* [PATCH 6 of 53] ipath - forbid creation of AH with DLID of 0
From: Bryan O'Sullivan @ 2006-05-12 23:42 UTC (permalink / raw)
To: rdreier; +Cc: openib-general, linux-kernel
In-Reply-To: <patchbomb.1147477365@eng-12.pathscale.com>
Don't allow an AH to be created with a DLID of 0.
Signed-off-by: Bryan O'Sullivan <bos@pathscale.com>
diff -r db56c0ab6a64 -r def81ab50644 drivers/infiniband/hw/ipath/ipath_verbs.c
--- a/drivers/infiniband/hw/ipath/ipath_verbs.c Fri May 12 15:55:27 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.c Fri May 12 15:55:27 2006 -0700
@@ -810,6 +810,11 @@ static struct ib_ah *ipath_create_ah(str
if (ah_attr->dlid >= IPS_MULTICAST_LID_BASE &&
ah_attr->dlid != IPS_PERMISSIVE_LID &&
!(ah_attr->ah_flags & IB_AH_GRH)) {
+ ret = ERR_PTR(-EINVAL);
+ goto bail;
+ }
+
+ if (ah_attr->dlid == 0) {
ret = ERR_PTR(-EINVAL);
goto bail;
}
^ permalink raw reply
* [PATCH 8 of 53] ipath - cap number of CQEs
From: Bryan O'Sullivan @ 2006-05-12 23:42 UTC (permalink / raw)
To: rdreier; +Cc: openib-general, linux-kernel
In-Reply-To: <patchbomb.1147477365@eng-12.pathscale.com>
Cap the number of CQEs. Not a real limitation for us, but expected by
the verbs code.
Signed-off-by: Bryan O'Sullivan <bos@pathscale.com>
diff -r e823378bd19c -r 1d3e85454b53 drivers/infiniband/hw/ipath/ipath_cq.c
--- a/drivers/infiniband/hw/ipath/ipath_cq.c Fri May 12 15:55:27 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_cq.c Fri May 12 15:55:27 2006 -0700
@@ -161,6 +161,11 @@ struct ib_cq *ipath_create_cq(struct ib_
struct ib_wc *wc;
struct ib_cq *ret;
+ if (entries > ib_ipath_max_cqe) {
+ ret = ERR_PTR(-EINVAL);
+ goto bail;
+ }
+
/*
* Need to use vmalloc() if we want to support large #s of
* entries.
diff -r e823378bd19c -r 1d3e85454b53 drivers/infiniband/hw/ipath/ipath_verbs.c
--- a/drivers/infiniband/hw/ipath/ipath_verbs.c Fri May 12 15:55:27 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.c Fri May 12 15:55:27 2006 -0700
@@ -64,6 +64,11 @@ module_param_named(max_ahs, ib_ipath_max
module_param_named(max_ahs, ib_ipath_max_ahs, uint, S_IWUSR | S_IRUGO);
MODULE_PARM_DESC(max_ahs,
"Maximum number of address handles to support");
+
+unsigned int ib_ipath_max_cqe = 0xFFFF;
+module_param_named(max_cqe, ib_ipath_max_cqe, uint, S_IWUSR | S_IRUGO);
+MODULE_PARM_DESC(max_cqe,
+ "Maximum number of completion queue entries to support");
MODULE_LICENSE("GPL");
MODULE_AUTHOR("PathScale <support@pathscale.com>");
@@ -598,7 +603,7 @@ static int ipath_query_device(struct ib_
props->max_sge = 255;
props->max_cq = 0xffff;
props->max_ah = ib_ipath_max_ahs;
- props->max_cqe = 0xffff;
+ props->max_cqe = ib_ipath_max_cqe;
props->max_mr = dev->lk_table.max;
props->max_pd = ib_ipath_max_pds;
props->max_qp_rd_atom = 1;
diff -r e823378bd19c -r 1d3e85454b53 drivers/infiniband/hw/ipath/ipath_verbs.h
--- a/drivers/infiniband/hw/ipath/ipath_verbs.h Fri May 12 15:55:27 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.h Fri May 12 15:55:27 2006 -0700
@@ -690,6 +690,8 @@ extern const int ib_ipath_state_ops[];
extern unsigned int ib_ipath_lkey_table_size;
+extern unsigned int ib_ipath_max_cqe;
+
extern const u32 ib_ipath_rnr_table[];
#endif /* IPATH_VERBS_H */
^ permalink raw reply
* [PATCH 11 of 53] ipath - don't modify QP if changes fail
From: Bryan O'Sullivan @ 2006-05-12 23:42 UTC (permalink / raw)
To: rdreier; +Cc: openib-general, linux-kernel
In-Reply-To: <patchbomb.1147477365@eng-12.pathscale.com>
Make sure modify_qp won't modify the QP if any of the changes failed.
Signed-off-by: Bryan O'Sullivan <bos@pathscale.com>
diff -r 2fea0d127a41 -r cc6d7f2537b2 drivers/infiniband/hw/ipath/ipath_qp.c
--- a/drivers/infiniband/hw/ipath/ipath_qp.c Fri May 12 15:55:27 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_qp.c Fri May 12 15:55:28 2006 -0700
@@ -427,6 +427,7 @@ int ipath_modify_qp(struct ib_qp *ibqp,
int ipath_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
int attr_mask)
{
+ struct ipath_ibdev *dev = to_idev(ibqp->device);
struct ipath_qp *qp = to_iqp(ibqp);
enum ib_qp_state cur_state, new_state;
unsigned long flags;
@@ -443,6 +444,19 @@ int ipath_modify_qp(struct ib_qp *ibqp,
attr_mask))
goto inval;
+ if (attr_mask & IB_QP_AV)
+ if (attr->ah_attr.dlid == 0 ||
+ attr->ah_attr.dlid >= IPS_MULTICAST_LID_BASE)
+ goto inval;
+
+ if (attr_mask & IB_QP_PKEY_INDEX)
+ if (attr->pkey_index >= ipath_layer_get_npkeys(dev->dd))
+ goto inval;
+
+ if (attr_mask & IB_QP_MIN_RNR_TIMER)
+ if (attr->min_rnr_timer > 31)
+ goto inval;
+
switch (new_state) {
case IB_QPS_RESET:
ipath_reset_qp(qp);
@@ -457,13 +471,8 @@ int ipath_modify_qp(struct ib_qp *ibqp,
}
- if (attr_mask & IB_QP_PKEY_INDEX) {
- struct ipath_ibdev *dev = to_idev(ibqp->device);
-
- if (attr->pkey_index >= ipath_layer_get_npkeys(dev->dd))
- goto inval;
+ if (attr_mask & IB_QP_PKEY_INDEX)
qp->s_pkey_index = attr->pkey_index;
- }
if (attr_mask & IB_QP_DEST_QPN)
qp->remote_qpn = attr->dest_qp_num;
@@ -479,12 +488,8 @@ int ipath_modify_qp(struct ib_qp *ibqp,
if (attr_mask & IB_QP_ACCESS_FLAGS)
qp->qp_access_flags = attr->qp_access_flags;
- if (attr_mask & IB_QP_AV) {
- if (attr->ah_attr.dlid == 0 ||
- attr->ah_attr.dlid >= IPS_MULTICAST_LID_BASE)
- goto inval;
+ if (attr_mask & IB_QP_AV)
qp->remote_ah_attr = attr->ah_attr;
- }
if (attr_mask & IB_QP_PATH_MTU)
qp->path_mtu = attr->path_mtu;
@@ -499,11 +504,8 @@ int ipath_modify_qp(struct ib_qp *ibqp,
qp->s_rnr_retry_cnt = qp->s_rnr_retry;
}
- if (attr_mask & IB_QP_MIN_RNR_TIMER) {
- if (attr->min_rnr_timer > 31)
- goto inval;
+ if (attr_mask & IB_QP_MIN_RNR_TIMER)
qp->s_min_rnr_timer = attr->min_rnr_timer;
- }
if (attr_mask & IB_QP_QKEY)
qp->qkey = attr->qkey;
^ permalink raw reply
* [PATCH 51 of 53] ipath - fix reporting of vendor ID and a few other trivial bits
From: Bryan O'Sullivan @ 2006-05-12 23:43 UTC (permalink / raw)
To: rdreier; +Cc: openib-general, linux-kernel
In-Reply-To: <patchbomb.1147477365@eng-12.pathscale.com>
Signed-off-by: Bryan O'Sullivan <bos@pathscale.com>
diff -r bd1de2e983db -r 5f665c503f0d drivers/infiniband/hw/ipath/ipath_layer.c
--- a/drivers/infiniband/hw/ipath/ipath_layer.c Fri May 12 15:55:29 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_layer.c Fri May 12 16:41:45 2006 -0700
