RE: [PATCH v3 1/4] nitro_enclaves: Merge contiguous physical memory regions

["Longpeng (Mike, Cloud Infrastructure Service Product Dept.)" <longpeng2@huawei.com>]

Hi Andra,

Sorry for the long delay.

I've read all your suggestions in each patch, there's no objection. I'll
send v4 later, please review when you free, thanks.

> -----Original Message-----
> From: Paraschiv, Andra-Irina [mailto:andraprs@amazon.com]
> Sent: Friday, October 15, 2021 9:34 PM
> To: Longpeng (Mike, Cloud Infrastructure Service Product Dept.)
> <longpeng2@huawei.com>
> Cc: Gonglei (Arei) <arei.gonglei@huawei.com>; gregkh@linuxfoundation.org;
> kamal@canonical.com; pbonzini@redhat.com; sgarzare@redhat.com;
> stefanha@redhat.com; vkuznets@redhat.com; linux-kernel@vger.kernel.org;
> ne-devel-upstream@amazon.com; lexnv@amazon.com; alcioa@amazon.com
> Subject: Re: [PATCH v3 1/4] nitro_enclaves: Merge contiguous physical memory
> regions
> 
> 
> 
> On 09/10/2021 04:32, Longpeng(Mike) wrote:
> >
> > From: Longpeng <longpeng2@huawei.com>
> >
> > There can be cases when there are more memory regions that need to be
> > set for an enclave than the maximum supported number of memory regions
> > per enclave. One example can be when the memory regions are backed by 2
> > MiB hugepages (the minimum supported hugepage size).
> >
> > Let's merge the adjacent regions if they are physical contiguous. This
> 
> physical contiguous => physically contiguous
> 
> > way the final number of memory regions is less than before merging and
> > could potentially avoid reaching maximum.
> >
> > Signed-off-by: Longpeng <longpeng2@huawei.com>
> > ---
> > Changes v2 -> v3:
> >    - update the commit title and commit message.  [Andra]
> >    - use 'struct range' to instead of 'struct phys_mem_region'.  [Andra, Greg
> KH]
> >    - add comments before the function definition.  [Andra]
> >    - rename several variables, parameters and function.  [Andra]
> > ---
> >   drivers/virt/nitro_enclaves/ne_misc_dev.c | 79
> +++++++++++++++++++++----------
> >   1 file changed, 55 insertions(+), 24 deletions(-)
> >
> > diff --git a/drivers/virt/nitro_enclaves/ne_misc_dev.c
> b/drivers/virt/nitro_enclaves/ne_misc_dev.c
> > index e21e1e8..eea53e9 100644
> > --- a/drivers/virt/nitro_enclaves/ne_misc_dev.c
> > +++ b/drivers/virt/nitro_enclaves/ne_misc_dev.c
> > @@ -26,6 +26,7 @@
> >   #include <linux/poll.h>
> >   #include <linux/slab.h>
> >   #include <linux/types.h>
> > +#include <linux/range.h>
> 
> Please keep the alphabetical order and move this before "#include
> <linux/slab.h>."
> 
> >   #include <uapi/linux/vm_sockets.h>
> >
> >   #include "ne_misc_dev.h"
> > @@ -126,6 +127,16 @@ struct ne_cpu_pool {
> >   static struct ne_cpu_pool ne_cpu_pool;
> >
> >   /**
> > + * struct phys_contig_mem_regions - Physical contiguous memory regions
> 
> Physical contiguous memory regions => Contiguous physical memory regions
> 
> > + * @num:       The number of regions that currently has.
> > + * @region:    The array of physical memory regions.
> > + */
> > +struct phys_contig_mem_regions {
> > +       unsigned long num;
> > +       struct range region[0];
> 
> Should use "struct range *regions" instead, to keep a similar style with
> regard to arrays throughout the code.
> 
> Please align the fields name (e.g. [1]) so that it's easier to
> distinguish between fields type and name.
> 
> Let's also prefix with "ne" the data structure naming e.g.
> ne_phys_contig_mem_regions. So that is similar to other data structures
> defined in the NE kernel driver codebase.
> 
> [1]
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/dri
> vers/virt/nitro_enclaves/ne_misc_dev.c#n118
> 
> > +};
> > +
> > +/**
> >    * ne_check_enclaves_created() - Verify if at least one enclave has been
> created.
> >    * @void:      No parameters provided.
> >    *
> > @@ -825,6 +836,33 @@ static int ne_sanity_check_user_mem_region_page(struct
> ne_enclave *ne_enclave,
> >   }
> >
> >   /**
> > + * ne_merge_phys_contig_memory_regions() - Add a memory region and merge the
> adjacent
> > + *                                         regions if they are physical
> contiguous.
> 
> physical contiguous => physically contiguous
> 
> > + * @regions   : Private data associated with the physical contiguous memory
> regions.
> 
> physical contiguous memory regions => contiguous physical memory regions
> 
> @regions => @phys_contig_regions
> 
> > + * @page_paddr: Physical start address of the region to be added.
> > + * @page_size : Length of the region to be added.
> > + *
> > + * Return:
> > + * * No return value.
> 
> Please add "Context: Process context. This function is called with the
> ne_enclave mutex held." before "Return", similar to other defined functions.
> 
> Can remove these lines, as for now there is no return value:
> 
> * Return:
> * * No return value.
> 
> And then can add this in the second patch in the series:
> 
> * Context: Process context. This function is called with the ne_enclave
> mutex held. (note: already available in the first patch)
> * Return: (note: added in the second patch)
> * * 0 ...
> 
> > + */
> > +static void
> > +ne_merge_phys_contig_memory_regions(struct phys_contig_mem_regions
> *regions,
> > +                                   u64 page_paddr, u64 page_size)
> 
> phys_contig_mem_regions => ne_phys_contig_mem_regions
> *regions => *phys_contig_regions
> 
> Can define something like this:
> 
> unsigned long num = phys_contig_regions->num;
> 
> and then use it inside the function, except the last line that
> increments the num.
> 
> > +{
> > +       /* Physical contiguous, just merge */
> 
> Physical contiguous => Physically contiguous
> 
> > +       if (regions->num &&
> > +           (regions->region[regions->num - 1].end + 1) == page_paddr) {
> 
> e.g. phys_contig_regions->regions[num - 1]
> 
> > +               regions->region[regions->num - 1].end += page_size;
> > +
> > +               return;
> > +       }
> > +
> > +       regions->region[regions->num].start = page_paddr;
> > +       regions->region[regions->num].end = page_paddr + page_size - 1;
> > +       regions->num++;
> > +}
> > +
> > +/**
> >    * ne_set_user_memory_region_ioctl() - Add user space memory region to the
> slot
> >    *                                    associated with the current enclave.
> >    * @ne_enclave :       Private data associated with the current enclave.
> > @@ -843,9 +881,9 @@ static int ne_set_user_memory_region_ioctl(struct
> ne_enclave *ne_enclave,
> >          unsigned long max_nr_pages = 0;
> >          unsigned long memory_size = 0;
> >          struct ne_mem_region *ne_mem_region = NULL;
> > -       unsigned long nr_phys_contig_mem_regions = 0;
> >          struct pci_dev *pdev = ne_devs.ne_pci_dev->pdev;
> > -       struct page **phys_contig_mem_regions = NULL;
> > +       struct phys_contig_mem_regions *phys_contig_mem_regions = NULL;
> 
> struct phys_contig_mem_regions *phys_contig_mem_regions = NULL; =>
> struct ne_phys_contig_mem_regions phys_contig_mem_regions = {};
> 
> > +       size_t size_to_alloc = 0;
> >          int rc = -EINVAL;
> >
> >          rc = ne_sanity_check_user_mem_region(ne_enclave, mem_region);
> > @@ -866,8 +904,9 @@ static int ne_set_user_memory_region_ioctl(struct
> ne_enclave *ne_enclave,
> >                  goto free_mem_region;
> >          }
> >
> > -       phys_contig_mem_regions = kcalloc(max_nr_pages,
> sizeof(*phys_contig_mem_regions),
> > -                                         GFP_KERNEL);
> > +       size_to_alloc = sizeof(*phys_contig_mem_regions) +
> > +                       max_nr_pages * sizeof(struct range);
> > +       phys_contig_mem_regions = kzalloc(size_to_alloc, GFP_KERNEL);
> 
> Then can alloc memory for the regions field.
> 
> phys_contig_mem_regions.regions = kcalloc(max_nr_pages,
> sizeof(*phys_contig_mem_regions.regions), GFP_KERNEL);
> 
> The "size_of_alloc" will not be needed in this case.
> 
> >          if (!phys_contig_mem_regions) {
> >                  rc = -ENOMEM;
> >
> > @@ -901,26 +940,16 @@ static int ne_set_user_memory_region_ioctl(struct
> ne_enclave *ne_enclave,
> >                  if (rc < 0)
> >                          goto put_pages;
> >
> > -               /*
> > -                * TODO: Update once handled non-contiguous memory regions
> > -                * received from user space or contiguous physical memory regions
> > -                * larger than 2 MiB e.g. 8 MiB.
> > -                */
> > -               phys_contig_mem_regions[i] = ne_mem_region->pages[i];
> > +
> ne_merge_phys_contig_memory_regions(phys_contig_mem_regions,
> 
> And can pass it here like this: &phys_contig_mem_regions.
> 
> > +
> page_to_phys(ne_mem_region->pages[i]),
> > +
> page_size(ne_mem_region->pages[i]));
> >
> >                  memory_size += page_size(ne_mem_region->pages[i]);
> >
> >                  ne_mem_region->nr_pages++;
> >          } while (memory_size < mem_region.memory_size);
> >
> > -       /*
> > -        * TODO: Update once handled non-contiguous memory regions received
> > -        * from user space or contiguous physical memory regions larger than
> > -        * 2 MiB e.g. 8 MiB.
> > -        */
> > -       nr_phys_contig_mem_regions = ne_mem_region->nr_pages;
> > -
> > -       if ((ne_enclave->nr_mem_regions + nr_phys_contig_mem_regions) >
> > +       if ((ne_enclave->nr_mem_regions + phys_contig_mem_regions->num) >
> >              ne_enclave->max_mem_regions) {
> >                  dev_err_ratelimited(ne_misc_dev.this_device,
> >                                      "Reached max memory regions %lld\n",
> > @@ -931,9 +960,10 @@ static int ne_set_user_memory_region_ioctl(struct
> ne_enclave *ne_enclave,
> >                  goto put_pages;
> >          }
> >
> > -       for (i = 0; i < nr_phys_contig_mem_regions; i++) {
> > -               u64 phys_region_addr =
> page_to_phys(phys_contig_mem_regions[i]);
> > -               u64 phys_region_size = page_size(phys_contig_mem_regions[i]);
> > +       for (i = 0; i < phys_contig_mem_regions->num; i++) {
> > +               struct range *range = phys_contig_mem_regions->region + i;
> > +               u64 phys_region_addr = range->start;
> > +               u64 phys_region_size = range_len(range);
> 
> With the updated data structure field, could be:
> 
> u64 phys_region_addr = phys_contig_mem_regions.regions[i].start;
> u64 phys_region_size = range_len(&phys_contig_mem_regions.regions[i]);
> 
> >
> >                  if (phys_region_size & (NE_MIN_MEM_REGION_SIZE - 1)) {
> >                          dev_err_ratelimited(ne_misc_dev.this_device,
> > @@ -959,13 +989,14 @@ static int ne_set_user_memory_region_ioctl(struct
> ne_enclave *ne_enclave,
> >
> >          list_add(&ne_mem_region->mem_region_list_entry,
> &ne_enclave->mem_regions_list);
> >
> > -       for (i = 0; i < nr_phys_contig_mem_regions; i++) {
> > +       for (i = 0; i < phys_contig_mem_regions->num; i++) {
> >                  struct ne_pci_dev_cmd_reply cmd_reply = {};
> >                  struct slot_add_mem_req slot_add_mem_req = {};
> > +               struct range *range = phys_contig_mem_regions->region + i;
> >
> >                  slot_add_mem_req.slot_uid = ne_enclave->slot_uid;
> > -               slot_add_mem_req.paddr =
> page_to_phys(phys_contig_mem_regions[i]);
> > -               slot_add_mem_req.size =
> page_size(phys_contig_mem_regions[i]);
> > +               slot_add_mem_req.paddr = range->start;
> > +               slot_add_mem_req.size = range_len(range);
> 
> Similarly here:
> 
> slot_add_mem_req.paddr = phys_contig_mem_regions.regions[i].start;
> slot_add_mem_req.size = range_len(&phys_contig_mem_regions.regions[i]);
> 
> Then remain the kfree paths for "phys_contig_mem_regions.regions"
> instead of "phys_contig_mem_regions".
> 
> Thanks,
> Andra
> 
> >
> >                  rc = ne_do_request(pdev, SLOT_ADD_MEM,
> >                                     &slot_add_mem_req,
> sizeof(slot_add_mem_req),
> > --
> > 1.8.3.1
> >
> 
> 
> 
> Amazon Development Center (Romania) S.R.L. registered office: 27A Sf. Lazar
> Street, UBC5, floor 2, Iasi, Iasi County, 700045, Romania. Registered in Romania.
> Registration number J22/2621/2005.