| Message ID | 1586420572-28353-2-git-send-email-peng.fan@nxp.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [1/2] remoteproc: drop memset when loading elf segments | expand |
On Thu 09 Apr 01:22 PDT 2020, Peng Fan wrote: > Since we no need memset if memsz is larger than filesz, we could > use filesz for the da to va translation when memsz translation fail. > To me this implies that the firmware has a segment that's larger than the memory that it's going to run in. I think even if we're not writing to the entire memsz, asking da_to_va for the entire memsz provides a valuable sanity check. Regards, Bjorn > Signed-off-by: Peng Fan <peng.fan@nxp.com> > --- > drivers/remoteproc/remoteproc_elf_loader.c | 12 ++++++++++-- > 1 file changed, 10 insertions(+), 2 deletions(-) > > diff --git a/drivers/remoteproc/remoteproc_elf_loader.c b/drivers/remoteproc/remoteproc_elf_loader.c > index cc50fe70d50c..74d425a4b34c 100644 > --- a/drivers/remoteproc/remoteproc_elf_loader.c > +++ b/drivers/remoteproc/remoteproc_elf_loader.c > @@ -229,8 +229,16 @@ int rproc_elf_load_segments(struct rproc *rproc, const struct firmware *fw) > if (!ptr) { > dev_err(dev, "bad phdr da 0x%llx mem 0x%llx\n", da, > memsz); > - ret = -EINVAL; > - break; > + > + ptr = rproc_da_to_va(rproc, da, filesz); > + if (!ptr) { > + dev_err(dev, > + "bad phdr da 0x%llx mem 0x%llx\n", > + da, filesz); > + ret = -EINVAL; > + break; > + } > + > } > > /* put the segment where the remote processor expects it */ > -- > 2.16.4 >
Hi Bjorn, > Subject: Re: [PATCH 2/2] remoteproc: use filesz as backup when translate > memsz fail > > On Thu 09 Apr 01:22 PDT 2020, Peng Fan wrote: > > > Since we no need memset if memsz is larger than filesz, we could use > > filesz for the da to va translation when memsz translation fail. > > > > To me this implies that the firmware has a segment that's larger than the > memory that it's going to run in. I think even if we're not writing to the entire > memsz, asking da_to_va for the entire memsz provides a valuable sanity > check. da_to_va implies that Linux should have the va map to da. However that will be case that Linux is not able to touch all da, it only able touch half. Then Linux should also map all da? Thanks, Peng. > > Regards, > Bjorn > > > Signed-off-by: Peng Fan <peng.fan@nxp.com> > > --- > > drivers/remoteproc/remoteproc_elf_loader.c | 12 ++++++++++-- > > 1 file changed, 10 insertions(+), 2 deletions(-) > > > > diff --git a/drivers/remoteproc/remoteproc_elf_loader.c > > b/drivers/remoteproc/remoteproc_elf_loader.c > > index cc50fe70d50c..74d425a4b34c 100644 > > --- a/drivers/remoteproc/remoteproc_elf_loader.c > > +++ b/drivers/remoteproc/remoteproc_elf_loader.c > > @@ -229,8 +229,16 @@ int rproc_elf_load_segments(struct rproc *rproc, > const struct firmware *fw) > > if (!ptr) { > > dev_err(dev, "bad phdr da 0x%llx mem 0x%llx\n", da, > > memsz); > > - ret = -EINVAL; > > - break; > > + > > + ptr = rproc_da_to_va(rproc, da, filesz); > > + if (!ptr) { > > + dev_err(dev, > > + "bad phdr da 0x%llx mem 0x%llx\n", > > + da, filesz); > > + ret = -EINVAL; > > + break; > > + } > > + > > } > > > > /* put the segment where the remote processor expects it */ > > -- > > 2.16.4 > >