@@ -339,18 +339,26 @@ u32 ipath_layer_get_nguid(struct ipath_d
EXPORT_SYMBOL_GPL(ipath_layer_get_nguid);
-int ipath_layer_query_device(struct ipath_devdata *dd, u32 * vendor,
- u32 * boardrev, u32 * majrev, u32 * minrev)
-{
- *vendor = dd->ipath_vendorid;
- *boardrev = dd->ipath_boardrev;
- *majrev = dd->ipath_majrev;
- *minrev = dd->ipath_minrev;
-
- return 0;
-}
-
-EXPORT_SYMBOL_GPL(ipath_layer_query_device);
+u32 ipath_layer_get_majrev(struct ipath_devdata *dd)
+{
+ return dd->ipath_majrev;
+}
+
+EXPORT_SYMBOL_GPL(ipath_layer_get_majrev);
+
+u32 ipath_layer_get_minrev(struct ipath_devdata *dd)
+{
+ return dd->ipath_minrev;
+}
+
+EXPORT_SYMBOL_GPL(ipath_layer_get_minrev);
+
+u32 ipath_layer_get_pcirev(struct ipath_devdata *dd)
+{
+ return dd->ipath_pcirev;
+}
+
+EXPORT_SYMBOL_GPL(ipath_layer_get_pcirev);
u32 ipath_layer_get_flags(struct ipath_devdata *dd)
{
@@ -372,6 +380,13 @@ u16 ipath_layer_get_deviceid(struct ipat
}
EXPORT_SYMBOL_GPL(ipath_layer_get_deviceid);
+
+u32 ipath_layer_get_vendorid(struct ipath_devdata *dd)
+{
+ return dd->ipath_vendorid;
+}
+
+EXPORT_SYMBOL_GPL(ipath_layer_get_vendorid);
u64 ipath_layer_get_lastibcstat(struct ipath_devdata *dd)
{
diff -r bd1de2e983db -r 5f665c503f0d drivers/infiniband/hw/ipath/ipath_layer.h
--- a/drivers/infiniband/hw/ipath/ipath_layer.h Fri May 12 15:55:29 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_layer.h Fri May 12 16:41:45 2006 -0700
@@ -145,11 +145,13 @@ int ipath_layer_set_guid(struct ipath_de
int ipath_layer_set_guid(struct ipath_devdata *, __be64 guid);
__be64 ipath_layer_get_guid(struct ipath_devdata *);
u32 ipath_layer_get_nguid(struct ipath_devdata *);
-int ipath_layer_query_device(struct ipath_devdata *, u32 * vendor,
- u32 * boardrev, u32 * majrev, u32 * minrev);
+u32 ipath_layer_get_majrev(struct ipath_devdata *);
+u32 ipath_layer_get_minrev(struct ipath_devdata *);
+u32 ipath_layer_get_pcirev(struct ipath_devdata *);
u32 ipath_layer_get_flags(struct ipath_devdata *dd);
struct device *ipath_layer_get_device(struct ipath_devdata *dd);
u16 ipath_layer_get_deviceid(struct ipath_devdata *dd);
+u32 ipath_layer_get_vendorid(struct ipath_devdata *);
u64 ipath_layer_get_lastibcstat(struct ipath_devdata *dd);
u32 ipath_layer_get_ibmtu(struct ipath_devdata *dd);
int ipath_layer_enable_timer(struct ipath_devdata *dd);
diff -r bd1de2e983db -r 5f665c503f0d drivers/infiniband/hw/ipath/ipath_mad.c
--- a/drivers/infiniband/hw/ipath/ipath_mad.c Fri May 12 15:55:29 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_mad.c Fri May 12 16:41:45 2006 -0700
@@ -84,7 +84,7 @@ static int recv_subn_get_nodeinfo(struct
{
struct nodeinfo *nip = (struct nodeinfo *)&smp->data;
struct ipath_devdata *dd = to_idev(ibdev)->dd;
- u32 vendor, boardid, majrev, minrev;
+ u32 vendor, majrev, minrev;
if (smp->attr_mod)
smp->status |= IB_SMP_INVALID_FIELD;
@@ -104,9 +104,11 @@ static int recv_subn_get_nodeinfo(struct
nip->port_guid = nip->sys_guid;
nip->partition_cap = cpu_to_be16(ipath_layer_get_npkeys(dd));
nip->device_id = cpu_to_be16(ipath_layer_get_deviceid(dd));
- ipath_layer_query_device(dd, &vendor, &boardid, &majrev, &minrev);
+ majrev = ipath_layer_get_majrev(dd);
+ minrev = ipath_layer_get_minrev(dd);
nip->revision = cpu_to_be32((majrev << 16) | minrev);
nip->local_port_num = port;
+ vendor = ipath_layer_get_vendorid(dd);
nip->vendor_id[0] = 0;
nip->vendor_id[1] = vendor >> 8;
nip->vendor_id[2] = vendor;
diff -r bd1de2e983db -r 5f665c503f0d drivers/infiniband/hw/ipath/ipath_verbs.c
--- a/drivers/infiniband/hw/ipath/ipath_verbs.c Fri May 12 15:55:29 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.c Fri May 12 16:41:45 2006 -0700
@@ -620,18 +620,15 @@ static int ipath_query_device(struct ib_
struct ib_device_attr *props)
{
struct ipath_ibdev *dev = to_idev(ibdev);
- u32 vendor, boardrev, majrev, minrev;
memset(props, 0, sizeof(*props));
props->device_cap_flags = IB_DEVICE_BAD_PKEY_CNTR |
IB_DEVICE_BAD_QKEY_CNTR | IB_DEVICE_SHUTDOWN_PORT |
IB_DEVICE_SYS_IMAGE_GUID;
- ipath_layer_query_device(dev->dd, &vendor, &boardrev,
- &majrev, &minrev);
- props->vendor_id = vendor;
- props->vendor_part_id = boardrev;
- props->hw_ver = boardrev << 16 | majrev << 8 | minrev;
+ props->vendor_id = ipath_layer_get_vendorid(dev->dd);
+ props->vendor_part_id = ipath_layer_get_deviceid(dev->dd);
+ props->hw_ver = ipath_layer_get_pcirev(dev->dd);
props->sys_image_guid = dev->sys_image_guid;
@@ -1220,11 +1217,8 @@ static ssize_t show_rev(struct class_dev
{
struct ipath_ibdev *dev =
container_of(cdev, struct ipath_ibdev, ibdev.class_dev);
- int vendor, boardrev, majrev, minrev;
-
- ipath_layer_query_device(dev->dd, &vendor, &boardrev,
- &majrev, &minrev);
- return sprintf(buf, "%d.%d\n", majrev, minrev);
+
+ return sprintf(buf, "%x\n", ipath_layer_get_pcirev(dev->dd));
}
static ssize_t show_hca(struct class_device *cdev, char *buf)
@@ -1253,7 +1247,7 @@ static ssize_t show_stats(struct class_d
len = sprintf(buf,
"RC resends %d\n"
"RC no QACK %d\n"
- "RC ACKs %d\n"
+ "RC ACKs %d\n"
"RC SEQ NAKs %d\n"
"RC RDMA seq %d\n"
"RC RNR NAKs %d\n"
@@ -1263,7 +1257,7 @@ static ssize_t show_stats(struct class_d
"piobuf wait %d\n"
"no piobuf %d\n"
"PKT drops %d\n"
- "WQE errs %d\n",
+ "WQE errs %d\n",
dev->n_rc_resends, dev->n_rc_qacks, dev->n_rc_acks,
dev->n_seq_naks, dev->n_rdma_seq, dev->n_rnr_naks,
dev->n_other_naks, dev->n_timeouts,
^ permalink raw reply
* [PATCH 2 of 53] ipath - purge sps_lid and sps_mlid arrays, and /sys entries
From: Bryan O'Sullivan @ 2006-05-12 23:42 UTC (permalink / raw)
To: rdreier; +Cc: openib-general, linux-kernel
In-Reply-To: <patchbomb.1147477365@eng-12.pathscale.com>
The two arrays only had space for 4 units, so didn't work for larger
numbers of units. I thought I'd eliminated these before submitting the
original driver patches.
Also fixed error return on ipath_sysfs_unit_write to not set an error
code if the sysfs code reports consuming more chars than we wrote (since
that can include the nul, and the user doesn't have to include the nul
in the write).
Also changed from ipath_set_sps_lid() to ipath_set_lid(); the sps
was a leftover piece of naming.
