| Message ID | 20240703065200.1438145-1-quic_ekangupt@quicinc.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [v1] misc: fastrpc: Add support for multiple PD from one process | expand |
On Wed, Jul 03, 2024 at 12:22:00PM +0530, Ekansh Gupta wrote: > @@ -268,6 +272,7 @@ struct fastrpc_channel_ctx { > struct fastrpc_session_ctx session[FASTRPC_MAX_SESSIONS]; > spinlock_t lock; > struct idr ctx_idr; > + struct ida dsp_pgid_ida; You have an idr and an ida? Why two different types for the same driver? > struct list_head users; > struct kref refcount; > /* Flag if dsp attributes are cached */ > @@ -299,6 +304,7 @@ struct fastrpc_user { > struct fastrpc_buf *init_mem; > > int tgid; > + int dsp_pgid; Are you sure this fits in an int? > +static int fastrpc_pgid_alloc(struct fastrpc_channel_ctx *cctx) > +{ > + int ret = -1; No need to initialize this. > + > + /* allocate unique id between MIN_FRPC_PGID and MAX_FRPC_PGID */ > + ret = ida_alloc_range(&cctx->dsp_pgid_ida, MIN_FRPC_PGID, > + MAX_FRPC_PGID, GFP_ATOMIC); > + if (ret < 0) > + return -1; Why is -1 a specific value here? Return a real error please. Or return 0 if that's not allowed. v > + > + return ret; > +} > + > static int fastrpc_device_open(struct inode *inode, struct file *filp) > { > struct fastrpc_channel_ctx *cctx; > @@ -1582,6 +1605,12 @@ static int fastrpc_device_open(struct inode *inode, struct file *filp) > fl->cctx = cctx; > fl->is_secure_dev = fdevice->secure; > > + fl->dsp_pgid = fastrpc_pgid_alloc(cctx); > + if (fl->dsp_pgid == -1) { > + dev_dbg(&cctx->rpdev->dev, "too many fastrpc clients, max %u allowed\n", MAX_DSP_PD); > + return -EUSERS; Why -EUSERS? And you obviously did not test this as you just leaked memory :( thanks, greg k-h
On Wed, Jul 03, 2024 at 12:22:00PM GMT, Ekansh Gupta wrote: > Memory intensive applications(which requires more tha 4GB) that wants > to offload tasks to DSP might have to split the tasks to multiple > user PD to make the resources available. For every call to DSP, > fastrpc driver passes the process tgid which works as an identifier > for the DSP to enqueue the tasks to specific PD. With current design, > if any process opens device node more than once and makes PD init > request, same tgid will be passed to DSP which will be considered a > bad request and this will result in failure as the same identifier > cannot be used for multiple DSP PD. Allocate and pass an effective > pgid to DSP which would be allocated during device open and will have > a lifetime till the device is closed. This will allow the same process > to open the device more than once and spawn multiple dynamic PD for > ease of processing. Consider adding line breaks to split the text into paragraphs to make it easier to read. The patch also raises the question whether the identifier should be unique per DSP or per the channel. > > Signed-off-by: Ekansh Gupta <quic_ekangupt@quicinc.com> > --- > drivers/misc/fastrpc.c | 48 ++++++++++++++++++++++++++++++++++-------- > 1 file changed, 39 insertions(+), 9 deletions(-) > > diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c > index 5204fda51da3..7250e30aa93f 100644 > --- a/drivers/misc/fastrpc.c > +++ b/drivers/misc/fastrpc.c > @@ -105,6 +105,10 @@ > > #define miscdev_to_fdevice(d) container_of(d, struct fastrpc_device, miscdev) > > +#define MAX_DSP_PD 64 > +#define MIN_FRPC_PGID 1000 > +#define MAX_FRPC_PGID (MIN_FRPC_PGID + MAX_DSP_PD) > + > static const char *domains[FASTRPC_DEV_MAX] = { "adsp", "mdsp", > "sdsp", "cdsp"}; > struct fastrpc_phy_page { > @@ -268,6 +272,7 @@ struct fastrpc_channel_ctx { > struct fastrpc_session_ctx session[FASTRPC_MAX_SESSIONS]; > spinlock_t lock; > struct idr ctx_idr; > + struct ida dsp_pgid_ida; > struct list_head users; > struct kref refcount; > /* Flag if dsp attributes are cached */ > @@ -299,6 +304,7 @@ struct fastrpc_user { > struct fastrpc_buf *init_mem; > > int tgid; Is it still used? What for? > + int dsp_pgid; unsigned int > int pd; > bool is_secure_dev; > /* Lock for lists */ > @@ -462,6 +468,7 @@ static void fastrpc_channel_ctx_free(struct kref *ref) > struct fastrpc_channel_ctx *cctx; > > cctx = container_of(ref, struct fastrpc_channel_ctx, refcount); > + ida_destroy(&cctx->dsp_pgid_ida); > > kfree(cctx); > } > @@ -1114,7 +1121,7 @@ static int fastrpc_invoke_send(struct fastrpc_session_ctx *sctx, > int ret; > > cctx = fl->cctx; > - msg->pid = fl->tgid; > + msg->pid = fl->dsp_pgid; > msg->tid = current->pid; > > if (kernel) > @@ -1292,7 +1299,7 @@ static int fastrpc_init_create_static_process(struct fastrpc_user *fl, > } > } > > - inbuf.pgid = fl->tgid; > + inbuf.pgid = fl->dsp_pgid; > inbuf.namelen = init.namelen; > inbuf.pageslen = 0; > fl->pd = USER_PD; > @@ -1394,7 +1401,7 @@ static int fastrpc_init_create_process(struct fastrpc_user *fl, > goto err; > } > > - inbuf.pgid = fl->tgid; > + inbuf.pgid = fl->dsp_pgid; > inbuf.namelen = strlen(current->comm) + 1; > inbuf.filelen = init.filelen; > inbuf.pageslen = 1; > @@ -1503,7 +1510,7 @@ static int fastrpc_release_current_dsp_process(struct fastrpc_user *fl) > int tgid = 0; > u32 sc; > > - tgid = fl->tgid; > + tgid = fl->dsp_pgid; > args[0].ptr = (u64)(uintptr_t) &tgid; > args[0].length = sizeof(tgid); > args[0].fd = -1; > @@ -1528,6 +1535,9 @@ static int fastrpc_device_release(struct inode *inode, struct file *file) > list_del(&fl->user); > spin_unlock_irqrestore(&cctx->lock, flags); > > + if (fl->dsp_pgid != -1) Why -1? On which path can it be left uninitialized? > + ida_free(&cctx->dsp_pgid_ida, fl->dsp_pgid); > + > if (fl->init_mem) > fastrpc_buf_free(fl->init_mem); > > @@ -1554,6 +1564,19 @@ static int fastrpc_device_release(struct inode *inode, struct file *file) > return 0; > } > > +static int fastrpc_pgid_alloc(struct fastrpc_channel_ctx *cctx) > +{ > + int ret = -1; There is no need to init it, is there? > + > + /* allocate unique id between MIN_FRPC_PGID and MAX_FRPC_PGID */ > + ret = ida_alloc_range(&cctx->dsp_pgid_ida, MIN_FRPC_PGID, > + MAX_FRPC_PGID, GFP_ATOMIC); > + if (ret < 0) > + return -1; Don't throw away error codes. Pass them as is. > + > + return ret; > +} > + > static int fastrpc_device_open(struct inode *inode, struct file *filp) > { > struct fastrpc_channel_ctx *cctx; > @@ -1582,6 +1605,12 @@ static int fastrpc_device_open(struct inode *inode, struct file *filp) > fl->cctx = cctx; > fl->is_secure_dev = fdevice->secure; > > + fl->dsp_pgid = fastrpc_pgid_alloc(cctx); > + if (fl->dsp_pgid == -1) { > + dev_dbg(&cctx->rpdev->dev, "too many fastrpc clients, max %u allowed\n", MAX_DSP_PD); > + return -EUSERS; Huh? Please return the error code that ida_alloc() returned; > + } > + > fl->sctx = fastrpc_session_alloc(cctx); > if (!fl->sctx) { > dev_err(&cctx->rpdev->dev, "No session available\n"); > @@ -1646,7 +1675,7 @@ static int fastrpc_dmabuf_alloc(struct