| Message ID | 1632386272-18139-1-git-send-email-jeyr@codeaurora.org (mailing list archive) |
|---|---|
| State | Not Applicable |
| Headers | show |
| Series | misc: fastrpc: copy to user only for non-DMA-BUF heap buffers | expand |
On Thu, Sep 23, 2021 at 02:07:52PM +0530, Jeya R wrote: > fastrpc_put_args is copying all the output buffers to user. For large > number of output context buffers, this might cause performance > degradation. Copying is not needed for DMA-BUF heap buffers. What does "performance degradation" really mean? > > Signed-off-by: Jeya R <jeyr@codeaurora.org> > --- > drivers/misc/fastrpc.c | 18 ++++++++++-------- > 1 file changed, 10 insertions(+), 8 deletions(-) > > diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c > index beda610..536eabf 100644 > --- a/drivers/misc/fastrpc.c > +++ b/drivers/misc/fastrpc.c > @@ -890,15 +890,17 @@ static int fastrpc_put_args(struct fastrpc_invoke_ctx *ctx, > inbufs = REMOTE_SCALARS_INBUFS(ctx->sc); > > for (i = inbufs; i < ctx->nbufs; ++i) { > - void *src = (void *)(uintptr_t)rpra[i].pv; > - void *dst = (void *)(uintptr_t)ctx->args[i].ptr; > - u64 len = rpra[i].len; > + if (!ctx->maps[i]) { > + void *src = (void *)(uintptr_t)rpra[i].pv; > + void *dst = (void *)(uintptr_t)ctx->args[i].ptr; uintptr_t is not a kernel variable type. Please use the real kernel type for this as you are touching these lines. > + u64 len = rpra[i].len; > > - if (!kernel) { > - if (copy_to_user((void __user *)dst, src, len)) > - return -EFAULT; > - } else { > - memcpy(dst, src, len); > + if (!kernel) { > + if (copy_to_user((void __user *)dst, src, len)) > + return -EFAULT; > + } else { > + memcpy(dst, src, len); > + } So you were copying buffers that didn't need to be copied? So you are now doing less work? Or is this fixing a bug where you were copying things that you should not have been copying? What commit does this fix? Does this need to go to the stable kernel trees? thanks, greg k-h
On 2021-09-23 14:19, Greg KH wrote: > On Thu, Sep 23, 2021 at 02:07:52PM +0530, Jeya R wrote: >> fastrpc_put_args is copying all the output buffers to user. For large >> number of output context buffers, this might cause performance >> degradation. Copying is not needed for DMA-BUF heap buffers. > > What does "performance degradation" really mean? Unnecessary copying for large number of buffers would cause some additional time which would get added to overall fastrpc call cost. > >> >> Signed-off-by: Jeya R <jeyr@codeaurora.org> >> --- >> drivers/misc/fastrpc.c | 18 ++++++++++-------- >> 1 file changed, 10 insertions(+), 8 deletions(-) >> >> diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c >> index beda610..536eabf 100644 >> --- a/drivers/misc/fastrpc.c >> +++ b/drivers/misc/fastrpc.c >> @@ -890,15 +890,17 @@ static int fastrpc_put_args(struct >> fastrpc_invoke_ctx *ctx, >> inbufs = REMOTE_SCALARS_INBUFS(ctx->sc); >> >> for (i = inbufs; i < ctx->nbufs; ++i) { >> - void *src = (void *)(uintptr_t)rpra[i].pv; >> - void *dst = (void *)(uintptr_t)ctx->args[i].ptr; >> - u64 len = rpra[i].len; >> + if (!ctx->maps[i]) { >> + void *src = (void *)(uintptr_t)rpra[i].pv; >> + void *dst = (void *)(uintptr_t)ctx->args[i].ptr; > > uintptr_t is not a kernel variable type. Please use the real kernel > type for this as you are touching these lines. > Sure, thanks for pointing