On Thu 09 Apr 18:32 PDT 2020, Peng Fan wrote: > Hi Bjorn, > > > Subject: Re: [PATCH 2/2] remoteproc: use filesz as backup when translate > > memsz fail > > > > On Thu 09 Apr 01:22 PDT 2020, Peng Fan wrote: > > > > > Since we no need memset if memsz is larger than filesz, we could use > > > filesz for the da to va translation when memsz translation fail. > > > > > > > To me this implies that the firmware has a segment that's larger than the > > memory that it's going to run in. I think even if we're not writing to the entire > > memsz, asking da_to_va for the entire memsz provides a valuable sanity > > check. > > da_to_va implies that Linux should have the va map to da. However > that will be case that Linux is not able to touch all da, it only able touch > half. Then Linux should also map all da? > So you have memory described in your ELF that can only be accessed by the remoteproc? And this memory is covered by segments of type PT_LOAD? What's your strategy for making sure that filesz stays within the boundaries that the ELF loader is allowed to touch? Regards, Bjorn > Thanks, > Peng. > > > > > Regards, > > Bjorn > > > > > Signed-off-by: Peng Fan <peng.fan@nxp.com> > > > --- > > > drivers/remoteproc/remoteproc_elf_loader.c | 12 ++++++++++-- > > > 1 file changed, 10 insertions(+), 2 deletions(-) > > > > > > diff --git a/drivers/remoteproc/remoteproc_elf_loader.c > > > b/drivers/remoteproc/remoteproc_elf_loader.c > > > index cc50fe70d50c..74d425a4b34c 100644 > > > --- a/drivers/remoteproc/remoteproc_elf_loader.c > > > +++ b/drivers/remoteproc/remoteproc_elf_loader.c > > > @@ -229,8 +229,16 @@ int rproc_elf_load_segments(struct rproc *rproc, > > const struct firmware *fw) > > > if (!ptr) { > > > dev_err(dev, "bad phdr da 0x%llx mem 0x%llx\n", da, > > > memsz); > > > - ret = -EINVAL; > > > - break; > > > + > > > + ptr = rproc_da_to_va(rproc, da, filesz); > > > + if (!ptr) { > > > + dev_err(dev, > > > + "bad phdr da 0x%llx mem 0x%llx\n", > > > + da, filesz); > > > + ret = -EINVAL; > > > + break; > > > + } > > > + > > > } > > > > > > /* put the segment where the remote processor expects it */ > > > -- > > > 2.16.4 > > >
On Thu, Apr 09, 2020 at 06:22:26PM -0700, Bjorn Andersson wrote: > On Thu 09 Apr 01:22 PDT 2020, Peng Fan wrote: > > > Since we no need memset if memsz is larger than filesz, we could > > use filesz for the da to va translation when memsz translation fail. > > > > To me this implies that the firmware has a segment that's larger than > the memory that it's going to run in. I think even if we're not writing > to the entire memsz, asking da_to_va for the entire memsz provides a > valuable sanity check. > > Regards, > Bjorn > > > Signed-off-by: Peng Fan <peng.fan@nxp.com> > > --- > > drivers/remoteproc/remoteproc_elf_loader.c | 12 ++++++++++-- > > 1 file changed, 10 insertions(+), 2 deletions(-) > > > > diff --git a/drivers/remoteproc/remoteproc_elf_loader.c b/drivers/remoteproc/remoteproc_elf_loader.c > > index cc50fe70d50c..74d425a4b34c 100644 > > --- a/drivers/remoteproc/remoteproc_elf_loader.c > > +++ b/drivers/remoteproc/remoteproc_elf_loader.c > > @@ -229,8 +229,16 @@ int rproc_elf_load_segments(struct rproc *rproc, const struct firmware *fw) > > if (!ptr) { > > dev_err(dev, "bad phdr da 0x%llx mem 0x%llx\n", da, > > memsz); > > - ret = -EINVAL; > > - break; > > + > > + ptr = rproc_da_to_va(rproc, da, filesz); > > + if (!ptr) { > > + dev_err(dev, > > + "bad phdr da 0x%llx mem 0x%llx\n", > > + da, filesz); > > + ret = -EINVAL; > > + break; > > + } > > + Adding to Bjorn's comment, I think if rproc_da_to_va() fails with memsz but succeeds with filesz something went wrong with how memory was laid out in the DT or the ELF resources. To me this patch offers the wrong solution - the focus should be on why rproc_da_to_va() fails. Thanks, Mathieu > > } > > > > /* put the segment where the remote processor expects it */ > > -- > > 2.16.4 > >