fastrpc_user *fl, char __user *argp) > static int fastrpc_init_attach(struct fastrpc_user *fl, int pd) > { > struct fastrpc_invoke_args args[1]; > - int tgid = fl->tgid; > + int tgid = fl->dsp_pgid; unsigned? > u32 sc; > > args[0].ptr = (u64)(uintptr_t) &tgid; > @@ -1802,7 +1831,7 @@ static int fastrpc_req_munmap_impl(struct fastrpc_user *fl, struct fastrpc_buf * > int err; > u32 sc; > > - req_msg.pgid = fl->tgid; > + req_msg.pgid = fl->dsp_pgid; > req_msg.size = buf->size; > req_msg.vaddr = buf->raddr; > > @@ -1888,7 +1917,7 @@ static int fastrpc_req_mmap(struct fastrpc_user *fl, char __user *argp) > return err; > } > > - req_msg.pgid = fl->tgid; > + req_msg.pgid = fl->dsp_pgid; > req_msg.flags = req.flags; > req_msg.vaddr = req.vaddrin; > req_msg.num = sizeof(pages); > @@ -1978,7 +2007,7 @@ static int fastrpc_req_mem_unmap_impl(struct fastrpc_user *fl, struct fastrpc_me > return -EINVAL; > } > > - req_msg.pgid = fl->tgid; > + req_msg.pgid = fl->dsp_pgid; > req_msg.len = map->len; > req_msg.vaddrin = map->raddr; > req_msg.fd = map->fd; > @@ -2031,7 +2060,7 @@ static int fastrpc_req_mem_map(struct fastrpc_user *fl, char __user *argp) > return err; > } > > - req_msg.pgid = fl->tgid; > + req_msg.pgid = fl->dsp_pgid; > req_msg.fd = req.fd; > req_msg.offset = req.offset; > req_msg.vaddrin = req.vaddrin; > @@ -2375,6 +2404,7 @@ static int fastrpc_rpmsg_probe(struct rpmsg_device *rpdev) > INIT_LIST_HEAD(&data->invoke_interrupted_mmaps); > spin_lock_init(&data->lock); > idr_init(&data->ctx_idr); > + ida_init(&data->dsp_pgid_ida); > data->domain_id = domain_id; > data->rpdev = rpdev; > > -- > 2.34.1 >
On 7/3/2024 4:09 PM, Greg KH wrote: > On Wed, Jul 03, 2024 at 12:22:00PM +0530, Ekansh Gupta wrote: >> @@ -268,6 +272,7 @@ struct fastrpc_channel_ctx { >> struct fastrpc_session_ctx session[FASTRPC_MAX_SESSIONS]; >> spinlock_t lock; >> struct idr ctx_idr; >> + struct ida dsp_pgid_ida; > You have an idr and an ida? Why two different types for the same > driver? Using ida for this because for this I just need to allocate and manage unique IDs without any associated data. So this looks more space efficient that idr. Should I keep it uniform for a driver? > >> struct list_head users; >> struct kref refcount; >> /* Flag if dsp attributes are cached */ >> @@ -299,6 +304,7 @@ struct fastrpc_user { >> struct fastrpc_buf *init_mem; >> >> int tgid; >> + int dsp_pgid; > Are you sure this fits in an int? I think this should be fine for IDs in rage of 1000-1064. > >> +static int fastrpc_pgid_alloc(struct fastrpc_channel_ctx *cctx) >> +{ >> + int ret = -1; > No need to initialize this. I'll update this. > >> + >> + /* allocate unique id between MIN_FRPC_PGID and MAX_FRPC_PGID */ >> + ret = ida_alloc_range(&cctx->dsp_pgid_ida, MIN_FRPC_PGID, >> + MAX_FRPC_PGID, GFP_ATOMIC); >> + if (ret < 0) >> + return -1; > Why is -1 a specific value here? Return a real error please. > Or return 0 if that's not allowed. Sure, will fix this in next spin. > > v >> + >> + return ret; >> +} >> + >> static int fastrpc_device_open(struct inode *inode, struct file *filp) >> { >> struct fastrpc_channel_ctx *cctx; >> @@ -1582,6 +1605,12 @@ static int fastrpc_device_open(struct inode *inode, struct file *filp) >> fl->cctx = cctx; >> fl->is_secure_dev = fdevice->secure; >> >> + fl->dsp_pgid = fastrpc_pgid_alloc(cctx); >> + if (fl->dsp_pgid == -1) { >> + dev_dbg(&cctx->rpdev->dev, "too many fastrpc clients, max %u allowed\n", MAX_DSP_PD); >> + return -EUSERS; > Why -EUSERS? This should be -EBUSY, I'll correct this. > > And you obviously did not test this as you just leaked memory :( My bad, I ran basic fastrpc tests and the working of this use case. Sorry for the miss. --Ekansh > > thanks, > > greg k-h