Signed-off-by: Bryan O'Sullivan <bos@pathscale.com>
diff -r 9b9f24aab350 -r 3ab7a7b10bf2 drivers/infiniband/hw/ipath/ipath_common.h
--- a/drivers/infiniband/hw/ipath/ipath_common.h Fri May 12 15:55:27 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_common.h Fri May 12 15:55:27 2006 -0700
@@ -121,8 +121,7 @@ struct infinipath_stats {
__u64 sps_ports;
/* list of pkeys (other than default) accepted (0 means not set) */
__u16 sps_pkeys[4];
- /* lids for up to 4 infinipaths, indexed by infinipath # */
- __u16 sps_lid[4];
+ __u16 sps_unused16[4]; /* available; maintaining compatible layout */
/* number of user ports per chip (not IB ports) */
__u32 sps_nports;
/* not our interrupt, or already handled */
@@ -140,10 +139,8 @@ struct infinipath_stats {
* packets if ipath not configured, sma/mad, etc.)
*/
__u64 sps_krdrops;
- /* mlids for up to 4 infinipaths, indexed by infinipath # */
- __u16 sps_mlid[4];
/* pad for future growth */
- __u64 __sps_pad[45];
+ __u64 __sps_pad[46];
};
/*
diff -r 9b9f24aab350 -r 3ab7a7b10bf2 drivers/infiniband/hw/ipath/ipath_init_chip.c
--- a/drivers/infiniband/hw/ipath/ipath_init_chip.c Fri May 12 15:55:27 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_init_chip.c Fri May 12 15:55:27 2006 -0700
@@ -836,8 +836,6 @@ int ipath_init_chip(struct ipath_devdata
/* clear any interrups up to this point (ints still not enabled) */
ipath_write_kreg(dd, dd->ipath_kregs->kr_intclear, -1LL);
- ipath_stats.sps_lid[dd->ipath_unit] = dd->ipath_lid;
-
/*
* Set up the port 0 (kernel) rcvhdr q and egr TIDs. If doing
* re-init, the simplest way to handle this is to free
diff -r 9b9f24aab350 -r 3ab7a7b10bf2 drivers/infiniband/hw/ipath/ipath_layer.c
--- a/drivers/infiniband/hw/ipath/ipath_layer.c Fri May 12 15:55:27 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_layer.c Fri May 12 15:55:27 2006 -0700
@@ -299,9 +299,8 @@ bail:
EXPORT_SYMBOL_GPL(ipath_layer_set_mtu);
-int ipath_set_sps_lid(struct ipath_devdata *dd, u32 arg, u8 lmc)
-{
- ipath_stats.sps_lid[dd->ipath_unit] = arg;
+int ipath_set_lid(struct ipath_devdata *dd, u32 arg, u8 lmc)
+{
dd->ipath_lid = arg;
dd->ipath_lmc = lmc;
@@ -315,7 +314,7 @@ int ipath_set_sps_lid(struct ipath_devda
return 0;
}
-EXPORT_SYMBOL_GPL(ipath_set_sps_lid);
+EXPORT_SYMBOL_GPL(ipath_set_lid);
int ipath_layer_set_guid(struct ipath_devdata *dd, __be64 guid)
{
@@ -616,9 +615,9 @@ int ipath_layer_open(struct ipath_devdat
if (*dd->ipath_statusp & IPATH_STATUS_IB_READY)
intval |= IPATH_LAYER_INT_IF_UP;
- if (ipath_stats.sps_lid[dd->ipath_unit])
+ if (dd->ipath_lid)
intval |= IPATH_LAYER_INT_LID;
- if (ipath_stats.sps_mlid[dd->ipath_unit])
+ if (dd->ipath_mlid)
intval |= IPATH_LAYER_INT_BCAST;
/*
* do this on open, in case low level is already up and
diff -r 9b9f24aab350 -r 3ab7a7b10bf2 drivers/infiniband/hw/ipath/ipath_layer.h
--- a/drivers/infiniband/hw/ipath/ipath_layer.h Fri May 12 15:55:27 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_layer.h Fri May 12 15:55:27 2006 -0700
@@ -126,7 +126,7 @@ u32 ipath_layer_get_cr_errpkey(struct ip
u32 ipath_layer_get_cr_errpkey(struct ipath_devdata *dd);
int ipath_layer_set_linkstate(struct ipath_devdata *dd, u8 state);
int ipath_layer_set_mtu(struct ipath_devdata *, u16);
-int ipath_set_sps_lid(struct ipath_devdata *, u32, u8);
+int ipath_set_lid(struct ipath_devdata *, u32, u8);
int ipath_layer_send_hdr(struct ipath_devdata *dd,
struct ether_header *hdr);
int ipath_verbs_send(struct ipath_devdata *dd, u32 hdrwords,
diff -r 9b9f24aab350 -r 3ab7a7b10bf2 drivers/infiniband/hw/ipath/ipath_mad.c
--- a/drivers/infiniband/hw/ipath/ipath_mad.c Fri May 12 15:55:27 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_mad.c Fri May 12 15:55:27 2006 -0700
@@ -341,7 +341,7 @@ static int recv_subn_set_portinfo(struct
/* Must be a valid unicast LID address. */
if (lid == 0 || lid >= IPS_MULTICAST_LID_BASE)
goto err;
- ipath_set_sps_lid(dev->dd, lid, pip->mkeyprot_resv_lmc & 7);
+ ipath_set_lid(dev->dd, lid, pip->mkeyprot_resv_lmc & 7);
event.event = IB_EVENT_LID_CHANGE;
ib_dispatch_event(&event);
}
diff -r 9b9f24aab350 -r 3ab7a7b10bf2 drivers/infiniband/hw/ipath/ipath_sysfs.c
--- a/drivers/infiniband/hw/ipath/ipath_sysfs.c Fri May 12 15:55:27 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_sysfs.c Fri May 12 15:55:27 2006 -0700
@@ -84,98 +84,6 @@ static ssize_t show_num_units(struct dev
ipath_count_units(NULL, NULL, NULL));
}
-#define DRIVER_STAT(name, attr) \
- static ssize_t show_stat_##name(struct device_driver *dev, \
- char *buf) \
- { \
- return scnprintf( \
- buf, PAGE_SIZE, "%llu\n", \
- (unsigned long long) ipath_stats.sps_ ##attr); \
- } \
- static DRIVER_ATTR(name, S_IRUGO, show_stat_##name, NULL)
-
-DRIVER_STAT(intrs, ints);
-DRIVER_STAT(err_intrs, errints);
-DRIVER_STAT(errs, errs);
-DRIVER_STAT(pkt_errs, pkterrs);
-DRIVER_STAT(crc_errs, crcerrs);
-DRIVER_STAT(hw_errs, hwerrs);
-DRIVER_STAT(ib_link, iblink);
-DRIVER_STAT(port0_pkts, port0pkts);
-DRIVER_STAT(ether_spkts, ether_spkts);
-DRIVER_STAT(ether_rpkts, ether_rpkts);
-DRIVER_STAT(sma_spkts, sma_spkts);
-DRIVER_STAT(sma_rpkts, sma_rpkts);
-DRIVER_STAT(hdrq_full, hdrqfull);
-DRIVER_STAT(etid_full, etidfull);
-DRIVER_STAT(no_piobufs, nopiobufs);
-DRIVER_STAT(ports, ports);
-DRIVER_STAT(pkey0, pkeys[0]);
-DRIVER_STAT(pkey1, pkeys[1]);
-DRIVER_STAT(pkey2, pkeys[2]);
-DRIVER_STAT(pkey3, pkeys[3]);
-/* XXX fix the following when dynamic table of devices used */
-DRIVER_STAT(lid0, lid[0]);
-DRIVER_STAT(lid1, lid[1]);
-DRIVER_STAT(lid2, lid[2]);
-DRIVER_STAT(lid3, lid[3]);
-
-DRIVER_STAT(nports, nports);
-DRIVER_STAT(null_intr, nullintr);
-DRIVER_STAT(max_pkts_call, maxpkts_call);
-DRIVER_STAT(avg_pkts_call, avgpkts_call);
-DRIVER_STAT(page_locks, pagelocks);
-DRIVER_STAT(page_unlocks, pageunlocks);
-DRIVER_STAT(krdrops, krdrops);
-/* XXX fix the following when dynamic table of devices used */
-DRIVER_STAT(mlid0, mlid[0]);
-DRIVER_STAT(mlid1, mlid[1]);
-DRIVER_STAT(mlid2, mlid[2]);
-DRIVER_STAT(mlid3, mlid[3]);
-
-static struct attribute *driver_stat_attributes[] = {
- &driver_attr_intrs.attr,
- &driver_attr_err_intrs.attr,
- &driver_attr_errs.attr,
- &driver_attr_pkt_errs.attr,
- &driver_attr_crc_errs.attr,
- &driver_attr_hw_errs.attr,
- &driver_attr_ib_link.attr,
- &driver_attr_port0_pkts.attr,
- &driver_attr_ether_spkts.attr,
- &driver_attr_ether_rpkts.attr,
- &driver_attr_sma_spkts.attr,
- &driver_attr_sma_rpkts.attr,
- &driver_attr_hdrq_full.attr,
- &driver_attr_etid_full.attr,
- &driver_attr_no_piobufs.attr,
- &driver_attr_ports.attr,
- &driver_attr_pkey0.attr,
- &driver_attr_pkey1.attr,
- &driver_attr_pkey2.attr,