this. Will update this in the next patch. >> + u64 len = rpra[i].len; >> >> - if (!kernel) { >> - if (copy_to_user((void __user *)dst, src, len)) >> - return -EFAULT; >> - } else { >> - memcpy(dst, src, len); >> + if (!kernel) { >> + if (copy_to_user((void __user *)dst, src, len)) >> + return -EFAULT; >> + } else { >> + memcpy(dst, src, len); >> + } > > So you were copying buffers that didn't need to be copied? So you are > now doing less work? Or is this fixing a bug where you were copying > things that you should not have been copying? > > What commit does this fix? Does this need to go to the stable kernel > trees? > Yes, not all buffer needs to be copied. This change would avoid unnecessary copying of buffers. Not adding fix tag as it's not exactly fixing any bug. This should go to stable kernel trees. Thanks > thanks, > > greg k-h
On Thu, Sep 23, 2021 at 03:54:15PM +0530, jeyr@codeaurora.org wrote: > On 2021-09-23 14:19, Greg KH wrote: > > On Thu, Sep 23, 2021 at 02:07:52PM +0530, Jeya R wrote: > > > fastrpc_put_args is copying all the output buffers to user. For large > > > number of output context buffers, this might cause performance > > > degradation. Copying is not needed for DMA-BUF heap buffers. > > > > What does "performance degradation" really mean? > > Unnecessary copying for large number of buffers would cause some > additional time which would get added to overall fastrpc call cost. > > > > > > > > > Signed-off-by: Jeya R <jeyr@codeaurora.org> > > > --- > > > drivers/misc/fastrpc.c | 18 ++++++++++-------- > > > 1 file changed, 10 insertions(+), 8 deletions(-) > > > > > > diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c > > > index beda610..536eabf 100644 > > > --- a/drivers/misc/fastrpc.c > > > +++ b/drivers/misc/fastrpc.c > > > @@ -890,15 +890,17 @@ static int fastrpc_put_args(struct > > > fastrpc_invoke_ctx *ctx, > > > inbufs = REMOTE_SCALARS_INBUFS(ctx->sc); > > > > > > for (i = inbufs; i < ctx->nbufs; ++i) { > > > - void *src = (void *)(uintptr_t)rpra[i].pv; > > > - void *dst = (void *)(uintptr_t)ctx->args[i].ptr; > > > - u64 len = rpra[i].len; > > > + if (!ctx->maps[i]) { > > > + void *src = (void *)(uintptr_t)rpra[i].pv; > > > + void *dst = (void *)(uintptr_t)ctx->args[i].ptr; > > > > uintptr_t is not a kernel variable type. Please use the real kernel > > type for this as you are touching these lines. > > > > Sure, thanks for pointing this. Will update this in the next patch. > > > > + u64 len = rpra[i].len; > > > > > > - if (!kernel) { > > > - if (copy_to_user((void __user *)dst, src, len)) > > > - return -EFAULT; > > > - } else { > > > - memcpy(dst, src, len); > > > + if (!kernel) { > > > + if (copy_to_user((void __user *)dst, src, len)) > > > + return -EFAULT; > > > + } else { > > > + memcpy(dst, src, len); > > > + } > > > > So you were copying buffers that didn't need to be copied? So you are > > now doing less work? Or is this fixing a bug where you were copying > > things that you should not have been copying? > > > > What commit does this fix? Does this need to go to the stable kernel > > trees? > > > > Yes, not all buffer needs to be copied. This change would avoid unnecessary > copying of buffers. Not adding fix tag as it's not exactly fixing any bug. > This should go to stable kernel trees. If it's not fixing a bug, why should it go into the stable trees?