> Subject: Re: [PATCH 2/2] remoteproc: use filesz as backup when translate > memsz fail > > On Thu, Apr 09, 2020 at 06:22:26PM -0700, Bjorn Andersson wrote: > > On Thu 09 Apr 01:22 PDT 2020, Peng Fan wrote: > > > > > Since we no need memset if memsz is larger than filesz, we could use > > > filesz for the da to va translation when memsz translation fail. > > > > > > > To me this implies that the firmware has a segment that's larger than > > the memory that it's going to run in. I think even if we're not > > writing to the entire memsz, asking da_to_va for the entire memsz > > provides a valuable sanity check. > > > > Regards, > > Bjorn > > > > > Signed-off-by: Peng Fan <peng.fan@nxp.com> > > > --- > > > drivers/remoteproc/remoteproc_elf_loader.c | 12 ++++++++++-- > > > 1 file changed, 10 insertions(+), 2 deletions(-) > > > > > > diff --git a/drivers/remoteproc/remoteproc_elf_loader.c > > > b/drivers/remoteproc/remoteproc_elf_loader.c > > > index cc50fe70d50c..74d425a4b34c 100644 > > > --- a/drivers/remoteproc/remoteproc_elf_loader.c > > > +++ b/drivers/remoteproc/remoteproc_elf_loader.c > > > @@ -229,8 +229,16 @@ int rproc_elf_load_segments(struct rproc *rproc, > const struct firmware *fw) > > > if (!ptr) { > > > dev_err(dev, "bad phdr da 0x%llx mem 0x%llx\n", da, > > > memsz); > > > - ret = -EINVAL; > > > - break; > > > + > > > + ptr = rproc_da_to_va(rproc, da, filesz); > > > + if (!ptr) { > > > + dev_err(dev, > > > + "bad phdr da 0x%llx mem 0x%llx\n", > > > + da, filesz); > > > + ret = -EINVAL; > > > + break; > > > + } > > > + > > Adding to Bjorn's comment, I think if rproc_da_to_va() fails with memsz but > succeeds with filesz something went wrong with how memory was laid out in > the DT or the ELF resources. To me this patch offers the wrong solution - the > focus should be on why rproc_da_to_va() fails. ok, I'll send out patch to fix imx_rproc. It is TCML and TCMU are not mapped continusly. Thanks, Peng. > > Thanks, > Mathieu > > > > } > > > > > > /* put the segment where the remote processor expects it */ > > > -- > > > 2.16.4 > > >
diff --git a/drivers/remoteproc/remoteproc_elf_loader.c b/drivers/remoteproc/remoteproc_elf_loader.c index cc50fe70d50c..74d425a4b34c 100644 --- a/drivers/remoteproc/remoteproc_elf_loader.c +++ b/drivers/remoteproc/remoteproc_elf_loader.c @@ -229,8 +229,16 @@ int rproc_elf_load_segments(struct rproc *rproc, const struct firmware *fw) if (!ptr) { dev_err(dev, "bad phdr da 0x%llx mem 0x%llx\n", da, memsz); - ret = -EINVAL; - break; + + ptr = rproc_da_to_va(rproc, da, filesz); + if (!ptr) { + dev_err(dev, + "bad phdr da 0x%llx mem 0x%llx\n", + da, filesz); + ret = -EINVAL; + break; + } + } /* put the segment where the remote processor expects it */
Since we no need memset if memsz is larger than filesz, we could use filesz for the da to va translation when memsz translation fail. Signed-off-by: Peng Fan <peng.fan@nxp.com> --- drivers/remoteproc/remoteproc_elf_loader.c | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-)