| Message ID | 20201205082957.12544-1-zhukeqian1@huawei.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | iommu: Up front sanity check in the arm_lpae_map | expand |
On Sat, Dec 05, 2020 at 04:29:57PM +0800, Keqian Zhu wrote: > ... then we have more chance to detect wrong code logic. This could do with being a bit more explicit. Something like: Although handling a mapping request with no permissions is a trivial no-op, defer the early return until after the size/range checks so that we are consistent with other mapping requests. > Signed-off-by: Keqian Zhu <zhukeqian1@huawei.com> > --- > drivers/iommu/io-pgtable-arm.c | 8 ++++---- > 1 file changed, 4 insertions(+), 4 deletions(-) > > diff --git a/drivers/iommu/io-pgtable-arm.c b/drivers/iommu/io-pgtable-arm.c > index a7a9bc08dcd1..8ade72adab31 100644 > --- a/drivers/iommu/io-pgtable-arm.c > +++ b/drivers/iommu/io-pgtable-arm.c > @@ -444,10 +444,6 @@ static int arm_lpae_map(struct io_pgtable_ops *ops, unsigned long iova, > arm_lpae_iopte prot; > long iaext = (s64)iova >> cfg->ias; > > - /* If no access, then nothing to do */ > - if (!(iommu_prot & (IOMMU_READ | IOMMU_WRITE))) > - return 0; > - > if (WARN_ON(!size || (size & cfg->pgsize_bitmap) != size)) > return -EINVAL; > > @@ -456,6 +452,10 @@ static int arm_lpae_map(struct io_pgtable_ops *ops, unsigned long iova, > if (WARN_ON(iaext || paddr >> cfg->oas)) > return -ERANGE; > > + /* If no access, then nothing to do */ > + if (!(iommu_prot & (IOMMU_READ | IOMMU_WRITE))) > + return 0; This looks sensible to me, but please can you make the same change for io-pgtable-arm-v7s.c so that the behaviour is consistent across the two formats? Thanks, Will
Hi Will, On 2020/12/7 18:59, Will Deacon wrote: > On Sat, Dec 05, 2020 at 04:29:57PM +0800, Keqian Zhu wrote: >> ... then we have more chance to detect wrong code logic. > > This could do with being a bit more explicit. Something like: > > Although handling a mapping request with no permissions is a > trivial no-op, defer the early return until after the size/range > checks so that we are consistent with other mapping requests. This looks well, thanks. > >> Signed-off-by: Keqian Zhu <zhukeqian1@huawei.com> >> --- >> drivers/iommu/io-pgtable-arm.c | 8 ++++---- >> 1 file changed, 4 insertions(+), 4 deletions(-) >> >> diff --git a/drivers/iommu/io-pgtable-arm.c b/drivers/iommu/io-pgtable-arm.c >> index a7a9bc08dcd1..8ade72adab31 100644 >> --- a/drivers/iommu/io-pgtable-arm.c >> +++ b/drivers/iommu/io-pgtable-arm.c >> @@ -444,10 +444,6 @@ static int arm_lpae_map(struct io_pgtable_ops *ops, unsigned long iova, >> arm_lpae_iopte prot; >> long iaext = (s64)iova >> cfg->ias; >> >> - /* If no access, then nothing to do */ >> - if (!(iommu_prot & (IOMMU_READ | IOMMU_WRITE))) >> - return 0; >> - >> if (WARN_ON(!size || (size & cfg->pgsize_bitmap) != size)) >> return -EINVAL; >> >> @@ -456,6 +452,10 @@ static int arm_lpae_map(struct io_pgtable_ops *ops, unsigned long iova, >> if (WARN_ON(iaext || paddr >> cfg->oas)) >> return -ERANGE; >> >> + /* If no access, then nothing to do */ >> + if (!(iommu_prot & (IOMMU_READ | IOMMU_WRITE))) >> + return 0; > > This looks sensible to me, but please can you make the same change for > io-pgtable-arm-v7s.c so that the behaviour is consistent across the two > formats? > OK. I can do it right now. Thanks, Keqian > Thanks, > > Will > . >
On 2020-12-05 08:29, Keqian Zhu wrote: > ... then we have more chance to detect wrong code logic. I don't follow that justification - it's still the same check with the same outcome, so how does moving it have any effect on the chance to detect errors? AFAICS the only difference it would make is to make some errors *less* obvious - if a sufficiently broken caller passes an empty prot value alongside an invalid size or already-mapped address, this will now quietly hide the warnings from the more serious condition(s). Yes, it will bail out a bit faster in the specific case where the prot value is the only thing