On 7/4/2024 11:47 AM, Ekansh Gupta wrote: > > On 7/3/2024 4:09 PM, Greg KH wrote: >> On Wed, Jul 03, 2024 at 12:22:00PM +0530, Ekansh Gupta wrote: >>> @@ -268,6 +272,7 @@ struct fastrpc_channel_ctx { >>> struct fastrpc_session_ctx session[FASTRPC_MAX_SESSIONS]; >>> spinlock_t lock; >>> struct idr ctx_idr; >>> + struct ida dsp_pgid_ida; >> You have an idr and an ida? Why two different types for the same >> driver? > Using ida for this because for this I just need to allocate and manage unique IDs > without any associated data. So this looks more space efficient that idr. > Should I keep it uniform for a driver? >>> struct list_head users; >>> struct kref refcount; >>> /* Flag if dsp attributes are cached */ >>> @@ -299,6 +304,7 @@ struct fastrpc_user { >>> struct fastrpc_buf *init_mem; >>> >>> int tgid; >>> + int dsp_pgid; >> Are you sure this fits in an int? > I think this should be fine for IDs in rage of 1000-1064. changing this to u32 as won't be storin any negative values here if allocation fails. >>> +static int fastrpc_pgid_alloc(struct fastrpc_channel_ctx *cctx) >>> +{ >>> + int ret = -1; >> No need to initialize this. > I'll update this. >>> + >>> + /* allocate unique id between MIN_FRPC_PGID and MAX_FRPC_PGID */ >>> + ret = ida_alloc_range(&cctx->dsp_pgid_ida, MIN_FRPC_PGID, >>> + MAX_FRPC_PGID, GFP_ATOMIC); >>> + if (ret < 0) >>> + return -1; >> Why is -1 a specific value here? Return a real error please. >> Or return 0 if that's not allowed. > Sure, will fix this in next spin. >> v >>> + >>> + return ret; >>> +} >>> + >>> static int fastrpc_device_open(struct inode *inode, struct file *filp) >>> { >>> struct fastrpc_channel_ctx *cctx; >>> @@ -1582,6 +1605,12 @@ static int fastrpc_device_open(struct inode *inode, struct file *filp) >>> fl->cctx = cctx; >>> fl->is_secure_dev = fdevice->secure; >>> >>> + fl->dsp_pgid = fastrpc_pgid_alloc(cctx); >>> + if (fl->dsp_pgid == -1) { >>> + dev_dbg(&cctx->rpdev->dev, "too many fastrpc clients, max %u allowed\n", MAX_DSP_PD); >>> + return -EUSERS; >> Why -EUSERS? > This should be -EBUSY, I'll correct this. >> And you obviously did not test this as you just leaked memory :( > My bad, I ran basic fastrpc tests and the working of this use case. Sorry for the miss. > > --Ekansh >> thanks, >> >> greg k-h >
On 7/3/2024 4:12 PM, Dmitry Baryshkov wrote: > On Wed, Jul 03, 2024 at 12:22:00PM GMT, Ekansh Gupta wrote: >> Memory intensive applications(which requires more tha 4GB) that wants >> to offload tasks to DSP might have to split the tasks to multiple >> user PD to make the resources available. For every call to DSP, >> fastrpc driver passes the process tgid which works as an identifier >> for the DSP to enqueue the tasks to specific PD. With current design, >> if any process opens device node more than once and makes PD init >> request, same tgid will be passed to DSP which will be considered a >> bad request and this will result in failure as the same identifier >> cannot be used for multiple DSP PD. Allocate and pass an effective >> pgid to DSP which would be allocated during device