| Message ID | 20181220093221.6624-1-yuval.shaia@oracle.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | [v1] IB/core: Warn if can't pin memory | expand |
On 20-Dec-18 11:32, Yuval Shaia wrote: > When memory registration is needed for page count which exceeds the > allowed lock limit the function exits with an error but no error is > printed. > Since this is a case that system administrator can fix (rlimit etc) we > need to give an alert. > This is not a data patch so performance penalty is not a consideration. > > Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com> > --- > v0 -> v1: > * Accept Leon's comment pr_err to pr_err_ratelimited > --- > drivers/infiniband/core/umem.c | 2 ++ > 1 file changed, 2 insertions(+) > > diff --git a/drivers/infiniband/core/umem.c b/drivers/infiniband/core/umem.c > index c6144df47ea4..73f85cc35418 100644 > --- a/drivers/infiniband/core/umem.c > +++ b/drivers/infiniband/core/umem.c > @@ -164,6 +164,8 @@ struct ib_umem *ib_umem_get(struct ib_ucontext *context, unsigned long addr, > if (check_add_overflow(mm->pinned_vm, npages, &new_pinned) || > (new_pinned > lock_limit && !capable(CAP_IPC_LOCK))) { > up_write(&mm->mmap_sem); > + pr_err_ratelimited("%s: Can't pin memory, requested %d, limit %d, ipc_lock off\n", > + _func__, new_pinned, lock_limit); Shouldn't these %d be %lu? > ret = -ENOMEM; > goto out; > } >
On Thu, Dec 20, 2018 at 11:39:54AM +0200, Gal Pressman wrote: > On 20-Dec-18 11:32, Yuval Shaia wrote: > > When memory registration is needed for page count which exceeds the > > allowed lock limit the function exits with an error but no error is > > printed. > > Since this is a case that system administrator can fix (rlimit etc) we > > need to give an alert. > > This is not a data patch so performance penalty is not a consideration. > > > > Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com> > > --- > > v0 -> v1: > > * Accept Leon's comment pr_err to pr_err_ratelimited > > --- > > drivers/infiniband/core/umem.c | 2 ++ > > 1 file changed, 2 insertions(+) > > > > diff --git a/drivers/infiniband/core/umem.c b/drivers/infiniband/core/umem.c > > index c6144df47ea4..73f85cc35418 100644 > > --- a/drivers/infiniband/core/umem.c > > +++ b/drivers/infiniband/core/umem.c > > @@ -164,6 +164,8 @@ struct ib_umem *ib_umem_get(struct ib_ucontext *context, unsigned long addr, > > if (check_add_overflow(mm->pinned_vm, npages, &new_pinned) || > > (new_pinned > lock_limit && !capable(CAP_IPC_LOCK))) { > > up_write(&mm->mmap_sem); > > + pr_err_ratelimited("%s: Can't pin memory, requested %d, limit %d, ipc_lock off\n", > > + _func__, new_pinned, lock_limit); > > Shouldn't these %d be %lu? Make sense, missed that, will fix > > > ret = -ENOMEM; > > goto out; > > } > > >
On 20-Dec-18 11:45, Yuval Shaia wrote: > On Thu, Dec 20, 2018 at 11:39:54AM +0200, Gal Pressman wrote: >> On 20-Dec-18 11:32, Yuval Shaia wrote: >>> When memory registration is needed for page count which exceeds the >>> allowed lock limit the function exits with an error but no error is >>> printed. >>> Since this is a case that system administrator can fix (rlimit etc) we >>> need to give an alert. >>> This is not a data patch so performance penalty is not a consideration. patch should be path. >>> >>> Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com> >>> --- >>> v0 -> v1: >>> * Accept Leon's comment pr_err to pr_err_ratelimited >>> --- >>> drivers/infiniband/core/umem.c | 2 ++ >>> 1 file changed, 2 insertions(+) >>> >>> diff --git a/drivers/infiniband/core/umem.c b/drivers/infiniband/core/umem.c >>> index c6144df47ea4..73f85cc35418 100644 >>> --- a/drivers/infiniband/core/umem.c >>> +++ b/drivers/infiniband/core/umem.c >>> @@ -164,6 +164,8 @@ struct ib_umem *ib_umem_get(struct ib_ucontext *context, unsigned long addr, >>> if (check_add_overflow(mm->pinned_vm, npages, &new_pinned) || >>> (new_pinned > lock_limit && !capable(CAP_IPC_LOCK))) { >>> up_write(&mm->mmap_sem); >>> + pr_err_ratelimited("%s: Can't pin memory, requested %d, limit %d, ipc_lock off\n", >>> + _func__, new_pinned, lock_limit); >> >> Shouldn't these %d be %lu? > > Make sense, missed that, will fix Also, looks like something bad happened to __func__. > >> >>> ret = -ENOMEM; >>> goto out; >>> } >>> >>
diff --git a/drivers/infiniband/core/umem.c b/drivers/infiniband/core/umem.c index c6144df47ea4..73f85cc35418 100644 --- a/drivers/infiniband/core/umem.c +++ b/drivers/infiniband/core/umem.c @@ -164,6 +164,8 @@ struct ib_umem *ib_umem_get(struct ib_ucontext *context, unsigned long addr, if (check_add_overflow(mm->pinned_vm, npages, &new_pinned) || (new_pinned > lock_limit && !capable(CAP_IPC_LOCK))) { up_write(&mm->mmap_sem); + pr_err_ratelimited("%s: Can't pin memory, requested %d, limit %d, ipc_lock off\n", + _func__, new_pinned, lock_limit); ret = -ENOMEM; goto out; }
When memory registration is needed for page count which exceeds the allowed lock limit the function exits with an error but no error is printed. Since this is a case that system administrator can fix (rlimit etc) we need to give an alert. This is not a data patch so performance penalty is not a consideration. Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com> --- v0 -> v1: * Accept Leon's comment pr_err to pr_err_ratelimited --- drivers/infiniband/core/umem.c | 2 ++ 1 file changed, 2 insertions(+)