wrong, but since when do we optimise for fundamentally incorrect API usage? Robin. > Signed-off-by: Keqian Zhu <zhukeqian1@huawei.com> > --- > drivers/iommu/io-pgtable-arm.c | 8 ++++---- > 1 file changed, 4 insertions(+), 4 deletions(-) > > diff --git a/drivers/iommu/io-pgtable-arm.c b/drivers/iommu/io-pgtable-arm.c > index a7a9bc08dcd1..8ade72adab31 100644 > --- a/drivers/iommu/io-pgtable-arm.c > +++ b/drivers/iommu/io-pgtable-arm.c > @@ -444,10 +444,6 @@ static int arm_lpae_map(struct io_pgtable_ops *ops, unsigned long iova, > arm_lpae_iopte prot; > long iaext = (s64)iova >> cfg->ias; > > - /* If no access, then nothing to do */ > - if (!(iommu_prot & (IOMMU_READ | IOMMU_WRITE))) > - return 0; > - > if (WARN_ON(!size || (size & cfg->pgsize_bitmap) != size)) > return -EINVAL; > > @@ -456,6 +452,10 @@ static int arm_lpae_map(struct io_pgtable_ops *ops, unsigned long iova, > if (WARN_ON(iaext || paddr >> cfg->oas)) > return -ERANGE; > > + /* If no access, then nothing to do */ > + if (!(iommu_prot & (IOMMU_READ | IOMMU_WRITE))) > + return 0; > + > prot = arm_lpae_prot_to_pte(data, iommu_prot); > ret = __arm_lpae_map(data, iova, paddr, size, prot, lvl, ptep, gfp); > /* >
On Mon, Dec 07, 2020 at 12:01:09PM +0000, Robin Murphy wrote: > On 2020-12-05 08:29, Keqian Zhu wrote: > > ... then we have more chance to detect wrong code logic. > > I don't follow that justification - it's still the same check with the same > outcome, so how does moving it have any effect on the chance to detect > errors? > > AFAICS the only difference it would make is to make some errors *less* > obvious - if a sufficiently broken caller passes an empty prot value > alongside an invalid size or already-mapped address, this will now quietly > hide the warnings from the more serious condition(s). > > Yes, it will bail out a bit faster in the specific case where the prot value > is the only thing wrong, but since when do we optimise for fundamentally > incorrect API usage? I thought it was the other way round -- doesn't this patch move the "empty prot" check later, so we have a chance to check the size and addresses first? Will > > Signed-off-by: Keqian Zhu <zhukeqian1@huawei.com> > > --- > > drivers/iommu/io-pgtable-arm.c | 8 ++++---- > > 1 file changed, 4 insertions(+), 4 deletions(-) > > > > diff --git a/drivers/iommu/io-pgtable-arm.c b/drivers/iommu/io-pgtable-arm.c > > index a7a9bc08dcd1..8ade72adab31 100644 > > --- a/drivers/iommu/io-pgtable-arm.c > > +++ b/drivers/iommu/io-pgtable-arm.c > > @@ -444,10 +444,6 @@ static int arm_lpae_map(struct io_pgtable_ops *ops, unsigned long iova, > > arm_lpae_iopte prot; > > long iaext = (s64)iova >> cfg->ias; > > - /* If no access, then nothing to do */ > > - if (!(iommu_prot & (IOMMU_READ | IOMMU_WRITE))) > > - return 0; > > - > > if (WARN_ON(!size || (size & cfg->pgsize_bitmap) != size)) > > return -EINVAL; > > @@ -456,6 +452,10 @@ static int arm_lpae_map(struct io_pgtable_ops *ops, unsigned long iova, > > if (WARN_ON(iaext || paddr >> cfg->oas)) > > return -ERANGE; > > + /* If no access, then nothing to do */ > > + if (!(iommu_prot & (IOMMU_READ | IOMMU_WRITE))) > > + return 0; > > + > > prot = arm_lpae_prot_to_pte(data, iommu_prot); > > ret = __arm_lpae_map(data, iova, paddr, size, prot, lvl, ptep, gfp); > > /* > >
Hi, On 2020/12/7 20:05, Will Deacon wrote: > On Mon, Dec 07, 2020 at 12:01:09PM +0000, Robin Murphy wrote: >> On 2020-12-05 08:29, Keqian Zhu wrote: >>> ... then we have more chance to detect wrong code logic. >> >> I don't follow that justification - it's still the same check with the same >> outcome, so how does moving it have any effect on the chance to detect >> errors? >> >> AFAICS the only difference it would make is to make some errors *less* >> obvious - if a sufficiently broken caller passes an empty prot value >> alongside an invalid size or already-mapped address, this will now quietly >> hide the warnings from the more serious condition(s). >> >> Yes, it will bail out a bit