From: Jonathan Cameron <Jonathan.Cameron@huawei.com>
To: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Cc: linuxarm@huawei.com, qemu-devel@nongnu.org,
"Alex Bennée" <alex.bennee@linaro.org>,
"Marcel Apfelbaum" <marcel@redhat.com>,
"Michael S . Tsirkin" <mst@redhat.com>,
"Igor Mammedov" <imammedo@redhat.com>,
"Markus Armbruster" <armbru@redhat.com>,
linux-cxl@vger.kernel.org,
"Ben Widawsky" <ben.widawsky@intel.com>,
"Peter Maydell" <peter.maydell@linaro.org>,
"Shameerali Kolothum Thodi"
<shameerali.kolothum.thodi@huawei.com>,
"Philippe Mathieu-Daudé" <f4bug@amsat.org>,
"Peter Xu" <peterx@redhat.com>,
"David Hildenbrand" <david@redhat.com>,
"Paolo Bonzini" <pbonzini@redhat.com>,
"Saransh Gupta1" <saransh@ibm.com>,
"Shreyas Shah" <shreyas.shah@elastics.cloud>,
"Chris Browy" <cbrowy@avery-design.com>,
"Samarth Saxena" <samarths@cadence.com>,
"Dan Williams" <dan.j.williams@intel.com>
Subject: Re: [PATCH v8 32/46] mem/cxl_type3: Add read and write functions for associated hostmem.
Date: Wed, 23 Mar 2022 15:43:53 +0000 [thread overview]
Message-ID: <20220323154353.0000617e@huawei.com> (raw)
In-Reply-To: <5e0dc4e5-d002-9fab-2b69-d849aee68acc@ilande.co.uk>
On Sat, 19 Mar 2022 08:53:40 +0000
Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk> wrote:
> On 18/03/2022 15:06, Jonathan Cameron wrote:
>
> > From: Jonathan Cameron <jonathan.cameron@huawei.com>
> >
> > Once a read or write reaches a CXL type 3 device, the HDM decoders
> > on the device are used to establish the Device Physical Address
> > which should be accessed. These functions peform the required maths
> > and then use a device specific address space to access the
> > hostmem->mr to fullfil the actual operation. Note that failed writes
> > are silent, but failed reads return poison. Note this is based
> > loosely on:
> >
> > https://lore.kernel.org/qemu-devel/20200817161853.593247-6-f4bug@amsat.org/
> > [RFC PATCH 0/9] hw/misc: Add support for interleaved memory accesses
> >
> > Only lightly tested so far. More complex test cases yet to be written.
> >
> > Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
> > ---
> > hw/mem/cxl_type3.c | 88 +++++++++++++++++++++++++++++++++++++
> > include/hw/cxl/cxl_device.h | 6 +++
> > 2 files changed, 94 insertions(+)
> >
> > diff --git a/hw/mem/cxl_type3.c b/hw/mem/cxl_type3.c
> > index 244eb5dc91..225155dac5 100644
> > --- a/hw/mem/cxl_type3.c
> > +++ b/hw/mem/cxl_type3.c
> > @@ -100,7 +100,9 @@ static void ct3_finalize(Object *obj)
> >
> > static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp)
> > {
> > + DeviceState *ds = DEVICE(ct3d);
> > MemoryRegion *mr;
> > + g_autofree char *name = NULL;
> >
> > if (!ct3d->hostmem) {
> > error_setg(errp, "memdev property must be set");
> > @@ -115,6 +117,13 @@ static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp)
> > memory_region_set_nonvolatile(mr, true);
> > memory_region_set_enabled(mr, true);
> > host_memory_backend_set_mapped(ct3d->hostmem, true);
>
> There is an existing example for generating names for PCI devices in SPAPR which you
> can borrow which looks something like this (not compile tested!):
>
> static char *cxl_type3_get_id(CXLType3Dev *ct3d)
> {
> uint32_t busnr = pci_bus_num(PCI_BUS(qdev_get_parent_bus(DEVICE(ct3d))));
> PCIDevice *pd = PCI_DEVICE(ct3d);
> DeviceState *ds = DEVICE(ct3d);
>
> if (ds->id) {
> return g_strdup_printf("%s:%02x:%02x.%x", ds->id, busnr,
> PCI_SLOT(pd->devfn), PCI_FUNC(pd->devfn));
> } else {
> return g_strdup_printf("%02x:%02x.%x", busnr,
> PCI_SLOT(pd->devfn), PCI_FUNC(pd->devfn));
> }
> }
The snag is at this point the PCI bus hasn't been enumerated so all of these
numbers are 0 for endpoints. They can change at any point (in theory) so
using them to help identify a device is unstable.