open and will have >> a lifetime till the device is closed. This will allow the same process >> to open the device more than once and spawn multiple dynamic PD for >> ease of processing. > Consider adding line breaks to split the text into paragraphs to make it > easier to read. > > The patch also raises the question whether the identifier should be > unique per DSP or per the channel. I'll update the commit text with more clear information. >> Signed-off-by: Ekansh Gupta <quic_ekangupt@quicinc.com> >> --- >> drivers/misc/fastrpc.c | 48 ++++++++++++++++++++++++++++++++++-------- >> 1 file changed, 39 insertions(+), 9 deletions(-) >> >> diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c >> index 5204fda51da3..7250e30aa93f 100644 >> --- a/drivers/misc/fastrpc.c >> +++ b/drivers/misc/fastrpc.c >> @@ -105,6 +105,10 @@ >> >> #define miscdev_to_fdevice(d) container_of(d, struct fastrpc_device, miscdev) >> >> +#define MAX_DSP_PD 64 >> +#define MIN_FRPC_PGID 1000 >> +#define MAX_FRPC_PGID (MIN_FRPC_PGID + MAX_DSP_PD) >> + >> static const char *domains[FASTRPC_DEV_MAX] = { "adsp", "mdsp", >> "sdsp", "cdsp"}; >> struct fastrpc_phy_page { >> @@ -268,6 +272,7 @@ struct fastrpc_channel_ctx { >> struct fastrpc_session_ctx session[FASTRPC_MAX_SESSIONS]; >> spinlock_t lock; >> struct idr ctx_idr; >> + struct ida dsp_pgid_ida; >> struct list_head users; >> struct kref refcount; >> /* Flag if dsp attributes are cached */ >> @@ -299,6 +304,7 @@ struct fastrpc_user { >> struct fastrpc_buf *init_mem; >> >> int tgid; > Is it still used? What for? I don't see any usage as of now, this can be removed. > >> + int dsp_pgid; > unsigned int planning to keep it u32 as the range is going to be from 1000-1064 > >> int pd; >> bool is_secure_dev; >> /* Lock for lists */ >> @@ -462,6 +468,7 @@ static void fastrpc_channel_ctx_free(struct kref *ref) >> struct fastrpc_channel_ctx *cctx; >> >> cctx = container_of(ref, struct fastrpc_channel_ctx, refcount); >> + ida_destroy(&cctx->dsp_pgid_ida); >> >> kfree(cctx); >> } >> @@ -1114,7 +1121,7 @@ static int fastrpc_invoke_send(struct fastrpc_session_ctx *sctx, >> int ret; >> >> cctx = fl->cctx; >> - msg->pid = fl->tgid; >> + msg->pid = fl->dsp_pgid; >> msg->tid = current->pid; >> >> if (kernel) >> @@ -1292,7 +1299,7 @@ static int fastrpc_init_create_static_process(struct fastrpc_user *fl, >> } >> } >> >> - inbuf.pgid = fl->tgid; >> + inbuf.pgid = fl->dsp_pgid; >> inbuf.namelen = init.namelen; >> inbuf.pageslen = 0; >> fl->pd = USER_PD; >> @@ -1394,7 +1401,7 @@ static int fastrpc_init_create_process(struct fastrpc_user *fl, >> goto err; >> } >> >> - inbuf.pgid = fl->tgid; >> + inbuf.pgid = fl->dsp_pgid; >> inbuf.namelen = strlen(current->comm) + 1; >> inbuf.filelen = init.filelen; >> inbuf.pageslen = 1; >> @@ -1503,7 +1510,7 @@ static int fastrpc_release_current_dsp_process(struct fastrpc_user *fl) >> int tgid = 0; >> u32 sc; >> >> - tgid = fl->tgid; >> + tgid = fl->dsp_pgid; >> args[0].ptr = (u64)(uintptr_t) &tgid; >> args[0].length = sizeof(tgid); >> args[0].fd = -1; >> @@ -1528,6 +1535,9 @@ static int fastrpc_device_release(struct inode *inode, struct file *file) >> list_del(&fl->user); >> spin_unlock_irqrestore(&cctx->lock, flags); >> >> + if (fl->dsp_pgid != -1) > Why -1? On which path can it be left uninitialized? in case ida allocation fails, I was setting this to -1 but now I don't see any reason for this check as if ida