faster in the specific case where the prot value >> is the only thing wrong, but since when do we optimise for fundamentally >> incorrect API usage? > > I thought it was the other way round -- doesn't this patch move the "empty > prot" check later, so we have a chance to check the size and addresses > first? Yes, this is my original idea. For that we treat iommu_prot with no permission as success at early start, defer this early return can expose hidden errors. Thanks, Keqian > > Will > >>> Signed-off-by: Keqian Zhu <zhukeqian1@huawei.com> >>> --- >>> drivers/iommu/io-pgtable-arm.c | 8 ++++---- >>> 1 file changed, 4 insertions(+), 4 deletions(-) >>> >>> diff --git a/drivers/iommu/io-pgtable-arm.c b/drivers/iommu/io-pgtable-arm.c >>> index a7a9bc08dcd1..8ade72adab31 100644 >>> --- a/drivers/iommu/io-pgtable-arm.c >>> +++ b/drivers/iommu/io-pgtable-arm.c >>> @@ -444,10 +444,6 @@ static int arm_lpae_map(struct io_pgtable_ops *ops, unsigned long iova, >>> arm_lpae_iopte prot; >>> long iaext = (s64)iova >> cfg->ias; >>> - /* If no access, then nothing to do */ >>> - if (!(iommu_prot & (IOMMU_READ | IOMMU_WRITE))) >>> - return 0; >>> - >>> if (WARN_ON(!size || (size & cfg->pgsize_bitmap) != size)) >>> return -EINVAL; >>> @@ -456,6 +452,10 @@ static int arm_lpae_map(struct io_pgtable_ops *ops, unsigned long iova, >>> if (WARN_ON(iaext || paddr >> cfg->oas)) >>> return -ERANGE; >>> + /* If no access, then nothing to do */ >>> + if (!(iommu_prot & (IOMMU_READ | IOMMU_WRITE))) >>> + return 0; >>> + >>> prot = arm_lpae_prot_to_pte(data, iommu_prot); >>> ret = __arm_lpae_map(data, iova, paddr, size, prot, lvl, ptep, gfp); >>> /* >>> > . >
On 2020-12-07 12:15, zhukeqian wrote: > Hi, > > On 2020/12/7 20:05, Will Deacon wrote: >> On Mon, Dec 07, 2020 at 12:01:09PM +0000, Robin Murphy wrote: >>> On 2020-12-05 08:29, Keqian Zhu wrote: >>>> ... then we have more chance to detect wrong code logic. >>> >>> I don't follow that justification - it's still the same check with the same >>> outcome, so how does moving it have any effect on the chance to detect >>> errors? > >>> >>> AFAICS the only difference it would make is to make some errors *less* >>> obvious - if a sufficiently broken caller passes an empty prot value >>> alongside an invalid size or already-mapped address, this will now quietly >>> hide the warnings from the more serious condition(s). >>> >>> Yes, it will bail out a bit faster in the specific case where the prot value >>> is the only thing wrong, but since when do we optimise for fundamentally >>> incorrect API usage? >> >> I thought it was the other way round -- doesn't this patch move the "empty >> prot" check later, so we have a chance to check the size and addresses >> first? > > Yes, this is my original idea. > For that we treat iommu_prot with no permission as success at early start, defer > this early return can expose hidden errors. ...oh dear, my apologies. I've just had a week off and apparently in that time I lost the ability to read :( I was somehow convinced that the existing check happened at the point where we go to install the PTE, and this patch was moving it earlier. Looking at the whole code in context now I see I got it completely backwards. Sorry for being an idiot. I guess that only goes to show that a more descriptive commit message would definitely be a good thing :) Robin. > > Thanks, > Keqian >> >> Will >> >>>> Signed-off-by: Keqian Zhu <zhukeqian1@huawei.com> >>>> --- >>>> drivers/iommu/io-pgtable-arm.c | 8 ++++---- >>>> 1 file changed, 4 insertions(+), 4 deletions(-) >>>> >>>> diff --git a/drivers/iommu/io-pgtable-arm.c b/drivers/iommu/io-pgtable-arm.c >>>> index a7a9bc08dcd1..8ade72adab31 100644 >>>> --- a/drivers/iommu/io-pgtable-arm.c >>>> +++ b/drivers/iommu/io-pgtable-arm.c >>>> @@ -444,10 +444,6 @@ static int arm_lpae_map(struct io_pgtable_ops *ops, unsigned long iova, >>>> arm_lpae_iopte prot; >>>> long iaext = (s64)iova >> cfg->ias; >>>> - /* If no access, then nothing to do */ >>>> - if (!