I switched to using a : as the separator as that is clear than
yet another -
>
> > +
> > + if (ds->id) {
> > + name = g_strdup_printf("cxl-type3-dpa-space-%s", ds->id);
> > + } else {
> > + name = g_strdup("cxl-type3-dpa-space");
> > + }
>
> This then becomes:
>
> char *id, *name;
> ...
>
> id = cxl_type3_get_id(ct3d);
> name = g_strdup_printf("cxl-type3-dpa-space:%s", id);
> address_space_init(&ct3d->hostmem_as, mr, name);
> g_free(id);
> g_free(name);
>
> > + address_space_init(&ct3d->hostmem_as, mr, name);
>
> There is an address_space_init() here but no associated address_space_destroy() -
> you'll need to add a ct3_finalize() function to remove the address space, otherwise
> there will be a memory leak when the device is removed because of the dangling reference.
I was lazy on this for two reasons:
a) I could actually figure out to do a finalize for a PCI device. I think after digging
more it's via the pc->exit callback (which is very oddly named - I guess for historic
reasons - why is the unwind of realize called exit? Ah well. I had previously
had it in instance_finalize which resulted in a qtest crash as it would
destroy an address space we hadn't created.
b) Only a tiny percentage of all the address spaces in qemu are ever destroyed (or at least
I couldn't find where they were destroyed). I guess using bad practice elsewhere isn't
the best way to write code :)
Having chased it through it seems to me that what I had in instance_finalize in 23
should have been in pc->exit as it is unwinding stuff done in pc->realize()
not the instance_init.
Thanks,
Jonathan
>
> > ct3d->cxl_dstate.pmem_size = ct3d->hostmem->size;
> >
> > if (!ct3d->lsa) {
> > @@ -160,6 +169,85 @@ static void ct3_realize(PCIDevice *pci_dev, Error **errp)
> > &ct3d->cxl_dstate.device_registers);
> > }
> >
> > +/* TODO: Support multiple HDM decoders and DPA skip */
> > +static bool cxl_type3_dpa(CXLType3Dev *ct3d, hwaddr host_addr, uint64_t *dpa)
> > +{
> > + uint32_t *cache_mem = ct3d->cxl_cstate.crb.cache_mem_registers;
> > + uint64_t decoder_base, decoder_size, hpa_offset;
> > + uint32_t hdm0_ctrl;
> > + int ig, iw;
> > +
> > + decoder_base = (((uint64_t)cache_mem[R_CXL_HDM_DECODER0_BASE_HI] << 32) |
> > + cache_mem[R_CXL_HDM_DECODER0_BASE_LO]);
> > + if ((uint64_t)host_addr < decoder_base) {
> > + return false;
> > + }
> > +
> > + hpa_offset = (uint64_t)host_addr - decoder_base;
> > +
> > + decoder_size = ((uint64_t)cache_mem[R_CXL_HDM_DECODER0_SIZE_HI] << 32) |
> > + cache_mem[R_CXL_HDM_DECODER0_SIZE_LO];
> > + if (hpa_offset >= decoder_size) {
> > + return false;
> > + }
> > +
> > + hdm0_ctrl = cache_mem[R_CXL_HDM_DECODER0_CTRL];
> > + iw = FIELD_EX32(hdm0_ctrl, CXL_HDM_DECODER0_CTRL, IW);
> > + ig = FIELD_EX32(hdm0_ctrl, CXL_HDM_DECODER0_CTRL, IG);
> > +
> > + *dpa = (MAKE_64BIT_MASK(0, 8 + ig) & hpa_offset) |
> > + ((MAKE_64BIT_MASK(8 + ig + iw, 64 - 8 - ig - iw) & hpa_offset) >> iw);
> > +
> > + return true;
> > +}
> > +
> > +MemTxResult cxl_type3_read(PCIDevice *d, hwaddr host_addr, uint64_t *data,
> > + unsigned size, MemTxAttrs attrs)
> > +{
> > + CXLType3Dev *ct3d = CT3(d);
> > + uint64_t dpa_offset;
> > + MemoryRegion *mr;
> > +
> > + /* TODO support volatile region */
> > + mr = host_memory_backend_get_memory(ct3d->hostmem);
> > + if (!mr) {
> > + return MEMTX_ERROR;
> > + }
> > +
> > + if (!cxl_type3_dpa(ct3d, host_addr, &dpa_offset)) {
> > + return MEMTX_ERROR;
> > + }
> > +
> > + if (dpa_offset > int128_get64(mr->size)) {
> > + return MEMTX_ERROR;
> > + }
> > +
> > + return address_space_read(&ct3d->hostmem_as, dpa_offset, attrs, data, size);
> > +}
> > +
> > +MemTxResult cxl_type3_write(PCIDevice *d, hwaddr host_addr, uint64_t data,
> > + unsigned size, MemTxAttrs attrs)
> > +{
> > + CXLType3Dev *ct3d = CT3(d);
> > + uint64_t dpa_offset;
> > + MemoryRegion *mr;
> > +
> > + mr = host_memory_backend_get_memory(ct3d->hostmem);
> > + if (!mr) {
> > + return MEMTX_OK;
> > + }
> > +
> > + if (!cxl_type3_dpa(ct3d, host_addr, &dpa_offset)) {
