|
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [Xen-devel] [RFC PATCH 23/23] X86/vvtd: Add queued invalidation (QI) support
From: Chao Gao <chao.gao@xxxxxxxxx>
Queued Invalidation Interface is an expanded invalidation interface with
extended capabilities. Hardware implementations report support for queued
invalidation interface through the Extended Capability Register. The queued
invalidation interface uses an Invalidation Queue (IQ), which is a circular
buffer in system memory. Software submits commands by writing Invalidation
Descriptors to the IQ.
In this patch, a new function viommu_process_iq() is used for emulating how
hardware handles invalidation requests through QI.
Signed-off-by: Chao Gao <chao.gao@xxxxxxxxx>
Signed-off-by: Lan Tianyu <tianyu.lan@xxxxxxxxx>
---
xen/arch/x86/hvm/vvtd.c | 248 ++++++++++++++++++++++++++++++++++++
xen/drivers/passthrough/vtd/iommu.h | 29 ++++-
2 files changed, 276 insertions(+), 1 deletion(-)
diff --git a/xen/arch/x86/hvm/vvtd.c b/xen/arch/x86/hvm/vvtd.c
index 28d8f36..6d11075 100644
--- a/xen/arch/x86/hvm/vvtd.c
+++ b/xen/arch/x86/hvm/vvtd.c
@@ -423,6 +423,185 @@ static int vvtd_log_fault(struct vvtd *vvtd,
return X86EMUL_OKAY;
}
+/*
+ * Process a invalidation descriptor. Currently, only Two types descriptors,
+ * Interrupt Entry Cache invalidation descritor and Invalidation Wait
+ * Descriptor are handled.
+ * @vvtd: the virtual vtd instance
+ * @i: the index of the invalidation descriptor to be processed
+ *
+ * If success return 0, or return -1 when failure.
+ */
+static int process_iqe(struct vvtd *vvtd, int i)
+{
+ uint64_t iqa, addr;
+ struct qinval_entry *qinval_page;
+ void *pg;
+ int ret;
+
+ vvtd_get_reg_quad(vvtd, DMAR_IQA_REG, iqa);
+ ret = map_guest_page(vvtd->domain, DMA_IQA_ADDR(iqa)>>PAGE_SHIFT,
+ (void**)&qinval_page);
+ if ( ret )
+ {
+ gdprintk(XENLOG_ERR, "Can't map guest IRT (rc %d)", ret);
+ return -1;
+ }
+
+ switch ( qinval_page[i].q.inv_wait_dsc.lo.type )
+ {
+ case TYPE_INVAL_WAIT:
+ if ( qinval_page[i].q.inv_wait_dsc.lo.sw )
+ {
+ addr = (qinval_page[i].q.inv_wait_dsc.hi.saddr << 2);
+ ret = map_guest_page(vvtd->domain, addr >> PAGE_SHIFT, &pg);
+ if ( ret )
+ {
+ gdprintk(XENLOG_ERR, "Can't map guest memory to inform guest "
+ "IWC completion (rc %d)", ret);
+ goto error;
+ }
+ *(uint32_t *)((uint64_t)pg + (addr & ~PAGE_MASK)) =
+ qinval_page[i].q.inv_wait_dsc.lo.sdata;
+ unmap_guest_page(pg);
+ }
+
+ /*
+ * The following code generates an invalidation completion event
+ * indicating the invalidation wait descriptor completion. Note that
+ * the following code fragment is not tested properly.