- &driver_attr_pkey3.attr,
- &driver_attr_lid0.attr,
- &driver_attr_lid1.attr,
- &driver_attr_lid2.attr,
- &driver_attr_lid3.attr,
- &driver_attr_nports.attr,
- &driver_attr_null_intr.attr,
- &driver_attr_max_pkts_call.attr,
- &driver_attr_avg_pkts_call.attr,
- &driver_attr_page_locks.attr,
- &driver_attr_page_unlocks.attr,
- &driver_attr_krdrops.attr,
- &driver_attr_mlid0.attr,
- &driver_attr_mlid1.attr,
- &driver_attr_mlid2.attr,
- &driver_attr_mlid3.attr,
- NULL
-};
-
-static struct attribute_group driver_stat_attr_group = {
- .name = "stats",
- .attrs = driver_stat_attributes
-};
static ssize_t show_status(struct device *dev,
struct device_attribute *attr,
@@ -272,7 +180,7 @@ static ssize_t store_lid(struct device *
size_t count)
{
struct ipath_devdata *dd = dev_get_drvdata(dev);
- u16 lid;
+ u16 lid = 0; /* gcc thinks might be un-initialized */
int ret;
ret = ipath_parse_ushort(buf, &lid);
@@ -284,11 +192,11 @@ static ssize_t store_lid(struct device *
goto invalid;
}
- ipath_set_sps_lid(dd, lid, 0);
+ ipath_set_lid(dd, lid, 0);
goto bail;
invalid:
- ipath_dev_err(dd, "attempt to set invalid LID\n");
+ ipath_dev_err(dd, "attempt to set invalid LID 0x%x\n", lid);
bail:
return ret;
}
@@ -319,7 +227,6 @@ static ssize_t store_mlid(struct device
unit = dd->ipath_unit;
dd->ipath_mlid = mlid;
- ipath_stats.sps_mlid[unit] = mlid;
ipath_layer_intr(dd, IPATH_LAYER_INT_BCAST);
goto bail;
@@ -737,17 +644,12 @@ int ipath_driver_create_group(struct dev
if (ret)
goto bail;
- ret = sysfs_create_group(&drv->kobj, &driver_stat_attr_group);
- if (ret)
- sysfs_remove_group(&drv->kobj, &driver_attr_group);
-
bail:
return ret;
}
void ipath_driver_remove_group(struct device_driver *drv)
{
- sysfs_remove_group(&drv->kobj, &driver_stat_attr_group);
sysfs_remove_group(&drv->kobj, &driver_attr_group);
}
^ permalink raw reply
* [PATCH 52 of 53] ipath - register as IB device owner
From: Bryan O'Sullivan @ 2006-05-12 23:43 UTC (permalink / raw)
To: rdreier; +Cc: openib-general, linux-kernel
In-Reply-To: <patchbomb.1147477365@eng-12.pathscale.com>
Signed-off-by: Bryan O'Sullivan <bos@pathscale.com>
diff -r 5f665c503f0d -r fd9bdeea5b10 drivers/infiniband/hw/ipath/ipath_verbs.c
--- a/drivers/infiniband/hw/ipath/ipath_verbs.c Fri May 12 16:41:45 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.c Fri May 12 16:42:39 2006 -0700
@@ -1060,6 +1060,7 @@ static void *ipath_register_ib_device(in
idev->dd = dd;
strlcpy(dev->name, "ipath%d", IB_DEVICE_NAME_MAX);
+ dev->owner = THIS_MODULE;
dev->node_guid = ipath_layer_get_guid(dd);
dev->uverbs_abi_ver = IPATH_UVERBS_ABI_VERSION;
dev->uverbs_cmd_mask =
^ permalink raw reply
* [PATCH 5 of 53] ipath - forbid creation of AHs with illegal ports
From: Bryan O'Sullivan @ 2006-05-12 23:42 UTC (permalink / raw)
To: rdreier; +Cc: openib-general, linux-kernel
In-Reply-To: <patchbomb.1147477365@eng-12.pathscale.com>
Don't allow an AH to be created with an illegal port.
Signed-off-by: Bryan O'Sullivan <bos@pathscale.com>
diff -r 300f0aa6f034 -r db56c0ab6a64 drivers/infiniband/hw/ipath/ipath_verbs.c
--- a/drivers/infiniband/hw/ipath/ipath_verbs.c Fri May 12 15:55:27 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.c Fri May 12 15:55:27 2006 -0700
@@ -810,6 +810,12 @@ static struct ib_ah *ipath_create_ah(str
if (ah_attr->dlid >= IPS_MULTICAST_LID_BASE &&
ah_attr->dlid != IPS_PERMISSIVE_LID &&
!(ah_attr->ah_flags & IB_AH_GRH)) {
+ ret = ERR_PTR(-EINVAL);
+ goto bail;
+ }
+
+ if (ah_attr->port_num != 1 ||
+ ah_attr->port_num > pd->device->phys_port_cnt) {
ret = ERR_PTR(-EINVAL);
goto bail;
}
^ permalink raw reply
* [PATCH 50 of 53] ipath - reduce maximum table sizes
From: Bryan O'Sullivan @ 2006-05-12 23:43 UTC (permalink / raw)
To: rdreier; +Cc: openib-general, linux-kernel
In-Reply-To: <patchbomb.1147477365@eng-12.pathscale.com>
Decrease the number of WRs and SGEs we support from 131071/255 to
16383/60. This decreases our maximum memory usage per QP from ~1800MB
down to about 40MB. This is still a lot, but it's better than 2GB.
Signed-off-by: Bryan O'Sullivan <bos@pathscale.com>
diff -r 40532fdc53f0 -r bd1de2e983db drivers/infiniband/hw/ipath/ipath_verbs.c
--- a/drivers/infiniband/hw/ipath/ipath_verbs.c Fri May 12 15:55:29 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.c Fri May 12 15:55:29 2006 -0700
@@ -73,12 +73,12 @@ module_param_named(max_cqs, ib_ipath_max
module_param_named(max_cqs, ib_ipath_max_cqs, uint, S_IWUSR | S_IRUGO);
MODULE_PARM_DESC(max_cqs, "Maximum number of completion queues to support");
-unsigned int ib_ipath_max_qp_wrs = 0x1FFFF;
+unsigned int ib_ipath_max_qp_wrs = 0x3FFF;
module_param_named(max_qp_wrs, ib_ipath_max_qp_wrs, uint,
S_IWUSR | S_IRUGO);
MODULE_PARM_DESC(max_qp_wrs, "Maximum number of QP WRs to support");
-unsigned int ib_ipath_max_sges = 0xFF;
+unsigned int ib_ipath_max_sges = 0x60;
module_param_named(max_sges, ib_ipath_max_sges, uint, S_IWUSR | S_IRUGO);
MODULE_PARM_DESC(max_sges, "Maximum number of SGEs to support");
^ permalink raw reply
* [PATCH 48 of 53] ipath - QP should ignore receive queue size if SRQ specified
From: Bryan O'Sullivan @ 2006-05-12 23:43 UTC (permalink / raw)
To: rdreier; +Cc: openib-general, linux-kernel
In-Reply-To: <patchbomb.1147477365@eng-12.pathscale.com>
Signed-off-by: Bryan O'Sullivan <bos@pathscale.com>
diff -r a1615956e57f -r 49b446b12f16 drivers/infiniband/hw/ipath/ipath_qp.c
--- a/drivers/infiniband/hw/ipath/ipath_qp.c Fri May 12 15:55:29 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_qp.c Fri May 12 15:55:29 2006 -0700
@@ -684,16 +684,22 @@ struct ib_qp *ipath_create_qp(struct ib_
ret = ERR_PTR(-ENOMEM);
goto bail;
}
- qp->r_rq.size = init_attr->cap.max_recv_wr + 1;
- sz = sizeof(struct ipath_sge) *
- init_attr->cap.max_recv_sge +
- sizeof(struct ipath_rwqe);
- qp->r_rq.wq = vmalloc(qp->r_rq.size * sz);
- if (!qp->r_rq.wq) {
- kfree(qp);
- vfree(swq);
- ret = ERR_PTR(-ENOMEM);
- goto bail;
+ if (init_attr->srq) {
+ qp->r_rq.size = 0;
+ qp->r_rq.max_sge = 0;
+ qp->r_rq.wq = NULL;
+ } else {
+ qp->r_rq.size = init_attr->cap.max_recv_wr + 1;
+ qp->r_rq.max_sge = init_attr->cap.max_recv_sge;
+ sz = (sizeof(struct ipath_sge) * qp->r_rq.max_sge) +
+ sizeof(struct ipath_rwqe);
+ qp->r_rq.wq = vmalloc(qp->r_rq.size * sz);
+ if (!qp->r_rq.wq) {
+ kfree(qp);
+ vfree(swq);
+ ret = ERR_PTR(-ENOMEM);
+ goto bail;
+ }
}
/*
@@ -712,7 +718,6 @@ struct ib_qp *ipath_create_qp(struct ib_
qp->s_wq = swq;
qp->s_size = init_attr->cap.max_send_wr + 1;
qp->s_max_sge = init_attr->cap.max_send_sge;
- qp->r_rq.max_sge = init_attr->cap.max_recv_sge;
qp->s_flags = init_attr->sq_sig_type == IB_SIGNAL_REQ_WR ?