On 2021-09-23 23:13, Greg KH wrote: > On Thu, Sep 23, 2021 at 03:54:15PM +0530, jeyr@codeaurora.org wrote: >> On 2021-09-23 14:19, Greg KH wrote: >> > On Thu, Sep 23, 2021 at 02:07:52PM +0530, Jeya R wrote: >> > > fastrpc_put_args is copying all the output buffers to user. For large >> > > number of output context buffers, this might cause performance >> > > degradation. Copying is not needed for DMA-BUF heap buffers. >> > >> > What does "performance degradation" really mean? >> >> Unnecessary copying for large number of buffers would cause some >> additional time which would get added to overall fastrpc call cost. >> >> > >> > > >> > > Signed-off-by: Jeya R <jeyr@codeaurora.org> >> > > --- >> > > drivers/misc/fastrpc.c | 18 ++++++++++-------- >> > > 1 file changed, 10 insertions(+), 8 deletions(-) >> > > >> > > diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c >> > > index beda610..536eabf 100644 >> > > --- a/drivers/misc/fastrpc.c >> > > +++ b/drivers/misc/fastrpc.c >> > > @@ -890,15 +890,17 @@ static int fastrpc_put_args(struct >> > > fastrpc_invoke_ctx *ctx, >> > > inbufs = REMOTE_SCALARS_INBUFS(ctx->sc); >> > > >> > > for (i = inbufs; i < ctx->nbufs; ++i) { >> > > - void *src = (void *)(uintptr_t)rpra[i].pv; >> > > - void *dst = (void *)(uintptr_t)ctx->args[i].ptr; >> > > - u64 len = rpra[i].len; >> > > + if (!ctx->maps[i]) { >> > > + void *src = (void *)(uintptr_t)rpra[i].pv; >> > > + void *dst = (void *)(uintptr_t)ctx->args[i].ptr; >> > >> > uintptr_t is not a kernel variable type. Please use the real kernel >> > type for this as you are touching these lines. >> > >> >> Sure, thanks for pointing this. Will update this in the next patch. >> >> > > + u64 len = rpra[i].len; >> > > >> > > - if (!kernel) { >> > > - if (copy_to_user((void __user *)dst, src, len)) >> > > - return -EFAULT; >> > > - } else { >> > > - memcpy(dst, src, len); >> > > + if (!kernel) { >> > > + if (copy_to_user((void __user *)dst, src, len)) >> > > + return -EFAULT; >> > > + } else { >> > > + memcpy(dst, src, len); >> > > + } >> > >> > So you were copying buffers that didn't need to be copied? So you are >> > now doing less work? Or is this fixing a bug where you were copying >> > things that you should not have been copying? >> > >> > What commit does this fix? Does this need to go to the stable kernel >> > trees? >> > >> >> Yes, not all buffer needs to be copied. This change would avoid >> unnecessary >> copying of buffers. Not adding fix tag as it's not exactly fixing any >> bug. >> This should go to stable kernel trees. > > If it's not fixing a bug, why should it go into the stable trees? This is not a bug fix, it can be considered as an enhancement and it can go in to new release.