+ */
+ if ( qinval_page[i].q.inv_wait_dsc.lo.iflag )
+ {
+ uint32_t ie_data, ie_addr;
+ if ( !vvtd_test_and_set_bit(vvtd, DMAR_ICS_REG, DMA_ICS_IWC_BIT) )
+ {
+ __vvtd_set_bit(vvtd, DMAR_IECTL_REG, DMA_IECTL_IP_BIT);
+ if ( !vvtd_test_bit(vvtd, DMAR_IECTL_REG, DMA_IECTL_IM_BIT) )
+ {
+ ie_data = vvtd_get_reg(vvtd, DMAR_IEDATA_REG);
+ ie_addr = vvtd_get_reg(vvtd, DMAR_IEADDR_REG);
+ vvtd_generate_interrupt(vvtd, ie_addr, ie_data);
+ __vvtd_clear_bit(vvtd, DMAR_IECTL_REG, DMA_IECTL_IP_BIT);
+ }
+ }
+ }
+ break;
+
+ case TYPE_INVAL_IEC:
+ /*
+ * Currently, no cache is preserved in hypervisor. Only need to update
+ * pIRTEs which are modified in binding process.
+ */
+ break;
+
+ default:
+ goto error;
+ }
+
+ unmap_guest_page((void*)qinval_page);
+ return 0;
+
+error:
+ unmap_guest_page((void*)qinval_page);
+ gdprintk(XENLOG_ERR, "Internal error in Queue Invalidation.\n");
+ domain_crash(vvtd->domain);
+ return -1;
+}
+
+/*
+ * Invalidate all the descriptors in Invalidation Queue.
+ */
+static void vvtd_process_iq(struct vvtd *vvtd)
+{
+ uint64_t iqh, iqt, iqa, max_entry, i;
+ int ret = 0;
+
+ /*
+ * No new descriptor is fetched from the Invalidation Queue until
+ * software clears the IQE field in the Fault Status Register
+ */
+ if ( vvtd_test_bit(vvtd, DMAR_FSTS_REG, DMA_FSTS_IQE_BIT) )
+ return;
+
+ vvtd_get_reg_quad(vvtd, DMAR_IQH_REG, iqh);
+ vvtd_get_reg_quad(vvtd, DMAR_IQT_REG, iqt);
+ vvtd_get_reg_quad(vvtd, DMAR_IQA_REG, iqa);
+
+ max_entry = DMA_IQA_ENTRY_PER_PAGE << DMA_IQA_QS(iqa);
+ iqh = DMA_IQH_QH(iqh);
+ iqt = DMA_IQT_QT(iqt);
+
+ ASSERT(iqt < max_entry);
+ if ( iqh == iqt )
+ return;
+
+ i = iqh;
+ while ( i != iqt )
+ {
+ ret = process_iqe(vvtd, i);
+ if ( ret )
+ break;
+ else
+ i = (i + 1) % max_entry;
+ vvtd_set_reg_quad(vvtd, DMAR_IQH_REG, i << DMA_IQH_QH_SHIFT);
+ }
+
+ /*
+ * When IQE set, IQH references the desriptor associated with the error.
+ */
+ if ( ret )
+ vvtd_report_non_recoverable_fault(vvtd, DMA_FSTS_IQE_BIT);
+}
+
+static int vvtd_write_iqt(struct vvtd *vvtd, unsigned long val)
+{
+ uint64_t iqa;
+
+ if ( val & DMA_IQT_RSVD )
+ {
+ VVTD_DEBUG(VVTD_DBG_RW, "Attempt to set reserved bits in "
+ "Invalidation Queue Tail.");
+ return X86EMUL_OKAY;
+ }
+
+ vvtd_get_reg_quad(vvtd, DMAR_IQA_REG, iqa);
+ if ( DMA_IQT_QT(val) >= DMA_IQA_ENTRY_PER_PAGE << DMA_IQA_QS(iqa) )
+ {