(iommu_prot & (IOMMU_READ | IOMMU_WRITE))) >>>> - return 0; >>>> - >>>> if (WARN_ON(!size || (size & cfg->pgsize_bitmap) != size)) >>>> return -EINVAL; >>>> @@ -456,6 +452,10 @@ static int arm_lpae_map(struct io_pgtable_ops *ops, unsigned long iova, >>>> if (WARN_ON(iaext || paddr >> cfg->oas)) >>>> return -ERANGE; >>>> + /* If no access, then nothing to do */ >>>> + if (!(iommu_prot & (IOMMU_READ | IOMMU_WRITE))) >>>> + return 0; >>>> + >>>> prot = arm_lpae_prot_to_pte(data, iommu_prot); >>>> ret = __arm_lpae_map(data, iova, paddr, size, prot, lvl, ptep, gfp); >>>> /* >>>> >> . >> > _______________________________________________ > iommu mailing list > iommu@lists.linux-foundation.org > https://lists.linuxfoundation.org/mailman/listinfo/iommu >
Hi Robin, On 2020/12/7 20:46, Robin Murphy wrote: > On 2020-12-07 12:15, zhukeqian wrote: >> Hi, >> >> On 2020/12/7 20:05, Will Deacon wrote: >>> On Mon, Dec 07, 2020 at 12:01:09PM +0000, Robin Murphy wrote: >>>> On 2020-12-05 08:29, Keqian Zhu wrote: >>>>> ... then we have more chance to detect wrong code logic. >>>> >>>> I don't follow that justification - it's still the same check with the same >>>> outcome, so how does moving it have any effect on the chance to detect >>>> errors? >> >>>> >>>> AFAICS the only difference it would make is to make some errors *less* >>>> obvious - if a sufficiently broken caller passes an empty prot value >>>> alongside an invalid size or already-mapped address, this will now quietly >>>> hide the warnings from the more serious condition(s). >>>> >>>> Yes, it will bail out a bit faster in the specific case where the prot value >>>> is the only thing wrong, but since when do we optimise for fundamentally >>>> incorrect API usage? >>> >>> I thought it was the other way round -- doesn't this patch move the "empty >>> prot" check later, so we have a chance to check the size and addresses >>> first? >> >> Yes, this is my original idea. >> For that we treat iommu_prot with no permission as success at early start, defer >> this early return can expose hidden errors. > > ...oh dear, my apologies. I've just had a week off and apparently in that time I lost the ability to read :( > > I was somehow convinced that the existing check happened at the point where we go to install the PTE, and this patch was moving it earlier. Looking at the whole code in context now I see I got it completely backwards. Sorry for being an idiot. > I see :-) I should make a more descriptive commit message. > I guess that only goes to show that a more descriptive commit message would definitely be a good thing :) > I have adapted Will's suggestion and sent v2, please check whether it is OK to you? Cheers, Keqian [...] >> _______________________________________________ >> iommu mailing list >> iommu@lists.linux-foundation.org >> https://lists.linuxfoundation.org/mailman/listinfo/iommu >> > . >
diff --git a/drivers/iommu/io-pgtable-arm.c b/drivers/iommu/io-pgtable-arm.c index a7a9bc08dcd1..8ade72adab31 100644 --- a/drivers/iommu/io-pgtable-arm.c +++ b/drivers/iommu/io-pgtable-arm.c @@ -444,10 +444,6 @@ static int arm_lpae_map(struct io_pgtable_ops *ops, unsigned long iova, arm_lpae_iopte prot; long iaext = (s64)iova >> cfg->ias; - /* If no access, then nothing to do */ - if (!(iommu_prot & (IOMMU_READ | IOMMU_WRITE))) - return 0; - if (WARN_ON(!size || (size & cfg->pgsize_bitmap) != size)) return -EINVAL; @@ -456,6 +452,10 @@ static int arm_lpae_map(struct io_pgtable_ops *ops, unsigned long iova, if (WARN_ON(iaext || paddr >> cfg->oas)) return -ERANGE; + /* If no access, then nothing to do */ + if (!(iommu_prot & (IOMMU_READ | IOMMU_WRITE))) + return 0; + prot = arm_lpae_prot_to_pte(data, iommu_prot); ret = __arm_lpae_map(data, iova, paddr, size, prot, lvl, ptep, gfp); /*
... then we have more chance to detect wrong code logic. Signed-off-by: Keqian Zhu <zhukeqian1@huawei.com> --- drivers/iommu/io-pgtable-arm.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)