| Message ID | 1362929953-63785-1-git-send-email-tim.gardner@canonical.com |
|---|---|
| State | Not Applicable, archived |
| Delegated to: | David Miller |
| Headers | show |
This is the result of 2773395b34883fe54418de188733a63bb38e0ad6. Steve might want to weigh in on this since it was done for performance reasons. Tom On 3/10/13 10:39 AM, Tim Gardner wrote: > rpcrdma_register_default_external() is several frames into > the call stack which goes deeper yet. You run the risk of stack > corruption by declaring such a large automatic variable, > so dynamically allocate the array of 'struct ib_phys_buf' objects in > order to silence the frame-larger-than warning. > > net/sunrpc/xprtrdma/verbs.c: In function 'rpcrdma_register_default_external': > net/sunrpc/xprtrdma/verbs.c:1774:1: warning: the frame size of 1056 bytes is larger than 1024 bytes [-Wframe-larger-than=] > > gcc version 4.6.3 > > Cc: Trond Myklebust <Trond.Myklebust@netapp.com> > Cc: "J. Bruce Fields" <bfields@fieldses.org> > Cc: "David S. Miller" <davem@davemloft.net> > Cc: Tom Tucker <tom@ogc.us> > Cc: Haggai Eran <haggaie@mellanox.com> > Cc: Or Gerlitz <ogerlitz@mellanox.com> > Cc: Shani Michaeli <shanim@mellanox.com> > Cc: linux-nfs@vger.kernel.org > Cc: netdev@vger.kernel.org > Signed-off-by: Tim Gardner <tim.gardner@canonical.com> > --- > net/sunrpc/xprtrdma/verbs.c | 7 ++++++- > 1 file changed, 6 insertions(+), 1 deletion(-) > > diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c > index 93726560..0916467 100644 > --- a/net/sunrpc/xprtrdma/verbs.c > +++ b/net/sunrpc/xprtrdma/verbs.c > @@ -1736,9 +1736,13 @@ rpcrdma_register_default_external(struct rpcrdma_mr_seg *seg, > int mem_priv = (writing ? IB_ACCESS_REMOTE_WRITE : > IB_ACCESS_REMOTE_READ); > struct rpcrdma_mr_seg *seg1 = seg; > - struct ib_phys_buf ipb[RPCRDMA_MAX_DATA_SEGS]; > + struct ib_phys_buf *ipb; > int len, i, rc = 0; > > + ipb = kmalloc(sizeof(*ipb) * RPCRDMA_MAX_DATA_SEGS, GFP_KERNEL); > + if (!ipb) > + return -ENOMEM; > + > if (*nsegs > RPCRDMA_MAX_DATA_SEGS) > *nsegs = RPCRDMA_MAX_DATA_SEGS; > for (len = 0, i = 0; i < *nsegs;) { > @@ -1770,6 +1774,7 @@ rpcrdma_register_default_external(struct rpcrdma_mr_seg *seg, > seg1->mr_len = len; > } > *nsegs = i; > + kfree(ipb); > return rc; > } > -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 03/10/2013 11:16 AM, Tom Tucker wrote: > > This is the result of 2773395b34883fe54418de188733a63bb38e0ad6. Steve > might want to weigh in on this since it was done for performance reasons. > > Tom > It seems to me that the cost of a kmalloc()/kfree() is negligible compared to network speeds. rtg
On Sun, Mar 10, 2013 at 09:39:13AM -0600, Tim Gardner wrote: > rpcrdma_register_default_external() is several frames into > the call stack which goes deeper yet. You run the risk of stack > corruption by declaring such a large automatic variable, > so dynamically allocate the array of 'struct ib_phys_buf' objects in > order to silence the frame-larger-than warning. > > net/sunrpc/xprtrdma/verbs.c: In function 'rpcrdma_register_default_external': > net/sunrpc/xprtrdma/verbs.c:1774:1: warning: the frame size of 1056 bytes is larger than 1024 bytes [-Wframe-larger-than=] > > gcc version 4.6.3 > > Cc: Trond Myklebust <Trond.Myklebust@netapp.com> > Cc: "J. Bruce Fields" <bfields@fieldses.org> > Cc: "David S. Miller" <davem@davemloft.net> > Cc: Tom Tucker <tom@ogc.us> > Cc: Haggai Eran <haggaie@mellanox.com> > Cc: Or Gerlitz <ogerlitz@mellanox.com> > Cc: Shani Michaeli <shanim@mellanox.com> > Cc: linux-nfs@vger.kernel.org > Cc: netdev@vger.kernel.org > Signed-off-by: Tim Gardner <tim.gardner@canonical.com> > --- > net/sunrpc/xprtrdma/verbs.c | 7 ++++++- > 1 file changed, 6 insertions(+), 1 deletion(-) > > diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c > index 93726560..0916467 100644 > --- a/net/sunrpc/xprtrdma/verbs.c > +++ b/net/sunrpc/xprtrdma/verbs.c > @@ -1736,9 +1736,13 @@ rpcrdma_register_default_external(struct rpcrdma_mr_seg *seg, > int mem_priv = (writing ? IB_ACCESS_REMOTE_WRITE : > IB_ACCESS_REMOTE_READ); > struct rpcrdma_mr_seg *seg1 = seg; > - struct ib_phys_buf ipb[RPCRDMA_MAX_DATA_SEGS]; > + struct ib_phys_buf *ipb; > int len, i, rc = 0; > > + ipb = kmalloc(sizeof(*ipb) * RPCRDMA_MAX_DATA_SEGS, GFP_KERNEL); Have you checked that this occurs in a context where allocations are OK? Checking very quickly through the callers I can't see any spinlocks or anything, but I also don't see any other allocations. Assuming this is just in rpciod context.... Trond's the authority, but I think we generally try to avoid allocations here, or make them GFP_NOFS if we must. Would it be possible to allocate this array as part of the rpcrdma_req? --b. > + if (!ipb) > + return -ENOMEM; > + > if (*nsegs > RPCRDMA_MAX_DATA_SEGS) > *nsegs = RPCRDMA_MAX_DATA_SEGS; > for (len = 0, i = 0; i < *nsegs;) { > @@ -1770,6 +1774,7 @@ rpcrdma_register_default_external(struct rpcrdma_mr_seg *seg, > seg1->mr_len = len; > } > *nsegs = i; > + kfree(ipb); > return rc; > } > > -- > 1.7.9.5 > -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c index 93726560..0916467 100644 --- a/net/sunrpc/xprtrdma/verbs.c +++ b/net/sunrpc/xprtrdma/verbs.c @@ -1736,9 +1736,13 @@ rpcrdma_register_default_external(struct rpcrdma_mr_seg *seg, int mem_priv = (writing ? IB_ACCESS_REMOTE_WRITE : IB_ACCESS_REMOTE_READ); struct rpcrdma_mr_seg *seg1 = seg; - struct ib_phys_buf ipb[RPCRDMA_MAX_DATA_SEGS]; + struct ib_phys_buf *ipb; int len, i, rc = 0; + ipb = kmalloc(sizeof(*ipb) * RPCRDMA_MAX_DATA_SEGS, GFP_KERNEL); + if (!ipb) + return -ENOMEM; + if (*nsegs > RPCRDMA_MAX_DATA_SEGS) *nsegs = RPCRDMA_MAX_DATA_SEGS; for (len = 0, i = 0; i < *nsegs;) { @@ -1770,6 +1774,7 @@ rpcrdma_register_default_external(struct rpcrdma_mr_seg *seg, seg1->mr_len = len; } *nsegs = i; + kfree(ipb); return rc; }
rpcrdma_register_default_external() is several frames into the call stack which goes deeper yet. You run the risk of stack corruption by declaring such a large automatic variable, so dynamically allocate the array of 'struct ib_phys_buf' objects in order to silence the frame-larger-than warning. net/sunrpc/xprtrdma/verbs.c: In function 'rpcrdma_register_default_external': net/sunrpc/xprtrdma/verbs.c:1774:1: warning: the frame size of 1056 bytes is larger than 1024 bytes [-Wframe-larger-than=] gcc version 4.6.3 Cc: Trond Myklebust <Trond.Myklebust@netapp.com> Cc: "J. Bruce Fields" <bfields@fieldses.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Tom Tucker <tom@ogc.us> Cc: Haggai Eran <haggaie@mellanox.com> Cc: Or Gerlitz <ogerlitz@mellanox.com> Cc: Shani Michaeli <shanim@mellanox.com> Cc: linux-nfs@vger.kernel.org Cc: netdev@vger.kernel.org Signed-off-by: Tim Gardner <tim.gardner@canonical.com> --- net/sunrpc/xprtrdma/verbs.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)