1 << IPATH_S_SIGNAL_REQ_WR : 0;
dev = to_idev(ibpd->device);
^ permalink raw reply
* [PATCH 44 of 53] ipath - allow diags on any unit
From: Bryan O'Sullivan @ 2006-05-12 23:43 UTC (permalink / raw)
To: rdreier; +Cc: openib-general, linux-kernel
In-Reply-To: <patchbomb.1147477365@eng-12.pathscale.com>
Previously, we hardwired all diags to unit 0.
Signed-off-by: Bryan O'Sullivan <bos@pathscale.com>
diff -r 7634b2f0fc40 -r 28d938eb0463 drivers/infiniband/hw/ipath/ipath_diag.c
--- a/drivers/infiniband/hw/ipath/ipath_diag.c Fri May 12 15:55:29 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_diag.c Fri May 12 15:55:29 2006 -0700
@@ -66,18 +66,20 @@ static struct file_operations diag_file_
.release = ipath_diag_release
};
-static struct cdev *diag_cdev;
-static struct class_device *diag_class_dev;
-
-int ipath_diag_init(void)
-{
- return ipath_cdev_init(IPATH_DIAG_MINOR, "ipath_diag",
- &diag_file_ops, &diag_cdev, &diag_class_dev);
-}
-
-void ipath_diag_cleanup(void)
-{
- ipath_cdev_cleanup(&diag_cdev, &diag_class_dev);
+int ipath_diag_add(struct ipath_devdata *dd)
+{
+ char name[16];
+
+ snprintf(name, sizeof(name), "ipath_diag%d", dd->ipath_unit);
+
+ return ipath_cdev_init(IPATH_DIAG_MINOR_BASE + dd->ipath_unit, name,
+ &diag_file_ops, &dd->diag_cdev,
+ &dd->diag_class_dev);
+}
+
+void ipath_diag_remove(struct ipath_devdata *dd)
+{
+ ipath_cdev_cleanup(&dd->diag_cdev, &dd->diag_class_dev);
}
/**
@@ -101,8 +103,7 @@ static int ipath_read_umem64(struct ipat
int ret;
/* not very efficient, but it works for now */
- if (reg_addr < dd->ipath_kregbase ||
- reg_end > dd->ipath_kregend) {
+ if (reg_addr < dd->ipath_kregbase || reg_end > dd->ipath_kregend) {
ret = -EINVAL;
goto bail;
}
@@ -139,8 +140,7 @@ static int ipath_write_umem64(struct ipa
int ret;
/* not very efficient, but it works for now */
- if (reg_addr < dd->ipath_kregbase ||
- reg_end > dd->ipath_kregend) {
+ if (reg_addr < dd->ipath_kregbase || reg_end > dd->ipath_kregend) {
ret = -EINVAL;
goto bail;
}
@@ -240,59 +240,45 @@ bail:
static int ipath_diag_open(struct inode *in, struct file *fp)
{
+ int unit = iminor(in) - IPATH_DIAG_MINOR_BASE;
struct ipath_devdata *dd;
- int unit = 0; /* XXX this is bogus */
- unsigned long flags;
- int ret;
-
- dd = ipath_lookup(unit);
+ int ret;
mutex_lock(&ipath_mutex);
- spin_lock_irqsave(&ipath_devs_lock, flags);
if (ipath_diag_inuse) {
ret = -EBUSY;
goto bail;
}
- list_for_each_entry(dd, &ipath_dev_list, ipath_list) {
- /*
- * we need at least one infinipath device to be present
- * (don't use INITTED, because we want to be able to open
- * even if device is in freeze mode, which cleared INITTED).
- * There is a small amount of risk to this, which is why we
- * also verify kregbase is set.
- */
-
- if (!(dd->ipath_flags & IPATH_PRESENT) ||
- !dd->ipath_kregbase)
- continue;
-
- ipath_diag_inuse = 1;
- diag_set_link = 0;
- ret = 0;
- goto bail;
- }
-
- ret = -ENODEV;
-
-bail:
- spin_unlock_irqrestore(&ipath_devs_lock, flags);
+ dd = ipath_lookup(unit);
+
+ if (dd == NULL || !(dd->ipath_flags & IPATH_PRESENT) ||
+ !dd->ipath_kregbase) {
+ ret = -ENODEV;
+ goto bail;
+ }
+
+ fp->private_data = dd;
+ ipath_diag_inuse = 1;
+ diag_set_link = 0;
+ ret = 0;
/* Only expose a way to reset the device if we
make it into diag mode. */
- if (ret == 0)
- ipath_expose_reset(&dd->pcidev->dev);
-
+ ipath_expose_reset(&dd->pcidev->dev);
+
+bail:
mutex_unlock(&ipath_mutex);
return ret;
}
-static int ipath_diag_release(struct inode *i, struct file *f)
+static int ipath_diag_release(struct inode *in, struct file *fp)
{
mutex_lock(&ipath_mutex);
ipath_diag_inuse = 0;
+ fp->private_data = NULL;
mutex_unlock(&ipath_mutex);
return 0;
}
@@ -300,16 +286,9 @@ static ssize_t ipath_diag_read(struct fi
static ssize_t ipath_diag_read(struct file *fp, char __user *data,
size_t count, loff_t *off)
{
- int unit = 0; /* XXX provide for reads on other units some day */
- struct ipath_devdata *dd;
+ struct ipath_devdata *dd = fp->private_data;
void __iomem *kreg_base;
ssize_t ret;
-
- dd = ipath_lookup(unit);
- if (!dd) {
- ret = -ENODEV;
- goto bail;
- }
kreg_base = dd->ipath_kregbase;
@@ -329,23 +308,16 @@ static ssize_t ipath_diag_read(struct fi
ret = count;
}
-bail:
return ret;
}
static ssize_t ipath_diag_write(struct file *fp, const char __user *data,
size_t count, loff_t *off)
{
- int unit = 0; /* XXX this is bogus */
- struct ipath_devdata *dd;
+ struct ipath_devdata *dd = fp->private_data;
void __iomem *kreg_base;
ssize_t ret;
- dd = ipath_lookup(unit);
- if (!dd) {
- ret = -ENODEV;
- goto bail;
- }
kreg_base = dd->ipath_kregbase;
if (count == 0)
@@ -364,6 +336,6 @@ static ssize_t ipath_diag_write(struct f
ret = count;
}
-bail:
- return ret;
-}
+ return ret;
+}
+
diff -r 7634b2f0fc40 -r 28d938eb0463 drivers/infiniband/hw/ipath/ipath_driver.c
--- a/drivers/infiniband/hw/ipath/ipath_driver.c Fri May 12 15:55:29 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_driver.c Fri May 12 15:55:29 2006 -0700
@@ -488,6 +488,7 @@ static int __devinit ipath_init_one(stru
ipath_device_create_group(&pdev->dev, dd);
ipathfs_add_device(dd);
ipath_user_add(dd);
+ ipath_diag_add(dd);
ipath_layer_add(dd);
goto bail;
@@ -517,8 +518,9 @@ static void __devexit ipath_remove_one(s
return;
dd = pci_get_drvdata(pdev);
- ipath_layer_del(dd);
- ipath_user_del(dd);
+ ipath_layer_remove(dd);
+ ipath_diag_remove(dd);
+ ipath_user_remove(dd);
ipathfs_remove_device(dd);
ipath_device_remove_group(&pdev->dev, dd);
ipath_cdbg(VERBOSE, "Releasing pci memory regions, dd %p, "
diff -r 7634b2f0fc40 -r 28d938eb0463 drivers/infiniband/hw/ipath/ipath_file_ops.c
--- a/drivers/infiniband/hw/ipath/ipath_file_ops.c Fri May 12 15:55:29 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_file_ops.c Fri May 12 15:55:29 2006 -0700
@@ -1390,16 +1390,16 @@ done:
static int ipath_open(struct inode *in, struct file *fp)
{
- int ret, minor;
+ int ret, user_minor;
mutex_lock(&ipath_mutex);
- minor = iminor(in);
+ user_minor = iminor(in) - IPATH_USER_MINOR_BASE;
