu) {                                                      \
++        st##suffix##_phys(addr, val);                                     \
++        return;                                                           \
++    }                                                                     \
++    err = dev->mmu->translate(dev->bus->iommu, dev,                       \
++                              addr, &paddr, &plen, IOMMU_PERM_WRITE);     \
++    if (err || (plen < size / 8))                                         \
++        return;                                                           \
++                                                                          \
++    st##suffix##_phys(paddr, val);                                        \
++}
+
+DEFINE_DMA_LD(ub, 8)
+DEFINE_DMA_LD(uw, 16)
+DEFINE_DMA_LD(l, 32)
+DEFINE_DMA_LD(q, 64)
+
+DEFINE_DMA_ST(b, 8)
+DEFINE_DMA_ST(w, 16)
+DEFINE_DMA_ST(l, 32)
+DEFINE_DMA_ST(q, 64)
+
+#endif
diff --git a/hw/pci.h b/hw/pci.h
index 1c6075e..9737f0e 100644
--- a/hw/pci.h
+++ b/hw/pci.h
@@ -5,6 +5,7 @@
 #include "qobject.h"
 
 #include "qdev.h"
+#include "dma_rw.h"
 
 /* PCI includes legacy ISA access.  */
 #include "isa.h"
@@ -119,6 +120,10 @@ enum {
 
 struct PCIDevice {
     DeviceState qdev;
+
+    /* For devices that do DMA. */
+    DMADevice dma;
+
     /* PCI config space */
     uint8_t *config;