Re: [PATCH 1/3] memory: add a sparse memory device

[Alexander Bulekov]

On 210311 1525, Philippe Mathieu-Daudé wrote:
> On 3/11/21 6:36 AM, Alexander Bulekov wrote:
> > For testing, it can be useful to simulate an enormous amount of memory
> > (e.g. 2^64 RAM). This adds an MMIO device that acts as sparse memory.
> > When something writes a nonzero value to a sparse-mem address, we
> > allocate a block of memory. This block is kept around, until all of the
> > bytes within the block are zero-ed. The device has a very low priority
> > (so it can be mapped beneath actual RAM, and virtual device MMIO
> > regions).
> 
> I'm not convinced we need this, but still added some comments while
> reviewing.
> 
> > 
> > Signed-off-by: Alexander Bulekov <alxndr@bu.edu>
> > ---
> >  MAINTAINERS         |   1 +
> >  hw/mem/meson.build  |   1 +
> >  hw/mem/sparse-mem.c | 154 ++++++++++++++++++++++++++++++++++++++++++++
> >  3 files changed, 156 insertions(+)
> >  create mode 100644 hw/mem/sparse-mem.c
> > 
> > diff --git a/MAINTAINERS b/MAINTAINERS
> > index f22d83c178..9e3d8b1401 100644
> > --- a/MAINTAINERS
> > +++ b/MAINTAINERS
> > @@ -2618,6 +2618,7 @@ R: Thomas Huth <thuth@redhat.com>
> >  S: Maintained
> >  F: tests/qtest/fuzz/
> >  F: scripts/oss-fuzz/
> > +F: hw/mem/sparse-mem.c
> >  F: docs/devel/fuzzing.rst
> >  
> >  Register API
> > diff --git a/hw/mem/meson.build b/hw/mem/meson.build
> > index 0d22f2b572..732f459e0a 100644
> > --- a/hw/mem/meson.build
> > +++ b/hw/mem/meson.build
> > @@ -1,5 +1,6 @@
> >  mem_ss = ss.source_set()
> >  mem_ss.add(files('memory-device.c'))
> > +mem_ss.add(files('sparse-mem.c'))
> >  mem_ss.add(when: 'CONFIG_DIMM', if_true: files('pc-dimm.c'))
> >  mem_ss.add(when: 'CONFIG_NPCM7XX', if_true: files('npcm7xx_mc.c'))
> >  mem_ss.add(when: 'CONFIG_NVDIMM', if_true: files('nvdimm.c'))
> > diff --git a/hw/mem/sparse-mem.c b/hw/mem/sparse-mem.c
> > new file mode 100644
> > index 0000000000..ffda6f76b4
> > --- /dev/null
> > +++ b/hw/mem/sparse-mem.c
> > @@ -0,0 +1,154 @@
> > +/*
> > + * A sparse memory device
> > + *
> > + * Copyright Red Hat Inc., 2021
> > + *
> > + * Authors:
> > + *  Alexander Bulekov   <alxndr@bu.edu>
> > + *
> > + * This work is licensed under the terms of the GNU GPL, version 2 or 
> > later.
> > + * See the COPYING file in the top-level directory.
> > + */
> > +
> > +#include "qemu/osdep.h"
> > +
> > +#include "exec/address-spaces.h"
> > +#include "hw/qdev-properties.h"
> > +
> > +#define TYPE_SPARSE_MEM "sparse-mem"
> > +#define SPARSE_MEM(obj) OBJECT_CHECK(SparseMemState, (obj), 
> > TYPE_SPARSE_MEM)
> > +
> > +#define SPARSE_BLOCK_SIZE 0x1000
> > +
> > +typedef struct SparseMemState {
> > +    DeviceState parent_obj;
> > +    MemoryRegion mmio;
> > +    uint64_t baseaddr;
> > +    uint64_t length;
> > +    uint64_t usage;
> 
> usage -> size_used?
> 

Ok - that's nicer.

> > +    uint64_t maxsize;
> > +    GHashTable *mapped;
> > +} SparseMemState;
> > +
> > +typedef struct sparse_mem_block {
> > +    uint16_t nonzeros;
> > +    uint8_t data[SPARSE_BLOCK_SIZE];
> > +} sparse_mem_block;
> > +
> 
> > +static const MemoryRegionOps sparse_mem_ops = {
> > +    .read = sparse_mem_read,
> > +    .write = sparse_mem_write,
> > +    .endianness = DEVICE_LITTLE_ENDIAN,
> > +    .valid = {
> > +            .min_access_size = 1,
> > +            .max_access_size = 8,
> > +            .unaligned = false,
> 
> Why restrict unaligned accesses?
> 

It is mostly a shortcut to avoid dealing with accesses that span
multiple "blocks". E.g. a read from (uint32_t*)0x1ffe would require
looking both at the 0x1000 and 0x2000 blocks.

> > +        },
> > +};
> > +
> > +static Property sparse_mem_properties[] = {
> > +    /* The base address of the memory */
> > +    DEFINE_PROP_UINT64("baseaddr", SparseMemState, baseaddr, 0x0),
> > +    /* The length of the sparse memory region */
> > +    DEFINE_PROP_UINT64("length", SparseMemState, length, UINT64_MAX),
> > +    /* Max amount of actual memory that can be used to back the sparse 
> > memory */
> > +    DEFINE_PROP_UINT64("maxsize", SparseMemState, maxsize, 0x100000),
> 
> 0x100000 -> 1 * MiB
> 

Ok.

> > +    DEFINE_PROP_END_OF_LIST(),
> > +};
> > +
> > +static void sparse_mem_realize(DeviceState *dev, Error **errp)
> > +{
> > +    SparseMemState *s = SPARSE_MEM(dev);
> 
> Anyhow, we should restrict this device to QTest accelerator, right?
> 
> Maybe:
>     if (!qtest_enabled()) {
>         error_setg(errp, "sparse_mem device requires QTest");
>         return;
>     }
> 
> > +
> > +    assert(s->baseaddr + s->length > s->baseaddr);
> 
> Don't you need more than 64-bit to do this check?

The check is to make sure that baseaddr + length doesn't overflow the
64-bit address-space.

> 
> > +
> > +    s->mapped = g_hash_table_new(NULL, NULL);
> > +    memory_region_init_io(&(s->mmio), OBJECT(s), &sparse_mem_ops, s,
> > +                          "sparse-mem", s->length);
> > +    memory_region_add_subregion_overlap(get_system_memory(), s->baseaddr,
> > +                                        &(s->mmio), -100);
> 
> mr_add() to sysmem from a non-sysbus device is odd... Maybe it is
> acceptable, I don't know enough.
> 

I will try to find a more standard way to do this.
Thanks
-Alex

> > +}
> > +
>