On Tue, Dec 20, 2016 at 09:42:46AM -0500, Konrad Rzeszutek Wilk wrote:
> On Mon, Dec 19, 2016 at 03:16:44PM +0100, Juergen Gross wrote:
> > On 19/12/16 13:29, Chris Wilson wrote:
> > > On Mon, Dec 19, 2016 at 12:39:16PM +0100, Juergen Gross wrote:
> > >> With recent 4.10 kernel the graphics isn't coming up under Xen. First
> > >> failure message is:
> > >>
> > >> [ 46.656649] i915 0000:00:02.0: swiotlb buffer is full (sz: 1630208
> > >> bytes)
> > >
> > > Do we get a silent failure? i915_gem_gtt_prepare_pages() is where we
> > > call dma_map_sg() and pass the sg to swiotlb (in this case) for
> > > remapping, and we do check for an error value of 0. After that error,
> > > SWIOTLB_MAP_ERROR is propagated back and converted to 0 for
> > > dma_map_sg(). That looks valid, and we should report ENOMEM back to the
> > > caller.
> > >
> > >> Later I see splats like:
> > >>
> > >> [ 49.393583] general protection fault: 0000 [#1] SMP
> > >
> > > What was the faulting address? RAX is particularly non-pointer-like so I
> > > wonder if we walked onto an uninitialised portion of the sgtable. We may
> > > have tripped over a bug in our sg_page iterator.
> >
> > During the bisect process there have been either GP or NULL pointer
> > dereferences or other page faults. Typical addresses where:
> >
> > xen_swiotlb_unmap_sg_attrs+0x1f/0x50: access to 0000000000000018
> > xen_swiotlb_unmap_sg_attrs+0x1f/0x50: access to 0000000003020118
> >
> > >
> > > The attached patch should prevent an early ENOMEM following the swiotlb
> > > allocation failure. But I suspect that we will still be tripping up the
> > > failure in the sg walker when binding to the GPU.
> > > -Chris
> > >
> >
> > The patch is working not too bad. :-)
> >
> > Still several "swiotlb buffer is full" messages (some with sz:, most
> > without), but no faults any more (neither GP nor NULL pointer
> > dereference). Graphical login is working now.
>
>
> I think I know why. The optimization that was added assumes that
> bus addresses is the same as physical address. Hence it packs all
> of the virtual addresses in the sg, and hands it off to SWIOTLB
> which walks each one and realizes that it has to use the bounce
> buffer.
>
> I am wondering if would make sense to pull 'swiotlb_max_size' inside
> of SWIOTLB and make it an library-ish - so Xen-SWIOTLB can register
> as well and report say that it can only provide one page
> (unless it is running under baremtal).
>
> Or make the usage of 'max_segement' and 'page_to_pfn(page) != last_pfn + 1'
> in i915_gem_object_Get_pages_gtt use something similar to
> xen_biovec_phys_mergeable?
Chris,
I was thinking of something like this (which Juergen has already tested).
>From f196f1294fd25f1402c3dd1231babb8d7f5db2e7 Mon Sep 17 00:00:00 2001
From: Konrad Rzeszutek Wilk <konrad.w...@oracle.com>
Date: Tue, 20 Dec 2016 10:02:02 -0500
Subject: [PATCH] swiotlb: Export swiotlb_max_segment to users
So they can figure out what is the optimal number of pages
that can be contingously stitched together without fear of
bounce buffer.
We also expose an mechanism for sub-users of SWIOTLB API, such
as Xen-SWIOTLB to set the max segment value. And lastly
if swiotlb=force is set (which mandates we bounce buffer everything)
we set max_segment so at least we can bounce buffer one 4K page
instead of a giant 512KB one for which we may not have space.
