On Thu, Sep 2, 2021 at 11:09 AM Jonathan Cameron <[email protected]> wrote: > > On Tue, 24 Aug 2021 09:07:13 -0700 > Dan Williams <[email protected]> wrote: > > > The CXL_PMEM driver expects exclusive control of the label storage area > > space. Similar to the LIBNVDIMM expectation that the label storage area > > is only writable from userspace when the corresponding memory device is > > not active in any region, the expectation is the native CXL_PCI UAPI > > path is disabled while the cxl_nvdimm for a given cxl_memdev device is > > active in LIBNVDIMM. > > > > Add the ability to toggle the availability of a given command for the > > UAPI path. Use that new capability to shutdown changes to partitions and > > the label storage area while the cxl_nvdimm device is actively proxying > > commands for LIBNVDIMM. > > > > Acked-by: Ben Widawsky <[email protected]> > > Signed-off-by: Dan Williams <[email protected]> > > --- > > drivers/cxl/core/mbox.c | 5 +++++ > > drivers/cxl/core/memdev.c | 31 +++++++++++++++++++++++++++++++ > > drivers/cxl/cxlmem.h | 4 ++++ > > drivers/cxl/pmem.c | 35 ++++++++++++++++++++++++++++------- > > 4 files changed, 68 insertions(+), 7 deletions(-) > > > > diff --git a/drivers/cxl/core/mbox.c b/drivers/cxl/core/mbox.c > > index 73107b302224..6a5c4f3679ba 100644 > > --- a/drivers/cxl/core/mbox.c > > +++ b/drivers/cxl/core/mbox.c > > @@ -230,6 +230,7 @@ static bool cxl_mem_raw_command_allowed(u16 opcode) > > * * %-EINVAL - Reserved fields or invalid values were used. > > * * %-ENOMEM - Input or output buffer wasn't sized properly. > > * * %-EPERM - Attempted to use a protected command. > > + * * %-EBUSY - Kernel has claimed exclusive access to this opcode > > * > > * The result of this command is a fully validated command in @out_cmd > > that is > > * safe to send to the hardware. > > @@ -305,6 +306,10 @@ static int cxl_validate_cmd_from_user(struct cxl_mem > > *cxlm, > > if (!test_bit(info->id, cxlm->enabled_cmds)) > > return -ENOTTY; > > > > + /* Check that the command is not claimed for exclusive kernel use */ > > + if (test_bit(info->id, cxlm->exclusive_cmds)) > > + return -EBUSY; > > + > > /* Check the input buffer is the expected size */ > > if (info->size_in >= 0 && info->size_in != send_cmd->in.size) > > return -ENOMEM; > > > > #endif /* __CXL_MEM_H__ */ > > diff --git a/drivers/cxl/pmem.c b/drivers/cxl/pmem.c > > index 9652c3ee41e7..469b984176a2 100644 > > --- a/drivers/cxl/pmem.c > > +++ b/drivers/cxl/pmem.c > > @@ -16,9 +16,21 @@ > > */ > > static struct workqueue_struct *cxl_pmem_wq; > > > > -static void unregister_nvdimm(void *nvdimm) > > +static __read_mostly DECLARE_BITMAP(exclusive_cmds, > > CXL_MEM_COMMAND_ID_MAX); > > + > > +static void unregister_nvdimm(void *_cxl_nvd) > > { > > - nvdimm_delete(nvdimm); > > + struct cxl_nvdimm *cxl_nvd = _cxl_nvd; > > + struct cxl_memdev *cxlmd = cxl_nvd->cxlmd; > > + struct cxl_mem *cxlm = cxlmd->cxlm; > > + struct device *dev = &cxl_nvd->dev; > > + struct nvdimm *nvdimm; > > + > > + nvdimm = dev_get_drvdata(dev); > > + if (nvdimm) > > + nvdimm_delete(nvdimm); > > + > > + clear_exclusive_cxl_commands(cxlm, exclusive_cmds); > > } > > > > static int match_nvdimm_bridge(struct device *dev, const void *data) > > @@ -39,9 +51,11 @@ static struct cxl_nvdimm_bridge > > *cxl_find_nvdimm_bridge(void) > > static int cxl_nvdimm_probe(struct device *dev) > > { > > struct cxl_nvdimm *cxl_nvd = to_cxl_nvdimm(dev); > > + struct cxl_memdev *cxlmd = cxl_nvd->cxlmd; > > + struct cxl_mem *cxlm = cxlmd->cxlm; > > struct cxl_nvdimm_bridge *cxl_nvb; > > + struct nvdimm *nvdimm = NULL; > > unsigned long flags = 0; > > - struct nvdimm *nvdimm; > > int rc = -ENXIO; > > > > cxl_nvb = cxl_find_nvdimm_bridge(); > > @@ -52,17 +66,20 @@ static int cxl_nvdimm_probe(struct device *dev) > > if (!cxl_nvb->nvdimm_bus) > > goto out; > > > > + set_exclusive_cxl_commands(cxlm, exclusive_cmds); > > + > > set_bit(NDD_LABELING, &flags); > > nvdimm = nvdimm_create(cxl_nvb->nvdimm_bus, cxl_nvd, NULL, flags, 0, > > 0, > > NULL); > > - if (!nvdimm) > > - goto out; > > - > > - rc = devm_add_action_or_reset(dev, unregister_nvdimm, nvdimm); > > + dev_set_drvdata(dev, nvdimm); > > + rc = devm_add_action_or_reset(dev, unregister_nvdimm, cxl_nvd); > > I think this ends up less readable than explicit devm handling of each part > rather than combining them. > > > set_exclusive...() > rc = devm_add_action_or_rset(dev, unset_exclusive, cxlm); > if (rc) > goto out; > > nvidimm = nvdim_create() > if (!nvdimm) //return value looks dubious in old code but I've not > checked it properly. > goto out; > > rc = devm_add_action_or_reset(dev, unregister_nvdimm, nvdimm); > if (rc) > goto out; > dev_set_drvdata(dev, nvdimm); > > and two simpler unwinding functions doing just one thing each.
Taking it a step further, the unwind is so simple here, might as well just have a typical cxl_nvdimm_remove() and drop the devm.