+ VVTD_DEBUG(VVTD_DBG_RW, "IQT: Value %lx exceeded supported max "
+ "index.", val);
+ return X86EMUL_OKAY;
+ }
+
+ vvtd_set_reg_quad(vvtd, DMAR_IQT_REG, val);
+ vvtd_process_iq(vvtd);
+ return X86EMUL_OKAY;
+}
+
+static int vvtd_write_iqa(struct vvtd *vvtd, unsigned long val)
+{
+ if ( val & DMA_IQA_RSVD )
+ {
+ VVTD_DEBUG(VVTD_DBG_RW, "Attempt to set reserved bits in "
+ "Invalidation Queue Address.");
+ return X86EMUL_OKAY;
+ }
+
+ vvtd_set_reg_quad(vvtd, DMAR_IQA_REG, val);
+ return X86EMUL_OKAY;
+}
+
+static int vvtd_write_ics(struct vvtd *vvtd, unsigned long val)
+{
+ if ( val & DMA_ICS_IWC )
+ {
+ __vvtd_clear_bit(vvtd, DMAR_ICS_REG, DMA_ICS_IWC_BIT);
+ /*When IWC field is cleared, the IP field needs to be cleared */
+ __vvtd_clear_bit(vvtd, DMAR_IECTL_REG, DMA_IECTL_IP_BIT);
+ }
+ return X86EMUL_OKAY;
+}
+
static int vvtd_write_frcd3(struct vvtd *vvtd, unsigned long val)
{
/* Writing a 1 means clear fault */
@@ -434,6 +613,29 @@ static int vvtd_write_frcd3(struct vvtd *vvtd, unsigned
long val)
return X86EMUL_OKAY;
}
+static int vvtd_write_iectl(struct vvtd *vvtd, unsigned long val)
+{
+ /*
+ * Only DMA_IECTL_IM bit is writable. Generate pending event when unmask.
+ */
+ if ( !(val & DMA_IECTL_IM) )
+ {
+ /* Clear IM and clear IP */
+ __vvtd_clear_bit(vvtd, DMAR_IECTL_REG, DMA_IECTL_IM_BIT);
+ if ( vvtd_test_and_clear_bit(vvtd, DMAR_IECTL_REG, DMA_IECTL_IP_BIT) )
+ {
+ uint32_t ie_data, ie_addr;
+ ie_data = vvtd_get_reg(vvtd, DMAR_IEDATA_REG);
+ ie_addr = vvtd_get_reg(vvtd, DMAR_IEADDR_REG);
+ vvtd_generate_interrupt(vvtd, ie_addr, ie_data);
+ }
+ }
+ else
+ __vvtd_set_bit(vvtd, DMAR_IECTL_REG, DMA_IECTL_IM_BIT);
+
+ return X86EMUL_OKAY;
+}
+
static int vvtd_write_fectl(struct vvtd *vvtd, unsigned long val)
{
/*
@@ -476,6 +678,10 @@ static int vvtd_write_fsts(struct vvtd *vvtd, unsigned
long val)
if ( !((vvtd_get_reg(vvtd, DMAR_FSTS_REG) & DMA_FSTS_FAULTS)) )
__vvtd_clear_bit(vvtd, DMAR_FECTL_REG, DMA_FECTL_IP_BIT);
+ /* Continue to deal invalidation when IQE is clear */
+ if ( !vvtd_test_bit(vvtd, DMAR_FSTS_REG, DMA_FSTS_IQE_BIT) )
+ vvtd_process_iq(vvtd);
+
return X86EMUL_OKAY;
}
@@ -636,6 +842,48 @@ static int vvtd_write(struct vcpu *v, unsigned long addr,
ret = vvtd_write_gcmd(vvtd, val_lo);
break;
+ case DMAR_IQT_REG:
+ if ( len == 8 )
+ ret = vvtd_write_iqt(vvtd, val);
+ else
+ ret = vvtd_write_iqt(vvtd, val_lo);
+ break;
+
+ case DMAR_IQA_REG:
+ if ( len == 8 )
+ ret = vvtd_write_iqa(vvtd, val);
+ else
+ {
+ unsigned long iqa_hi;
+
+ iqa_hi = vvtd_get_reg(vvtd, DMAR_IQA_REG_HI);
+ ret = vvtd_write_iqa(vvtd, val_lo | (iqa_hi << 32));
+ }
+ break;
+
+ case DMAR_IQA_REG_HI:
+ {
+ unsigned long iqa_lo;
+
+ if ( len == 8 )
+ goto error;
+ iqa_lo = vvtd_get_reg(vvtd, DMAR_IQA_REG);
+ ret = vvtd_write_iqa(vvtd, (val_lo << 32) | iqa_lo);
+ break;
+ }
+
+ case DMAR_ICS_REG:
+ if ( len == 8 )
+ goto error;
+ ret = vvtd_write_ics(vvtd, val_lo);
+ break;
+
+ case DMAR_IECTL_REG:
+ if ( len == 8 )
+ goto error;
+ ret = vvtd_write_iectl(vvtd, val_lo);
+ break;
+
case DMAR_IRTA_REG:
if ( len == 8 )
vvtd_set_reg_quad(vvtd, DMAR_IRTA_REG, val);
diff --git a/xen/drivers/passthrough/vtd/iommu.h
b/xen/drivers/passthrough/vtd/iommu.h
index 20e6172..2ee7c5c 100644
--- a/xen/drivers/passthrough/vtd/iommu.h
+++ b/xen/drivers/passthrough/vtd/iommu.h
@@ -207,6 +207,32 @@
#define DMA_IRTA_S(val) (val & 0xf)
#define DMA_IRTA_SIZE(val) (1UL << (DMA_IRTA_S(val) + 1))
+/* IQH_REG */
+#define DMA_IQH_QH_SHIFT 4
+#define DMA_IQH_QH(val) ((val >> 4) & 0x7fffULL)
+
+/* IQT_REG */
+#define DMA_IQT_QT_SHIFT 4
+#define DMA_IQT_QT(val) ((val >> 4) & 0x7fffULL)
+#define DMA_IQT_RSVD 0xfffffffffff80007ULL
+
+/* IQA_REG */
+#define DMA_MGAW 39 /* Maximum Guest Address Width */
+#define DMA_IQA_ADDR(val) (val & ~0xfffULL)
+#define DMA_IQA_QS(val) (val & 0x7)
+#define DMA_IQA_ENTRY_PER_PAGE (1 << 8)
+#define DMA_IQA_RSVD (~((1ULL << DMA_MGAW) -1 ) | 0xff8ULL)
+
+/* IECTL_REG */
+#define DMA_IECTL_IM_BIT 31
+#define DMA_IECTL_IM (1 << DMA_IECTL_IM_BIT)
+#define DMA_IECTL_IP_BIT 30
+#define DMA_IECTL_IP (((u64)1) << DMA_IECTL_IP_BIT)
+
+/* ICS_REG */
+#define DMA_ICS_IWC_BIT 0
+#define DMA_ICS_IWC (1 << DMA_ICS_IWC_BIT)
+
/* PMEN_REG */
#define DMA_PMEN_EPM (((u32)1) << 31)
#define DMA_PMEN_PRS (((u32)1) << 0)
@@ -241,7 +267,8 @@
#define DMA_FSTS_PPF ((u64)1 << DMA_FSTS_PPF_BIT)
#define DMA_FSTS_AFO ((u64)1 << 2)
#define DMA_FSTS_APF ((u64)1 << 3)
-#define DMA_FSTS_IQE ((u64)1 << 4)
+#define DMA_FSTS_IQE_BIT 4
+#define DMA_FSTS_IQE ((u64)1 << DMA_FSTS_IQE_BIT)
#define DMA_FSTS_ICE ((u64)1 << 5)
#define DMA_FSTS_ITE ((u64)1 << 6)
#define DMA_FSTS_PRO_BIT 7
--
1.8.3.1
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
https://lists.xen.org/xen-devel
|
 |
Lists.xenproject.org is hosted with RackSpace, monitoring our |