ipath_cdbg(VERBOSE, "open on dev %lx (minor %d)\n",
- (long)in->i_rdev, minor);
-
- if (minor)
- ret = find_free_port(minor - 1, fp);
+ (long)in->i_rdev, user_minor);
+
+ if (user_minor)
+ ret = find_free_port(user_minor - 1, fp);
else
ret = find_best_unit(fp);
@@ -1799,19 +1799,13 @@ int ipath_user_add(struct ipath_devdata
"error %d\n", -ret);
goto bail;
}
- ret = ipath_diag_init();
- if (ret < 0) {
- ipath_dev_err(dd, "Unable to set up diag support: "
- "error %d\n", -ret);
- goto bail_sma;
- }
ret = init_cdev(0, "ipath", &ipath_file_ops, &wildcard_cdev,
&wildcard_class_dev);
if (ret < 0) {
ipath_dev_err(dd, "Could not create wildcard "
"minor: error %d\n", -ret);
- goto bail_diag;
+ goto bail_sma;
}
atomic_set(&user_setup, 1);
@@ -1820,31 +1814,28 @@ int ipath_user_add(struct ipath_devdata
snprintf(name, sizeof(name), "ipath%d", dd->ipath_unit);
ret = init_cdev(dd->ipath_unit + 1, name, &ipath_file_ops,
- &dd->cdev, &dd->class_dev);
+ &dd->user_cdev, &dd->user_class_dev);
if (ret < 0)
ipath_dev_err(dd, "Could not create user minor %d, %s\n",
dd->ipath_unit + 1, name);
goto bail;
-bail_diag:
- ipath_diag_cleanup();
bail_sma:
user_cleanup();
bail:
return ret;
}
-void ipath_user_del(struct ipath_devdata *dd)
-{
- cleanup_cdev(&dd->cdev, &dd->class_dev);
+void ipath_user_remove(struct ipath_devdata *dd)
+{
+ cleanup_cdev(&dd->user_cdev, &dd->user_class_dev);
if (atomic_dec_return(&user_count) == 0) {
if (atomic_read(&user_setup) == 0)
goto bail;
cleanup_cdev(&wildcard_cdev, &wildcard_class_dev);
- ipath_diag_cleanup();
user_cleanup();
atomic_set(&user_setup, 0);
diff -r 7634b2f0fc40 -r 28d938eb0463 drivers/infiniband/hw/ipath/ipath_kernel.h
--- a/drivers/infiniband/hw/ipath/ipath_kernel.h Fri May 12 15:55:29 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_kernel.h Fri May 12 15:55:29 2006 -0700
@@ -347,8 +347,10 @@ struct ipath_devdata {
char *ipath_freezemsg;
/* pci access data structure */
struct pci_dev *pcidev;
- struct cdev *cdev;
- struct class_device *class_dev;
+ struct cdev *user_cdev;
+ struct cdev *diag_cdev;
+ struct class_device *user_class_dev;
+ struct class_device *diag_class_dev;
/* timer used to prevent stats overflow, error throttling, etc. */
struct timer_list ipath_stats_timer;
/* check for stale messages in rcv queue */
@@ -531,7 +533,7 @@ extern int __ipath_verbs_rcv(struct ipat
extern int __ipath_verbs_rcv(struct ipath_devdata *, void *, void *, u32);
void ipath_layer_add(struct ipath_devdata *);
-void ipath_layer_del(struct ipath_devdata *);
+void ipath_layer_remove(struct ipath_devdata *);
int ipath_init_chip(struct ipath_devdata *, int);
int ipath_enable_wc(struct ipath_devdata *dd);
@@ -545,14 +547,14 @@ void ipath_cdev_cleanup(struct cdev **cd
void ipath_cdev_cleanup(struct cdev **cdevp,
struct class_device **class_devp);
-int ipath_diag_init(void);
-void ipath_diag_cleanup(void);
+int ipath_diag_add(struct ipath_devdata *);
+void ipath_diag_remove(struct ipath_devdata *);
void ipath_diag_bringup_link(struct ipath_devdata *);
extern wait_queue_head_t ipath_sma_state_wait;
int ipath_user_add(struct ipath_devdata *dd);
-void ipath_user_del(struct ipath_devdata *dd);
+void ipath_user_remove(struct ipath_devdata *dd);
struct sk_buff *ipath_alloc_skb(struct ipath_devdata *dd, gfp_t);
@@ -831,9 +833,10 @@ extern struct mutex ipath_mutex;
#define IPATH_DRV_NAME "ipath_core"
#define IPATH_MAJOR 233
+#define IPATH_USER_MINOR_BASE 0
#define IPATH_SMA_MINOR 128
-#define IPATH_DIAG_MINOR 129
-#define IPATH_NMINORS 130
+#define IPATH_DIAG_MINOR_BASE 129
+#define IPATH_NMINORS 255
#define ipath_dev_err(dd,fmt,...) \
do { \
diff -r 7634b2f0fc40 -r 28d938eb0463 drivers/infiniband/hw/ipath/ipath_layer.c
--- a/drivers/infiniband/hw/ipath/ipath_layer.c Fri May 12 15:55:29 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_layer.c Fri May 12 15:55:29 2006 -0700
@@ -402,7 +402,7 @@ void ipath_layer_add(struct ipath_devdat
mutex_unlock(&ipath_layer_mutex);
}
-void ipath_layer_del(struct ipath_devdata *dd)
+void ipath_layer_remove(struct ipath_devdata *dd)
{
mutex_lock(&ipath_layer_mutex);
^ permalink raw reply
* [PATCH 53 of 53] ipath - add memory barrier when waiting for writes
From: Bryan O'Sullivan @ 2006-05-12 23:43 UTC (permalink / raw)
To: rdreier; +Cc: openib-general, linux-kernel
In-Reply-To: <patchbomb.1147477365@eng-12.pathscale.com>
Signed-off-by: Bryan O'Sullivan <bos@pathscale.com>
diff -r fd9bdeea5b10 -r f8ebb8c1e436 drivers/infiniband/hw/ipath/ipath_eeprom.c
--- a/drivers/infiniband/hw/ipath/ipath_eeprom.c Fri May 12 16:42:39 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_eeprom.c Fri May 12 16:42:39 2006 -0700
@@ -185,6 +185,7 @@ bail:
*/
static void i2c_wait_for_writes(struct ipath_devdata *dd)
{
+ mb();
(void)ipath_read_kreg32(dd, dd->ipath_kregs->kr_scratch);
}
^ permalink raw reply
* [PATCH 33 of 53] ipath - clean up some comments
From: Bryan O'Sullivan @ 2006-05-12 23:43 UTC (permalink / raw)
To: rdreier; +Cc: openib-general, linux-kernel
In-Reply-To: <patchbomb.1147477365@eng-12.pathscale.com>
Signed-off-by: Bryan O'Sullivan <bos@pathscale.com>
diff -r b9fd1a46c910 -r 5ddaf7c07cdf drivers/infiniband/hw/ipath/ipath_kernel.h
--- a/drivers/infiniband/hw/ipath/ipath_kernel.h Fri May 12 15:55:28 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_kernel.h Fri May 12 15:55:28 2006 -0700
@@ -720,13 +720,8 @@ u64 ipath_read_kreg64_port(const struct
* @port: port number
*
* Return the contents of a register that is virtualized to be per port.
- * Prints a debug message and returns -1 on errors (not distinguishable from
- * valid contents at runtime; we may add a separate error variable at some
- * point).
- *
- * This is normally not used by the kernel, but may be for debugging, and
- * has a different implementation than user mode, which is why it's not in
- * _common.h.
+ * Returns -1 on errors (not distinguishable from valid contents at
+ * runtime; we may add a separate error variable at some point).