> > + return MEMTX_OK;
> > + }
> > +
> > + if (dpa_offset > int128_get64(mr->size)) {
> > + return MEMTX_OK;
> > + }
> > + return address_space_write(&ct3d->hostmem_as, dpa_offset, attrs,
> > + &data, size);
> > +}
> > +
> > static void ct3d_reset(DeviceState *dev)
> > {
> > CXLType3Dev *ct3d = CT3(dev);
> > diff --git a/include/hw/cxl/cxl_device.h b/include/hw/cxl/cxl_device.h
> > index 288cc11772..eb998791d7 100644
> > --- a/include/hw/cxl/cxl_device.h
> > +++ b/include/hw/cxl/cxl_device.h
> > @@ -235,6 +235,7 @@ typedef struct cxl_type3_dev {
> > PCIDevice parent_obj;
> >
> > /* Properties */
> > + AddressSpace hostmem_as;
> > uint64_t size;
> > HostMemoryBackend *hostmem;
> > HostMemoryBackend *lsa;
> > @@ -262,4 +263,9 @@ struct CXLType3Class {
> > uint64_t offset);
> > };
> >
> > +MemTxResult cxl_type3_read(PCIDevice *d, hwaddr host_addr, uint64_t *data,
> > + unsigned size, MemTxAttrs attrs);
> > +MemTxResult cxl_type3_write(PCIDevice *d, hwaddr host_addr, uint64_t data,
> > + unsigned size, MemTxAttrs attrs);
> > +
> > #endif
>
>
> ATB,
>
> Mark.
next prev parent reply other threads:[~2022-03-23 15:44 UTC|newest]
Thread overview: 71+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-03-18 15:05 [PATCH v8 00/46] CXl 2.0 emulation Support Jonathan Cameron
2022-03-18 15:05 ` [PATCH v8 01/46] hw/pci/cxl: Add a CXL component type (interface) Jonathan Cameron
2022-03-27 13:32 ` Adam Manzanares
2022-03-18 15:05 ` [PATCH v8 02/46] hw/cxl/component: Introduce CXL components (8.1.x, 8.2.5) Jonathan Cameron
2022-03-25 13:45 ` Jonathan Cameron
2022-03-28 14:28 ` Adam Manzanares
2022-03-30 16:55 ` Jonathan Cameron
2022-03-31 12:20 ` Jonathan Cameron
2022-03-18 15:05 ` [PATCH v8 03/46] MAINTAINERS: Add entry for Compute Express Link Emulation Jonathan Cameron
2022-03-18 15:05 ` [PATCH v8 04/46] hw/cxl/device: Introduce a CXL device (8.2.8) Jonathan Cameron
2022-03-29 18:13 ` Adam Manzanares
2022-03-29 19:53 ` Davidlohr Bueso
2022-03-30 12:15 ` Jonathan Cameron
2022-03-31 21:42 ` Adam Manzanares
2022-03-30 17:48 ` Jonathan Cameron
2022-03-31 22:13 ` Adam Manzanares
2022-04-01 13:30 ` Jonathan Cameron
2022-04-04 15:15 ` Adam Manzanares
2022-04-05 9:10 ` Jonathan Cameron
2022-03-18 15:05 ` [PATCH v8 05/46] hw/cxl/device: Implement the CAP array (8.2.8.1-2) Jonathan Cameron
2022-03-18 15:05 ` [PATCH v8 06/46] hw/cxl/device: Implement basic mailbox (8.2.8.4) Jonathan Cameron
2022-03-18 15:05 ` [PATCH v8 07/46] hw/cxl/device: Add memory device utilities Jonathan Cameron
2022-03-18 15:05 ` [PATCH v8 08/46] hw/cxl/device: Add cheap EVENTS implementation (8.2.9.1) Jonathan Cameron
2022-03-18 15:05 ` [PATCH v8 09/46] hw/cxl/device: Timestamp implementation (8.2.9.3) Jonathan Cameron
2022-03-18 15:05 ` [PATCH v8 10/46] hw/cxl/device: Add log commands (8.2.9.4) + CEL Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 11/46] hw/pxb: Use a type for realizing expanders Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 12/46] hw/pci/cxl: Create a CXL bus type Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 13/46] cxl: Machine level control on whether CXL support is enabled Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 14/46] hw/pxb: Allow creation of a CXL PXB (host bridge) Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 15/46] qtest/cxl: Introduce initial test for pxb-cxl only Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 16/46] hw/cxl/rp: Add a root port Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 17/46] hw/cxl/device: Add a memory device (8.2.8.5) Jonathan Cameron
2022-03-19 8:32 ` Mark Cave-Ayland
2022-03-23 18:18 ` Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 18/46] hw/cxl/device: Implement MMIO HDM decoding (8.2.5.12) Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 19/46] hw/cxl/device: Add some trivial commands Jonathan Cameron
2022-03-18 16:56 ` Alison Schofield
