| Message ID | 20210412064150.40064-1-galpress@amazon.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | [for-next,v3] RDMA/nldev: Add copy-on-fork attribute to get sys command | expand |
On Mon, Apr 12, 2021 at 09:41:50AM +0300, Gal Pressman wrote: > The new attribute indicates that the kernel copies DMA pages on fork, > hence libibverbs' fork support through madvise and MADV_DONTFORK is not > needed. > > The introduced attribute is always reported as supported since the > kernel has the patch that added the copy-on-fork behavior. This allows > the userspace library to identify older vs newer kernel versions. > Extra care should be taken when backporting this patch as it relies on > the fact that the copy-on-fork patch is merged, hence no check for > support is added. > > Don't backport this patch unless you also have the following series: > 70e806e4e645 ("mm: Do early cow for pinned pages during fork() for ptes") > and 4eae4efa2c29 ("hugetlb: do early cow when page pinned on src mm"). > > Fixes: 70e806e4e645 ("mm: Do early cow for pinned pages during fork() for ptes") > Fixes: 4eae4efa2c29 ("hugetlb: do early cow when page pinned on src mm") > Signed-off-by: Gal Pressman <galpress@amazon.com> > --- > PR was sent: > https://github.com/linux-rdma/rdma-core/pull/975 > > Changelog - > v2->v3: https://lore.kernel.org/linux-rdma/21317d2c-9a8e-0dd7-3678-d2933c5053c4@amazon.com/ > * Remove check if copy-on-fork attribute was provided from nldev_set_sys_set_doit() > > v1->v2: https://lore.kernel.org/linux-rdma/20210405114722.98904-1-galpress@amazon.com/ > * Remove nla_put_u8() return value check > * Add commit hashes to commit message and code comment > --- > drivers/infiniband/core/nldev.c | 11 +++++++++++ > include/uapi/rdma/rdma_netlink.h | 2 ++ > 2 files changed, 13 insertions(+) > > diff --git a/drivers/infiniband/core/nldev.c b/drivers/infiniband/core/nldev.c > index b8dc002a2478..4889e06a581a 100644 > --- a/drivers/infiniband/core/nldev.c > +++ b/drivers/infiniband/core/nldev.c > @@ -146,6 +146,7 @@ static const struct nla_policy nldev_policy[RDMA_NLDEV_ATTR_MAX] = { > [RDMA_NLDEV_ATTR_UVERBS_DRIVER_ID] = { .type = NLA_U32 }, > [RDMA_NLDEV_NET_NS_FD] = { .type = NLA_U32 }, > [RDMA_NLDEV_SYS_ATTR_NETNS_MODE] = { .type = NLA_U8 }, > + [RDMA_NLDEV_SYS_ATTR_COPY_ON_FORK] = { .type = NLA_U8 }, > }; > > static int put_driver_name_print_type(struct sk_buff *msg, const char *name, > @@ -1697,6 +1698,16 @@ static int nldev_sys_get_doit(struct sk_buff *skb, struct nlmsghdr *nlh, > nlmsg_free(msg); > return err; > } > + > + /* > + * Copy-on-fork is supported. > + * See commits: > + * 70e806e4e645 ("mm: Do early cow for pinned pages during fork() for ptes") > + * 4eae4efa2c29 ("hugetlb: do early cow when page pinned on src mm") > + * for more details. Don't backport this without them. > + */ > + nla_put_u8(msg, RDMA_NLDEV_SYS_ATTR_COPY_ON_FORK, 1); > + Nit, it is good to write here that we don't check nla_put_u8() on purpose. Thanks, Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
On 12/04/2021 10:51, Leon Romanovsky wrote: > On Mon, Apr 12, 2021 at 09:41:50AM +0300, Gal Pressman wrote: >> The new attribute indicates that the kernel copies DMA pages on fork, >> hence libibverbs' fork support through madvise and MADV_DONTFORK is not >> needed. >> >> The introduced attribute is always reported as supported since the >> kernel has the patch that added the copy-on-fork behavior. This allows >> the userspace library to identify older vs newer kernel versions. >> Extra care should be taken when backporting this patch as it relies on >> the fact that the copy-on-fork patch is merged, hence no check for >> support is added. >> >> Don't backport this patch unless you also have the following series: >> 70e806e4e645 ("mm: Do early cow for pinned pages during fork() for ptes") >> and 4eae4efa2c29 ("hugetlb: do early cow when page pinned on src mm"). >> >> Fixes: 70e806e4e645 ("mm: Do early cow for pinned pages during fork() for ptes") >> Fixes: 4eae4efa2c29 ("hugetlb: do early cow when page pinned on src mm") >> Signed-off-by: Gal Pressman <galpress@amazon.com> >> --- >> PR was sent: >> https://github.com/linux-rdma/rdma-core/pull/975 >> >> Changelog - >> v2->v3: https://lore.kernel.org/linux-rdma/21317d2c-9a8e-0dd7-3678-d2933c5053c4@amazon.com/ >> * Remove check if copy-on-fork attribute was provided from nldev_set_sys_set_doit() >> >> v1->v2: https://lore.kernel.org/linux-rdma/20210405114722.98904-1-galpress@amazon.com/ >> * Remove nla_put_u8() return value check >> * Add commit hashes to commit message and code comment >> --- >> drivers/infiniband/core/nldev.c | 11 +++++++++++ >> include/uapi/rdma/rdma_netlink.h | 2 ++ >> 2 files changed, 13 insertions(+) >> >> diff --git a/drivers/infiniband/core/nldev.c b/drivers/infiniband/core/nldev.c >> index b8dc002a2478..4889e06a581a 100644 >> --- a/drivers/infiniband/core/nldev.c >> +++ b/drivers/infiniband/core/nldev.c >> @@ -146,6 +146,7 @@ static const struct nla_policy nldev_policy[RDMA_NLDEV_ATTR_MAX] = { >> [RDMA_NLDEV_ATTR_UVERBS_DRIVER_ID] = { .type = NLA_U32 }, >> [RDMA_NLDEV_NET_NS_FD] = { .type = NLA_U32 }, >> [RDMA_NLDEV_SYS_ATTR_NETNS_MODE] = { .type = NLA_U8 }, >> + [RDMA_NLDEV_SYS_ATTR_COPY_ON_FORK] = { .type = NLA_U8 }, >> }; >> >> static int put_driver_name_print_type(struct sk_buff *msg, const char *name, >> @@ -1697,6 +1698,16 @@ static int nldev_sys_get_doit(struct sk_buff *skb, struct nlmsghdr *nlh, >> nlmsg_free(msg); >> return err; >> } >> + >> + /* >> + * Copy-on-fork is supported. >> + * See commits: >> + * 70e806e4e645 ("mm: Do early cow for pinned pages during fork() for ptes") >> + * 4eae4efa2c29 ("hugetlb: do early cow when page pinned on src mm") >> + * for more details. Don't backport this without them. >> + */ >> + nla_put_u8(msg, RDMA_NLDEV_SYS_ATTR_COPY_ON_FORK, 1); >> + > > Nit, it is good to write here that we don't check nla_put_u8() on purpose. > > Thanks, > Reviewed-by: Leon Romanovsky <leonro@nvidia.com> Thanks Leon!
diff --git a/drivers/infiniband/core/nldev.c b/drivers/infiniband/core/nldev.c index b8dc002a2478..4889e06a581a 100644 --- a/drivers/infiniband/core/nldev.c +++ b/drivers/infiniband/core/nldev.c @@ -146,6 +146,7 @@ static const struct nla_policy nldev_policy[RDMA_NLDEV_ATTR_MAX] = { [RDMA_NLDEV_ATTR_UVERBS_DRIVER_ID] = { .type = NLA_U32 }, [RDMA_NLDEV_NET_NS_FD] = { .type = NLA_U32 }, [RDMA_NLDEV_SYS_ATTR_NETNS_MODE] = { .type = NLA_U8 }, + [RDMA_NLDEV_SYS_ATTR_COPY_ON_FORK] = { .type = NLA_U8 }, }; static int put_driver_name_print_type(struct sk_buff *msg, const char *name, @@ -1697,6 +1698,16 @@ static int nldev_sys_get_doit(struct sk_buff *skb, struct nlmsghdr *nlh, nlmsg_free(msg); return err; } + + /* + * Copy-on-fork is supported. + * See commits: + * 70e806e4e645 ("mm: Do early cow for pinned pages during fork() for ptes") + * 4eae4efa2c29 ("hugetlb: do early cow when page pinned on src mm") + * for more details. Don't backport this without them. + */ + nla_put_u8(msg, RDMA_NLDEV_SYS_ATTR_COPY_ON_FORK, 1); + nlmsg_end(msg, nlh); return rdma_nl_unicast(sock_net(skb->sk), msg, NETLINK_CB(skb).portid); } diff --git a/include/uapi/rdma/rdma_netlink.h b/include/uapi/rdma/rdma_netlink.h index d2f5b8396243..342c9db5b3c1 100644 --- a/include/uapi/rdma/rdma_netlink.h +++ b/include/uapi/rdma/rdma_netlink.h @@ -533,6 +533,8 @@ enum rdma_nldev_attr { RDMA_NLDEV_ATTR_RES_RAW, /* binary */ + RDMA_NLDEV_SYS_ATTR_COPY_ON_FORK, /* u8 */ + /* * Always the end */
The new attribute indicates that the kernel copies DMA pages on fork, hence libibverbs' fork support through madvise and MADV_DONTFORK is not needed. The introduced attribute is always reported as supported since the kernel has the patch that added the copy-on-fork behavior. This allows the userspace library to identify older vs newer kernel versions. Extra care should be taken when backporting this patch as it relies on the fact that the copy-on-fork patch is merged, hence no check for support is added. Don't backport this patch unless you also have the following series: 70e806e4e645 ("mm: Do early cow for pinned pages during fork() for ptes") and 4eae4efa2c29 ("hugetlb: do early cow when page pinned on src mm"). Fixes: 70e806e4e645 ("mm: Do early cow for pinned pages during fork() for ptes") Fixes: 4eae4efa2c29 ("hugetlb: do early cow when page pinned on src mm") Signed-off-by: Gal Pressman <galpress@amazon.com> --- PR was sent: https://github.com/linux-rdma/rdma-core/pull/975 Changelog - v2->v3: https://lore.kernel.org/linux-rdma/21317d2c-9a8e-0dd7-3678-d2933c5053c4@amazon.com/ * Remove check if copy-on-fork attribute was provided from nldev_set_sys_set_doit() v1->v2: https://lore.kernel.org/linux-rdma/20210405114722.98904-1-galpress@amazon.com/ * Remove nla_put_u8() return value check * Add commit hashes to commit message and code comment --- drivers/infiniband/core/nldev.c | 11 +++++++++++ include/uapi/rdma/rdma_netlink.h | 2 ++ 2 files changed, 13 insertions(+)