*/
static inline u32 ipath_read_ureg32(const struct ipath_devdata *dd,
ipath_ureg regno, int port)
^ permalink raw reply
* [PATCH 39 of 53] ipath - count PE800 receive interrupts on user ports
From: Bryan O'Sullivan @ 2006-05-12 23:43 UTC (permalink / raw)
To: rdreier; +Cc: openib-general, linux-kernel
In-Reply-To: <patchbomb.1147477365@eng-12.pathscale.com>
Fixed so it works on the PE-800. It had not previously been updated to
match PE-800 receive interrupt differences from HT-400.
Signed-off-by: Bryan O'Sullivan <bos@pathscale.com>
diff -r e9306861dc6a -r 5b565c24d62a drivers/infiniband/hw/ipath/ipath_file_ops.c
--- a/drivers/infiniband/hw/ipath/ipath_file_ops.c Fri May 12 15:55:28 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_file_ops.c Fri May 12 15:55:28 2006 -0700
@@ -1172,6 +1172,10 @@ static unsigned int ipath_poll(struct fi
if (tail == head) {
set_bit(IPATH_PORT_WAITING_RCV, &pd->port_flag);
+ if(dd->ipath_rhdrhead_intr_off) /* arm rcv interrupt */
+ (void)ipath_write_ureg(dd, ur_rcvhdrhead,
+ dd->ipath_rhdrhead_intr_off
+ | head, pd->port_port);
poll_wait(fp, &pd->port_wait, pt);
if (test_bit(IPATH_PORT_WAITING_RCV, &pd->port_flag)) {
^ permalink raw reply
* [PATCH 45 of 53] ipath - fix memory leak when create of QP fails
From: Bryan O'Sullivan @ 2006-05-12 23:43 UTC (permalink / raw)
To: rdreier; +Cc: openib-general, linux-kernel
In-Reply-To: <patchbomb.1147477365@eng-12.pathscale.com>
Signed-off-by: Bryan O'Sullivan <bos@pathscale.com>
diff -r 28d938eb0463 -r b41e576e5202 drivers/infiniband/hw/ipath/ipath_init_chip.c
--- a/drivers/infiniband/hw/ipath/ipath_init_chip.c Fri May 12 15:55:29 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_init_chip.c Fri May 12 15:55:29 2006 -0700
@@ -114,6 +114,7 @@ static int create_port0_egr(struct ipath
"eager TID %u\n", e);
while (e != 0)
dev_kfree_skb(skbs[--e]);
+ vfree(skbs);
ret = -ENOMEM;
goto bail;
}
^ permalink raw reply
* [PATCH 27 of 53] ipath - fix accounting of data packets with bad VLs
From: Bryan O'Sullivan @ 2006-05-12 23:43 UTC (permalink / raw)
To: rdreier; +Cc: openib-general, linux-kernel
In-Reply-To: <patchbomb.1147477365@eng-12.pathscale.com>
For better IB conformance.
Signed-off-by: Bryan O'Sullivan <bos@pathscale.com>
diff -r 8e2d63833cf2 -r 551966b88d7c drivers/infiniband/hw/ipath/ipath_layer.c
--- a/drivers/infiniband/hw/ipath/ipath_layer.c Fri May 12 15:55:28 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_layer.c Fri May 12 15:55:28 2006 -0700
@@ -1019,13 +1019,11 @@ int ipath_layer_get_counters(struct ipat
ipath_snap_cntr(dd, dd->ipath_cregs->cr_rxdroppktcnt) +
ipath_snap_cntr(dd, dd->ipath_cregs->cr_rcvovflcnt) +
ipath_snap_cntr(dd, dd->ipath_cregs->cr_portovflcnt) +
- ipath_snap_cntr(dd, dd->ipath_cregs->cr_errrcvflowctrlcnt) +
ipath_snap_cntr(dd, dd->ipath_cregs->cr_err_rlencnt) +
ipath_snap_cntr(dd, dd->ipath_cregs->cr_invalidrlencnt) +
ipath_snap_cntr(dd, dd->ipath_cregs->cr_erricrccnt) +
ipath_snap_cntr(dd, dd->ipath_cregs->cr_errvcrccnt) +
ipath_snap_cntr(dd, dd->ipath_cregs->cr_errlpcrccnt) +
- ipath_snap_cntr(dd, dd->ipath_cregs->cr_errlinkcnt) +
ipath_snap_cntr(dd, dd->ipath_cregs->cr_badformatcnt);
cntrs->port_rcv_remphys_errors =
ipath_snap_cntr(dd, dd->ipath_cregs->cr_rcvebpcnt);
diff -r 8e2d63833cf2 -r 551966b88d7c drivers/infiniband/hw/ipath/ipath_mad.c
--- a/drivers/infiniband/hw/ipath/ipath_mad.c Fri May 12 15:55:28 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_mad.c Fri May 12 15:55:28 2006 -0700
@@ -1316,32 +1316,8 @@ int ipath_process_mad(struct ib_device *
struct ib_wc *in_wc, struct ib_grh *in_grh,
struct ib_mad *in_mad, struct ib_mad *out_mad)
{
- struct ipath_ibdev *dev = to_idev(ibdev);
int ret;
- /*
- * Snapshot current HW counters to "clear" them.
- * This should be done when the driver is loaded except that for
- * some reason we get a zillion errors when brining up the link.
- */
- if (dev->rcv_errors == 0) {
- struct ipath_layer_counters cntrs;
-
- ipath_layer_get_counters(to_idev(ibdev)->dd, &cntrs);
- dev->rcv_errors++;
- dev->n_symbol_error_counter = cntrs.symbol_error_counter;
- dev->n_link_error_recovery_counter =
- cntrs.link_error_recovery_counter;
- dev->n_link_downed_counter = cntrs.link_downed_counter;
- dev->n_port_rcv_errors = cntrs.port_rcv_errors + 1;
- dev->n_port_rcv_remphys_errors =
- cntrs.port_rcv_remphys_errors;
- dev->n_port_xmit_discards = cntrs.port_xmit_discards;
- dev->n_port_xmit_data = cntrs.port_xmit_data;
- dev->n_port_rcv_data = cntrs.port_rcv_data;
- dev->n_port_xmit_packets = cntrs.port_xmit_packets;
- dev->n_port_rcv_packets = cntrs.port_rcv_packets;
- }
switch (in_mad->mad_hdr.mgmt_class) {
case IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE:
case IB_MGMT_CLASS_SUBN_LID_ROUTED:
diff -r 8e2d63833cf2 -r 551966b88d7c drivers/infiniband/hw/ipath/ipath_verbs.c
--- a/drivers/infiniband/hw/ipath/ipath_verbs.c Fri May 12 15:55:28 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.c Fri May 12 15:55:28 2006 -0700
@@ -981,6 +981,7 @@ static int ipath_verbs_register_sysfs(st
*/
static void *ipath_register_ib_device(int unit, struct ipath_devdata *dd)
{
+ struct ipath_layer_counters cntrs;
struct ipath_ibdev *idev;
struct ib_device *dev;
int ret;
@@ -1030,6 +1031,21 @@ static void *ipath_register_ib_device(in
idev->pma_counter_select[3] = IB_PMA_PORT_RCV_PKTS;
idev->pma_counter_select[5] = IB_PMA_PORT_XMIT_WAIT;
idev->link_width_enabled = 3; /* 1x or 4x */
+
+ /* Snapshot current HW counters to "clear" them. */
+ ipath_layer_get_counters(dd, &cntrs);
+ idev->n_symbol_error_counter = cntrs.symbol_error_counter;
+ idev->n_link_error_recovery_counter =
+ cntrs.link_error_recovery_counter;
+ idev->n_link_downed_counter = cntrs.link_downed_counter;
+ idev->n_port_rcv_errors = cntrs.port_rcv_errors;
+ idev->n_port_rcv_remphys_errors =
+ cntrs.port_rcv_remphys_errors;
+ idev->n_port_xmit_discards = cntrs.port_xmit_discards;
+ idev->n_port_xmit_data = cntrs.port_xmit_data;
+ idev->n_port_rcv_data = cntrs.port_rcv_data;
+ idev->n_port_xmit_packets = cntrs.port_xmit_packets;
+ idev->n_port_rcv_packets = cntrs.port_rcv_packets;
/*
* The system image GUID is supposed to be the same for all
^ permalink raw reply
* [PATCH 32 of 53] ipath - fix NULL dereference during cleanup
From: Bryan O'Sullivan @ 2006-05-12 23:43 UTC (permalink / raw)
To: rdreier; +Cc: openib-general, linux-kernel
In-Reply-To: <patchbomb.1147477365@eng-12.pathscale.com>
Fix NULL deref due to pcidev being clobbered before dd->ipath_f_cleanup()
was called.