Reported-and-Tested-by: Juergen Gross <jgr...@suse.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.w...@oracle.com>
---
drivers/gpu/drm/i915/i915_gem.c | 11 +----------
drivers/xen/swiotlb-xen.c | 4 ++++
include/linux/swiotlb.h | 3 +++
lib/swiotlb.c | 26 ++++++++++++++++++++++++++
4 files changed, 34 insertions(+), 10 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index d0dcaf3..115fa39 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -2290,15 +2290,6 @@ void __i915_gem_object_put_pages(struct
drm_i915_gem_object *obj,
mutex_unlock(&obj->mm.lock);
}
-static unsigned int swiotlb_max_size(void)
-{
-#if IS_ENABLED(CONFIG_SWIOTLB)
- return rounddown(swiotlb_nr_tbl() << IO_TLB_SHIFT, PAGE_SIZE);
-#else
- return 0;
-#endif
-}
-
static void i915_sg_trim(struct sg_table *orig_st)
{
struct sg_table new_st;
@@ -2345,7 +2336,7 @@ i915_gem_object_get_pages_gtt(struct drm_i915_gem_object
*obj)
GEM_BUG_ON(obj->base.read_domains & I915_GEM_GPU_DOMAINS);
GEM_BUG_ON(obj->base.write_domain & I915_GEM_GPU_DOMAINS);
- max_segment = swiotlb_max_size();
+ max_segment = swiotlb_max_segment();
if (!max_segment)
max_segment = rounddown(UINT_MAX, PAGE_SIZE);
diff --git a/drivers/xen/swiotlb-xen.c b/drivers/xen/swiotlb-xen.c
index aba1200..f905d6e 100644
--- a/drivers/xen/swiotlb-xen.c
+++ b/drivers/xen/swiotlb-xen.c
@@ -275,6 +275,10 @@ int __ref xen_swiotlb_init(int verbose, bool early)
rc = 0;
} else
rc = swiotlb_late_init_with_tbl(xen_io_tlb_start,
xen_io_tlb_nslabs);
+
+ if (!rc)
+ swiotlb_set_max_segment(PAGE_SIZE);
+
return rc;
error:
if (repeat--) {
diff --git a/include/linux/swiotlb.h b/include/linux/swiotlb.h
index d9c84a9..d7eacd8 100644
--- a/include/linux/swiotlb.h
+++ b/include/linux/swiotlb.h
@@ -114,11 +114,14 @@ swiotlb_dma_supported(struct device *hwdev, u64 mask);
#ifdef CONFIG_SWIOTLB
extern void __init swiotlb_free(void);
+unsigned int swiotlb_max_segment(void);
#else
static inline void swiotlb_free(void) { }
+unsigned int swiotlb_max_segment(void) { return 0; }
#endif
extern void swiotlb_print_info(void);
extern int is_swiotlb_buffer(phys_addr_t paddr);
+extern void swiotlb_set_max_segment(unsigned int);
#endif /* __LINUX_SWIOTLB_H */
diff --git a/lib/swiotlb.c b/lib/swiotlb.c
index 9def738..975b8fc 100644
--- a/lib/swiotlb.c
+++ b/lib/swiotlb.c
@@ -83,6 +83,12 @@ static unsigned int *io_tlb_list;
static unsigned int io_tlb_index;
/*
+ * Max segment that we can provide which (if pages are contingous) will
+ * not be bounced (unless SWIOTLB_FORCE is set).
+ */
+unsigned int max_segment;
+
+/*
* We need to save away the original address corresponding to a mapped entry
* for the sync operations.
*/
@@ -124,6 +130,20 @@ unsigned long swiotlb_nr_tbl(void)
}
EXPORT_SYMBOL_GPL(swiotlb_nr_tbl);
+unsigned int swiotlb_max_segment(void)
+{
+ return max_segment;
+}
+EXPORT_SYMBOL_GPL(swiotlb_max_segment);
+
+void swiotlb_set_max_segment(unsigned int val)
+{
+ if (swiotlb_force == SWIOTLB_FORCE)
+ max_segment = 1;
+ else
+ max_segment = rounddown(val, PAGE_SIZE);
+}
+
/* default to 64MB */
#define IO_TLB_DEFAULT_SIZE (64UL<<20)
unsigned long swiotlb_size_or_default(void)
@@ -205,6 +225,7 @@ int __init swiotlb_init_with_tbl(char *tlb, unsigned long
nslabs, int verbose)
if (verbose)
swiotlb_print_info();
+ swiotlb_set_max_segment(io_tlb_nslabs << IO_TLB_SHIFT);
return 0;
}
@@ -283,6 +304,7 @@ swiotlb_late_init_with_default_size(size_t default_size)
rc = swiotlb_late_init_with_tbl(vstart, io_tlb_nslabs);
if (rc)
free_pages((unsigned long)vstart, order);
+
return rc;
}
@@ -337,6 +359,8 @@ swiotlb_late_init_with_tbl(char *tlb, unsigned long nslabs)
late_alloc = 1;
+ swiotlb_set_max_segment(io_tlb_nslabs << IO_TLB_SHIFT);
+
return 0;
cleanup4:
@@ -351,6 +375,7 @@ swiotlb_late_init_with_tbl(char *tlb, unsigned long nslabs)
io_tlb_end = 0;
io_tlb_start = 0;
io_tlb_nslabs = 0;
+ max_segment = 0;
return -ENOMEM;
}
@@ -379,6 +404,7 @@ void __init swiotlb_free(void)
PAGE_ALIGN(io_tlb_nslabs << IO_TLB_SHIFT));
}
io_tlb_nslabs = 0;
+ max_segment = 0;
}
int is_swiotlb_buffer(phys_addr_t paddr)
--
2.7.4