alloc fails, that will result in open() failure. > >> + ida_free(&cctx->dsp_pgid_ida, fl->dsp_pgid); >> + >> if (fl->init_mem) >> fastrpc_buf_free(fl->init_mem); >> >> @@ -1554,6 +1564,19 @@ static int fastrpc_device_release(struct inode *inode, struct file *file) >> return 0; >> } >> >> +static int fastrpc_pgid_alloc(struct fastrpc_channel_ctx *cctx) >> +{ >> + int ret = -1; > There is no need to init it, is there? no, I'll remove this. > >> + >> + /* allocate unique id between MIN_FRPC_PGID and MAX_FRPC_PGID */ >> + ret = ida_alloc_range(&cctx->dsp_pgid_ida, MIN_FRPC_PGID, >> + MAX_FRPC_PGID, GFP_ATOMIC); >> + if (ret < 0) >> + return -1; > Don't throw away error codes. Pass them as is. Ack. > >> + >> + return ret; >> +} >> + >> static int fastrpc_device_open(struct inode *inode, struct file *filp) >> { >> struct fastrpc_channel_ctx *cctx; >> @@ -1582,6 +1605,12 @@ static int fastrpc_device_open(struct inode *inode, struct file *filp) >> fl->cctx = cctx; >> fl->is_secure_dev = fdevice->secure; >> >> + fl->dsp_pgid = fastrpc_pgid_alloc(cctx); >> + if (fl->dsp_pgid == -1) { >> + dev_dbg(&cctx->rpdev->dev, "too many fastrpc clients, max %u allowed\n", MAX_DSP_PD); >> + return -EUSERS; > Huh? Please return the error code that ida_alloc() returned; Ack. > >> + } >> + >> fl->sctx = fastrpc_session_alloc(cctx); >> if (!fl->sctx) { >> dev_err(&cctx->rpdev->dev, "No session available\n"); >> @@ -1646,7 +1675,7 @@ static int fastrpc_dmabuf_alloc(struct fastrpc_user *fl, char __user *argp) >> static int fastrpc_init_attach(struct fastrpc_user *fl, int pd) >> { >> struct fastrpc_invoke_args args[1]; >> - int tgid = fl->tgid; >> + int tgid = fl->dsp_pgid; > unsigned? will update this. Thanks. --Ekansh > >> u32 sc; >> >> args[0].ptr = (u64)(uintptr_t) &tgid; >> @@ -1802,7 +1831,7 @@ static int fastrpc_req_munmap_impl(struct fastrpc_user *fl, struct fastrpc_buf * >> int err; >> u32 sc; >> >> - req_msg.pgid = fl->tgid; >> + req_msg.pgid = fl->dsp_pgid; >> req_msg.size = buf->size; >> req_msg.vaddr = buf->raddr; >> >> @@ -1888,7 +1917,7 @@ static int fastrpc_req_mmap(struct fastrpc_user *fl, char __user *argp) >> return err; >> } >> >> - req_msg.pgid = fl->tgid; >> + req_msg.pgid = fl->dsp_pgid; >> req_msg.flags = req.flags; >> req_msg.vaddr = req.vaddrin; >> req_msg.num = sizeof(pages); >> @@ -1978,7 +2007,7 @@ static int fastrpc_req_mem_unmap_impl(struct fastrpc_user *fl, struct fastrpc_me >> return -EINVAL; >> } >> >> - req_msg.pgid = fl->tgid; >> + req_msg.pgid = fl->dsp_pgid; >> req_msg.len = map->len; >> req_msg.vaddrin = map->raddr; >> req_msg.fd = map->fd; >> @@ -2031,7 +2060,7 @@ static int fastrpc_req_mem_map(struct fastrpc_user *fl, char __user *argp) >> return err; >> } >> >> - req_msg.pgid = fl->tgid; >> + req_msg.pgid = fl->dsp_pgid; >> req_msg.fd = req.fd; >> req_msg.offset = req.offset; >> req_msg.vaddrin = req.vaddrin; >> @@ -2375,6 +2404,7 @@ static int fastrpc_rpmsg_probe(struct rpmsg_device *rpdev) >> INIT_LIST_HEAD(&data->invoke_interrupted_mmaps); >> spin_lock_init(&data->lock); >> idr_init(&data->ctx_idr); >> + ida_init(&data->dsp_pgid_ida); >> data->domain_id = domain_id; >> data->rpdev = rpdev; >> >> -- >> 2.34.1 >>