On Fri, Sep 24, 2021 at 11:55:32AM +0530, jeyr@codeaurora.org wrote: > On 2021-09-23 23:13, Greg KH wrote: > > On Thu, Sep 23, 2021 at 03:54:15PM +0530, jeyr@codeaurora.org wrote: > > > On 2021-09-23 14:19, Greg KH wrote: > > > > On Thu, Sep 23, 2021 at 02:07:52PM +0530, Jeya R wrote: > > > > > fastrpc_put_args is copying all the output buffers to user. For large > > > > > number of output context buffers, this might cause performance > > > > > degradation. Copying is not needed for DMA-BUF heap buffers. > > > > > > > > What does "performance degradation" really mean? > > > > > > Unnecessary copying for large number of buffers would cause some > > > additional time which would get added to overall fastrpc call cost. > > > > > > > > > > > > > > > > > Signed-off-by: Jeya R <jeyr@codeaurora.org> > > > > > --- > > > > > drivers/misc/fastrpc.c | 18 ++++++++++-------- > > > > > 1 file changed, 10 insertions(+), 8 deletions(-) > > > > > > > > > > diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c > > > > > index beda610..536eabf 100644 > > > > > --- a/drivers/misc/fastrpc.c > > > > > +++ b/drivers/misc/fastrpc.c > > > > > @@ -890,15 +890,17 @@ static int fastrpc_put_args(struct > > > > > fastrpc_invoke_ctx *ctx, > > > > > inbufs = REMOTE_SCALARS_INBUFS(ctx->sc); > > > > > > > > > > for (i = inbufs; i < ctx->nbufs; ++i) { > > > > > - void *src = (void *)(uintptr_t)rpra[i].pv; > > > > > - void *dst = (void *)(uintptr_t)ctx->args[i].ptr; > > > > > - u64 len = rpra[i].len; > > > > > + if (!ctx->maps[i]) { > > > > > + void *src = (void *)(uintptr_t)rpra[i].pv; > > > > > + void *dst = (void *)(uintptr_t)ctx->args[i].ptr; > > > > > > > > uintptr_t is not a kernel variable type. Please use the real kernel > > > > type for this as you are touching these lines. > > > > > > > > > > Sure, thanks for pointing this. Will update this in the next patch. > > > > > > > > + u64 len = rpra[i].len; > > > > > > > > > > - if (!kernel) { > > > > > - if (copy_to_user((void __user *)dst, src, len)) > > > > > - return -EFAULT; > > > > > - } else { > > > > > - memcpy(dst, src, len); > > > > > + if (!kernel) { > > > > > + if (copy_to_user((void __user *)dst, src, len)) > > > > > + return -EFAULT; > > > > > + } else { > > > > > + memcpy(dst, src, len); > > > > > + } > > > > > > > > So you were copying buffers that didn't need to be copied? So you are > > > > now doing less work? Or is this fixing a bug where you were copying > > > > things that you should not have been copying? > > > > > > > > What commit does this fix? Does this need to go to the stable kernel > > > > trees? > > > > > > > > > > Yes, not all buffer needs to be copied. This change would avoid > > > unnecessary > > > copying of buffers. Not adding fix tag as it's not exactly fixing > > > any bug. > > > This should go to stable kernel trees. > > > > If it's not fixing a bug, why should it go into the stable trees? > > This is not a bug fix, it can be considered as an enhancement and it can go > in > to new release. Ok, please update and send a v2 with the needed changes. thanks, greg k-h
diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c index beda610..536eabf 100644 --- a/drivers/misc/fastrpc.c +++ b/drivers/misc/fastrpc.c @@ -890,15 +890,17 @@ static int fastrpc_put_args(struct fastrpc_invoke_ctx *ctx, inbufs = REMOTE_SCALARS_INBUFS(ctx->sc); for (i = inbufs; i < ctx->nbufs; ++i) { - void *src = (void *)(uintptr_t)rpra[i].pv; - void *dst = (void *)(uintptr_t)ctx->args[i].ptr; - u64 len = rpra[i].len; + if (!ctx->maps[i]) { + void *src = (void *)(uintptr_t)rpra[i].pv; + void *dst = (void *)(uintptr_t)ctx->args[i].ptr; + u64 len = rpra[i].len; - if (!kernel) { - if (copy_to_user((void __user *)dst, src, len)) - return -EFAULT; - } else { - memcpy(dst, src, len); + if (!kernel) { + if (copy_to_user((void __user *)dst, src, len)) + return -EFAULT; + } else { + memcpy(dst, src, len); + } } }
fastrpc_put_args is copying all the output buffers to user. For large number of output context buffers, this might cause performance degradation. Copying is not needed for DMA-BUF heap buffers. Signed-off-by: Jeya R <jeyr@codeaurora.org> --- drivers/misc/fastrpc.c | 18 ++++++++++-------- 1 file changed, 10 insertions(+), 8 deletions(-)