2022-03-23 15:57 ` Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 20/46] hw/cxl/device: Plumb real Label Storage Area (LSA) sizing Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 21/46] hw/cxl/device: Implement get/set Label Storage Area (LSA) Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 22/46] qtests/cxl: Add initial root port and CXL type3 tests Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 23/46] hw/cxl/component: Implement host bridge MMIO (8.2.5, table 142) Jonathan Cameron
2022-03-19 8:35 ` Mark Cave-Ayland
2022-03-23 18:37 ` Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 24/46] acpi/cxl: Add _OSC implementation (9.14.2) Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 25/46] acpi/cxl: Create the CEDT (9.14.1) Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 26/46] hw/cxl/component: Add utils for interleave parameter encoding/decoding Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 27/46] hw/cxl/host: Add support for CXL Fixed Memory Windows Jonathan Cameron
2022-03-28 12:50 ` Markus Armbruster
2022-03-31 12:12 ` Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 28/46] acpi/cxl: Introduce CFMWS structures in CEDT Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 29/46] hw/pci-host/gpex-acpi: Add support for dsdt construction for pxb-cxl Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 30/46] pci/pcie_port: Add pci_find_port_by_pn() Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 31/46] CXL/cxl_component: Add cxl_get_hb_cstate() Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 32/46] mem/cxl_type3: Add read and write functions for associated hostmem Jonathan Cameron
2022-03-19 8:53 ` Mark Cave-Ayland
2022-03-23 15:43 ` Jonathan Cameron [this message]
2022-03-18 15:06 ` [PATCH v8 33/46] cxl/cxl-host: Add memops for CFMWS region Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 34/46] hw/cxl/component Add a dumb HDM decoder handler Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 35/46] i386/pc: Enable CXL fixed memory windows Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 36/46] tests/acpi: q35: Allow addition of a CXL test Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 37/46] qtests/bios-tables-test: Add a test for CXL emulation Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 38/46] tests/acpi: Add tables " Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 39/46] qtest/cxl: Add more complex test cases with CFMWs Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 40/46] hw/arm/virt: Basic CXL enablement on pci_expander_bridge instances pxb-cxl Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 41/46] qtest/cxl: Add aarch64 virt test for CXL Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 42/46] docs/cxl: Add initial Compute eXpress Link (CXL) documentation Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 43/46] pci-bridge/cxl_upstream: Add a CXL switch upstream port Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 44/46] pci-bridge/cxl_downstream: Add a CXL switch downstream port Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 45/46] cxl/cxl-host: Support interleave decoding with one level of switches Jonathan Cameron
2022-03-18 15:06 ` [PATCH v8 46/46] docs/cxl: Add switch documentation Jonathan Cameron
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20220323154353.0000617e@huawei.com \
--to=jonathan.cameron@huawei.com \
--cc=alex.bennee@linaro.org \
--cc=armbru@redhat.com \
--cc=ben.widawsky@intel.com \
--cc=cbrowy@avery-design.com \
--cc=dan.j.williams@intel.com \
--cc=david@redhat.com \
--cc=f4bug@amsat.org \
--cc=imammedo@redhat.com \
--cc=linux-cxl@vger.kernel.org \
--cc=linuxarm@huawei.com \
--cc=marcel@redhat.com \
--cc=mark.cave-ayland@ilande.co.uk \
--cc=mst@redhat.com \
--cc=pbonzini@redhat.com \
--cc=peter.maydell@linaro.org \
--cc=peterx@redhat.com \
--cc=qemu-devel@nongnu.org \
--cc=samarths@cadence.com \
--cc=saransh@ibm.com \
--cc=shameerali.kolothum.thodi@huawei.com \
--cc=shreyas.shah@elastics.cloud \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox