Re: [PATCH v1] misc: fastrpc: Fix remote heap allocation request

[Ekansh Gupta <quic_ekangupt@xxxxxxxxxxx>]

On 6/12/2023 5:34 PM, Srinivas Kandagatla wrote:
> On 05/06/2023 15:11, Ekansh Gupta wrote:
> > Remote heap is used by DSP audioPD on need basis. This memory is
> > allocated from reserved CMA memory region and is then shared with
> > audioPD to use it for it's functionality.
> >
> > Current implementation of remote heap is not allocating the memory
> > from CMA region, instead it is allocating the memory from SMMU
> > context bank. The arguments passed to scm call for the reassignment
> > of ownership is also not correct. Added changes to allocate CMA
> > memory and have a proper ownership reassignment.
> >
> > Fixes: 532ad70c6d44 ("misc: fastrpc: Add mmap request assigning for 
> > static PD pool")
> > Cc: stable <stable@xxxxxxxxxx>
> > Tested-by: Ekansh Gupta <quic_ekangupt@xxxxxxxxxxx>
> > Signed-off-by: Ekansh Gupta <quic_ekangupt@xxxxxxxxxxx>
> > ---
> >   drivers/misc/fastrpc.c | 26 ++++++++++++++++++++------
> >   1 file changed, 20 insertions(+), 6 deletions(-)
> >
> > diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c
> > index 30d4d04..f5fc2de 100644
> > --- a/drivers/misc/fastrpc.c
> > +++ b/drivers/misc/fastrpc.c
> > @@ -1866,7 +1866,11 @@ static int fastrpc_req_mmap(struct fastrpc_user 
> > *fl, char __user *argp)
> >           return -EINVAL;
> >       }
> > -    err = fastrpc_buf_alloc(fl, fl->sctx->dev, req.size, &buf);
> > +    if (req.flags == ADSP_MMAP_REMOTE_HEAP_ADDR)
> > +        err = fastrpc_remote_heap_alloc(fl, dev, req.size, &buf);
> > +    else
> > +        err = fastrpc_buf_alloc(fl, dev, req.size, &buf);
> > +
> >       if (err) {
> >           dev_err(dev, "failed to allocate buffer\n");
> >           return err;
> > @@ -1905,12 +1909,22 @@ static int fastrpc_req_mmap(struct 
> > fastrpc_user *fl, char __user *argp)
> >       /* Add memory to static PD pool, protection thru hypervisor */
> >       if (req.flags == ADSP_MMAP_REMOTE_HEAP_ADDR && fl->cctx->vmcount) {
> > -        struct qcom_scm_vmperm perm;
> > +        u64 src_perms = BIT(QCOM_SCM_VMID_HLOS);
> > +        struct qcom_scm_vmperm *dst_perms;
> > +        u32 i;
> > -        perm.vmid = QCOM_SCM_VMID_HLOS;
> > -        perm.perm = QCOM_SCM_PERM_RWX;
> > -        err = qcom_scm_assign_mem(buf->phys, buf->size,
> > -            &fl->cctx->perms, &perm, 1);
> > +        dst_perms = kcalloc(fl->cctx->vmcount,
> > +            sizeof(struct qcom_scm_vmperm), GFP_KERNEL);
> > +        if (!dst_perms)
> > +            return -ENOMEM;
> > +        for (i = 0; i < fl->cctx->vmcount; i++) {
> > +            dst_perms[i].vmid = fl->cctx->vmperms[i].vmid;
> > +            dst_perms[i].perm = fl->cctx->vmperms[i].perm;
> > +        }
> > +
> Why not use fl->cctx->vmperms instead of allocating a new dst_perms?
>
> --srini
Thanks Srini for your comments. I'll address them in next patch.
> > +        err = qcom_scm_assign_mem(buf->phys,(u64)buf->size,
> > +            &src_perms, dst_perms, fl->cctx->vmcount);
> > +        kfree(dst_perms);
> >           if (err) {
> >               dev_err(fl->sctx->dev, "Failed to assign memory phys 
> > 0x%llx size 0x%llx err %d",
> >                       buf->phys, buf->size, err);