On Thu, Jul 04, 2024 at 11:47:22AM +0530, Ekansh Gupta wrote: > > > On 7/3/2024 4:09 PM, Greg KH wrote: > > On Wed, Jul 03, 2024 at 12:22:00PM +0530, Ekansh Gupta wrote: > >> @@ -268,6 +272,7 @@ struct fastrpc_channel_ctx { > >> struct fastrpc_session_ctx session[FASTRPC_MAX_SESSIONS]; > >> spinlock_t lock; > >> struct idr ctx_idr; > >> + struct ida dsp_pgid_ida; > > You have an idr and an ida? Why two different types for the same > > driver? > Using ida for this because for this I just need to allocate and manage unique IDs > without any associated data. So this looks more space efficient that idr. > Should I keep it uniform for a driver? Be consistant, it will make your life easier over the next 10+ years that you have to maintain it. > >> struct list_head users; > >> struct kref refcount; > >> /* Flag if dsp attributes are cached */ > >> @@ -299,6 +304,7 @@ struct fastrpc_user { > >> struct fastrpc_buf *init_mem; > >> > >> int tgid; > >> + int dsp_pgid; > > Are you sure this fits in an int? > I think this should be fine for IDs in rage of 1000-1064. Where is that range specified anywhere? I don't see it documented here for us to know that :( And if that's all you care about, then use a u16. thanks, greg k-h
diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c index 5204fda51da3..7250e30aa93f 100644 --- a/drivers/misc/fastrpc.c +++ b/drivers/misc/fastrpc.c @@ -105,6 +105,10 @@ #define miscdev_to_fdevice(d) container_of(d, struct fastrpc_device, miscdev) +#define MAX_DSP_PD 64 +#define MIN_FRPC_PGID 1000 +#define MAX_FRPC_PGID (MIN_FRPC_PGID + MAX_DSP_PD) + static const char *domains[FASTRPC_DEV_MAX] = { "adsp", "mdsp", "sdsp", "cdsp"}; struct fastrpc_phy_page { @@ -268,6 +272,7 @@ struct fastrpc_channel_ctx { struct fastrpc_session_ctx session[FASTRPC_MAX_SESSIONS]; spinlock_t lock; struct idr ctx_idr; + struct ida dsp_pgid_ida; struct list_head users; struct kref refcount; /* Flag if dsp attributes are cached */ @@ -299,6 +304,7 @@ struct fastrpc_user { struct fastrpc_buf *init_mem; int tgid; + int dsp_pgid; int pd; bool is_secure_dev; /* Lock for lists */ @@ -462,6 +468,7 @@ static void fastrpc_channel_ctx_free(struct kref *ref) struct fastrpc_channel_ctx *cctx; cctx = container_of(ref, struct fastrpc_channel_ctx, refcount); + ida_destroy(&cctx->dsp_pgid_ida); kfree(cctx); } @@ -1114,7 +1121,7 @@ static int fastrpc_invoke_send(struct fastrpc_session_ctx *sctx, int ret; cctx = fl->cctx; - msg->pid = fl->tgid; + msg->pid = fl->dsp_pgid; msg->tid = current->pid; if (kernel) @@ -1292,7 +1299,7 @@ static int fastrpc_init_create_static_process(struct fastrpc_user *fl, } } - inbuf.pgid = fl->tgid; + inbuf.pgid = fl->dsp_pgid; inbuf.namelen = init.namelen; inbuf.pageslen = 0; fl->pd = USER_PD; @@ -1394,7 +1401,7 @@ static int fastrpc_init_create_process(struct fastrpc_user *fl, goto err; } - inbuf.pgid = fl->tgid; + inbuf.pgid = fl->dsp_pgid; inbuf.namelen = strlen(current->comm) + 1; inbuf.filelen = init.filelen; inbuf.pageslen = 1; @@ -1503,7 +1510,7 @@ static int fastrpc_release_current_dsp_process(struct fastrpc_user *fl) int tgid = 0; u32 sc; - tgid = fl->tgid; + tgid = fl->dsp_pgid; args[0].ptr = (u64)(uintptr_t) &tgid; args[0].length = sizeof(tgid); args[0].fd = -1; @@ -1528,6 +1535,9 @@ static int fastrpc_device_release(struct inode *inode, struct file *file) list_del(&fl->user); spin_unlock_irqrestore(&cctx->lock, flags); + if (fl->dsp_pgid != -1) + ida_free(&cctx->dsp_pgid_ida, fl->dsp_pgid); + if (fl->init_mem) fastrpc_buf_free(fl->init_mem); @@ -1554,6 +1564,19 @@ static int fastrpc_device_release(struct inode *inode, struct file *file) return 0; } +static int fastrpc_pgid_alloc(struct fastrpc_channel_ctx *cctx) +{ + int ret = -1; + + /* allocate unique id between MIN_FRPC_PGID and MAX_FRPC_PGID */ + ret = ida_alloc_range(&cctx->dsp_pgid_ida, MIN_FRPC_PGID, + MAX_FRPC_PGID, GFP_ATOMIC); + if (ret < 0) + return -1; + + return ret; +} + static int fastrpc_device_open(struct inode *inode, struct file *filp) { struct fastrpc_channel_ctx *cctx; @@ -1582,6 +1605,12 @@ static int fastrpc_device_open(struct inode *inode, struct file *filp) fl->cctx = cctx; fl->is_secure_dev = fdevice->secure; + fl->dsp_pgid = fastrpc_pgid_alloc(cctx); + if (fl->dsp_pgid == -1) { + dev_dbg(&cctx->rpdev->dev, "too many fastrpc clients, max %u allowed\n", MAX_DSP_PD); + return -EUSERS; + } + fl->sctx = fastrpc_session_alloc(cctx); if (!fl->sctx) { dev_err(&cctx->rpdev->dev, "No session available\n"); @@ -1646,7 +1675,7 @@ static int fastrpc_dmabuf_alloc(struct fastrpc_user *fl, char __user *argp) static int fastrpc_init_attach(struct fastrpc_user *fl, int pd) { struct fastrpc_invoke_args args[1]; - int tgid = fl->tgid; + int tgid = fl->dsp_pgid; u32 sc; args[0].ptr = (u64)(uintptr_t) &tgid; @@ -1802,7 +1831,7 @@ static int fastrpc_req_munmap_impl(struct fastrpc_user *fl, struct fastrpc_buf * int err; u32 sc; - req_msg.pgid = fl->tgid; + req_msg.pgid = fl->dsp_pgid; req_msg.size = buf->size; req_msg.vaddr = buf->raddr; @@ -1888,7 +1917,7 @@ static int fastrpc_req_mmap(struct fastrpc_user *fl, char __user *argp) return err; } - req_msg.pgid = fl->tgid; + req_msg.pgid = fl->dsp_pgid; req_msg.flags = req.flags; req_msg.vaddr = req.vaddrin; req_msg.num = sizeof(pages); @@ -1978,7 +2007,7 @@ static int fastrpc_req_mem_unmap_impl(struct fastrpc_user *fl, struct fastrpc_me return -EINVAL; } - req_msg.pgid = fl->tgid; + req_msg.pgid = fl->dsp_pgid; req_msg.len = map->len; req_msg.vaddrin = map->raddr; req_msg.fd = map->fd; @@ -2031,7 +2060,7 @@ static int fastrpc_req_mem_map(struct fastrpc_user *fl, char __user *argp) return err; } - req_msg.pgid = fl->tgid; + req_msg.pgid = fl->dsp_pgid; req_msg.fd = req.fd; req_msg.offset = req.offset; req_msg.vaddrin = req.vaddrin; @@ -2375,6 +2404,7 @@ static int fastrpc_rpmsg_probe(struct rpmsg_device *rpdev) INIT_LIST_HEAD(&data->invoke_interrupted_mmaps); spin_lock_init(&data->lock); idr_init(&data->ctx_idr); + ida_init(&data->dsp_pgid_ida); data->domain_id = domain_id; data->rpdev = rpdev;
Memory intensive applications(which requires more tha 4GB) that wants to offload tasks to DSP might have to split the tasks to multiple user PD to make the resources available. For every call to DSP, fastrpc driver passes the process tgid which works as an identifier for the DSP to enqueue the tasks to specific PD. With current design, if any process opens device node more than once and makes PD init request, same tgid will be passed to DSP which will be considered a bad request and this will result in failure as the same identifier cannot be used for multiple DSP PD. Allocate and pass an effective pgid to DSP which would be allocated during device open and will have a lifetime till the device is closed. This will allow the same process to open the device more than once and spawn multiple dynamic PD for ease of processing. Signed-off-by: Ekansh Gupta <quic_ekangupt@quicinc.com> --- drivers/misc/fastrpc.c | 48 ++++++++++++++++++++++++++++++++++-------- 1 file changed, 39 insertions(+), 9 deletions(-)