Signed-off-by: Bryan O'Sullivan <bos@pathscale.com>
diff -r 4868daa7f215 -r b9fd1a46c910 drivers/infiniband/hw/ipath/ipath_driver.c
--- a/drivers/infiniband/hw/ipath/ipath_driver.c Fri May 12 15:55:28 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_driver.c Fri May 12 15:55:28 2006 -0700
@@ -1897,19 +1897,19 @@ static void __exit infinipath_cleanup(vo
} else
ipath_dbg("irq is 0, not doing free_irq "
"for unit %u\n", dd->ipath_unit);
+
+ /*
+ * we check for NULL here, because it's outside
+ * the kregbase check, and we need to call it
+ * after the free_irq. Thus it's possible that
+ * the function pointers were never initialized.
+ */
+ if (dd->ipath_f_cleanup)
+ /* clean up chip-specific stuff */
+ dd->ipath_f_cleanup(dd);
+
dd->pcidev = NULL;
}
-
- /*
- * we check for NULL here, because it's outside the kregbase
- * check, and we need to call it after the free_irq. Thus
- * it's possible that the function pointers were never
- * initialized.
- */
- if (dd->ipath_f_cleanup)
- /* clean up chip-specific stuff */
- dd->ipath_f_cleanup(dd);
-
spin_lock_irqsave(&ipath_devs_lock, flags);
}
^ permalink raw reply
* [PATCH 18 of 53] ipath - make max mcast sizes configurable
From: Bryan O'Sullivan @ 2006-05-12 23:43 UTC (permalink / raw)
To: rdreier; +Cc: openib-general, linux-kernel
In-Reply-To: <patchbomb.1147477365@eng-12.pathscale.com>
Make the max IB mcast sizes configurable.
Signed-off-by: Bryan O'Sullivan <bos@pathscale.com>
diff -r c5f3731224bb -r df954e47ff67 drivers/infiniband/hw/ipath/ipath_verbs.c
--- a/drivers/infiniband/hw/ipath/ipath_verbs.c Fri May 12 15:55:28 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.c Fri May 12 15:55:28 2006 -0700
@@ -81,6 +81,32 @@ unsigned int ib_ipath_max_sges = 0xFF;
unsigned int ib_ipath_max_sges = 0xFF;
module_param_named(max_sges, ib_ipath_max_sges, uint, S_IWUSR | S_IRUGO);
MODULE_PARM_DESC(max_sges, "Maximum number of SGEs to support");
+
+unsigned int ib_ipath_max_mcast_grps = 16384;
+module_param_named(max_mcast_grps, ib_ipath_max_mcast_grps, uint,
+ S_IWUSR | S_IRUGO);
+MODULE_PARM_DESC(max_mcast_grps,
+ "Maximum number of multicast groups to support");
+
+unsigned int ib_ipath_max_mcast_qp_attached = 16;
+module_param_named(max_mcast_qp_attached, ib_ipath_max_mcast_qp_attached,
+ uint, S_IWUSR | S_IRUGO);
+MODULE_PARM_DESC(max_mcast_qp_attached,
+ "Maximum number of attached QPs to support");
+
+unsigned int ib_ipath_max_srqs = 1024;
+module_param_named(max_srqs, ib_ipath_max_srqs, uint, S_IWUSR | S_IRUGO);
+MODULE_PARM_DESC(max_srqs, "Maximum number of SRQs to support");
+
+unsigned int ib_ipath_max_srq_sges = 128;
+module_param_named(max_srq_sges, ib_ipath_max_srq_sges,
+ uint, S_IWUSR | S_IRUGO);
+MODULE_PARM_DESC(max_srq_sges, "Maximum number of SRQ SGEs to support");
+
+unsigned int ib_ipath_max_srq_wrs = 0x1FFFF;
+module_param_named(max_srq_wrs, ib_ipath_max_srq_wrs,
+ uint, S_IWUSR | S_IRUGO);
+MODULE_PARM_DESC(max_srq_wrs, "Maximum number of SRQ WRs support");
MODULE_LICENSE("GPL");
MODULE_AUTHOR("PathScale <support@pathscale.com>");
@@ -621,14 +647,14 @@ static int ipath_query_device(struct ib_
props->max_qp_rd_atom = 1;
props->max_qp_init_rd_atom = 1;
/* props->max_res_rd_atom */
- props->max_srq = 0xffff;
- props->max_srq_wr = 0xffff;
- props->max_srq_sge = 255;
+ props->max_srq = ib_ipath_max_srqs;
+ props->max_srq_wr = ib_ipath_max_srq_wrs;
+ props->max_srq_sge = ib_ipath_max_srq_sges;
/* props->local_ca_ack_delay */
props->atomic_cap = IB_ATOMIC_HCA;
props->max_pkeys = ipath_layer_get_npkeys(dev->dd);
- props->max_mcast_grp = 0xffff;
- props->max_mcast_qp_attach = 0xffff;
+ props->max_mcast_grp = ib_ipath_max_mcast_grps;
+ props->max_mcast_qp_attach = ib_ipath_max_mcast_qp_attached;
props->max_total_mcast_qp_attach = props->max_mcast_qp_attach *
props->max_mcast_grp;
diff -r c5f3731224bb -r df954e47ff67 drivers/infiniband/hw/ipath/ipath_verbs.h
--- a/drivers/infiniband/hw/ipath/ipath_verbs.h Fri May 12 15:55:28 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.h Fri May 12 15:55:28 2006 -0700
@@ -148,6 +148,7 @@ struct ipath_mcast {
struct list_head qp_list;
wait_queue_head_t wait;
atomic_t refcount;
+ int n_attached;
};
/* Memory region */
@@ -434,6 +435,7 @@ struct ipath_ibdev {
u32 n_pds_allocated; /* number of PDs allocated for device */
u32 n_ahs_allocated; /* number of AHs allocated for device */
u32 n_cqs_allocated; /* number of CQs allocated for device */
+ u32 n_mcast_grps_allocated; /* number of mcast groups allocated */
u64 ipath_sword; /* total dwords sent (sample result) */
u64 ipath_rword; /* total dwords received (sample result) */
u64 ipath_spkts; /* total packets sent (sample result) */
@@ -699,6 +701,16 @@ extern unsigned int ib_ipath_max_qp_wrs;
extern unsigned int ib_ipath_max_sges;
+extern unsigned int ib_ipath_max_mcast_grps;
+
+extern unsigned int ib_ipath_max_mcast_qp_attached;
+
+extern unsigned int ib_ipath_max_srqs;
+
+extern unsigned int ib_ipath_max_srq_sges;
+
+extern unsigned int ib_ipath_max_srq_wrs;
+
extern const u32 ib_ipath_rnr_table[];
#endif /* IPATH_VERBS_H */
diff -r c5f3731224bb -r df954e47ff67 drivers/infiniband/hw/ipath/ipath_verbs_mcast.c
--- a/drivers/infiniband/hw/ipath/ipath_verbs_mcast.c Fri May 12 15:55:28 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs_mcast.c Fri May 12 15:55:28 2006 -0700
@@ -92,6 +92,7 @@ static struct ipath_mcast *ipath_mcast_a
INIT_LIST_HEAD(&mcast->qp_list);
init_waitqueue_head(&mcast->wait);
atomic_set(&mcast->refcount, 0);
+ mcast->n_attached = 0;
bail:
return mcast;
@@ -157,7 +158,8 @@ bail:
* the table but the QP was added. Return ESRCH if the QP was already
* attached and neither structure was added.
*/
-static int ipath_mcast_add(struct ipath_mcast *mcast,
+static int ipath_mcast_add(struct ipath_ibdev *dev,
+ struct ipath_mcast *mcast,
struct ipath_mcast_qp *mqp)
{
struct rb_node **n = &mcast_tree.rb_node;
@@ -188,16 +190,28 @@ static int ipath_mcast_add(struct ipath_
/* Search the QP list to see if this is already there. */
list_for_each_entry_rcu(p, &tmcast->qp_list, list) {
if (p->qp == mqp->qp) {
- spin_unlock_irqrestore(&mcast_lock, flags);
ret = ESRCH;
goto bail;
}
}
+ if (tmcast->n_attached == ib_ipath_max_mcast_qp_attached) {
+ ret = ENOMEM;
+ goto bail;
+ }
+
+ tmcast->n_attached++;
+
list_add_tail_rcu(&mqp->list, &tmcast->qp_list);
- spin_unlock_irqrestore(&mcast_lock, flags);
ret = EEXIST;
goto bail;
}
+
+ if (dev->n_mcast_grps_allocated == ib_ipath_max_mcast_grps) {
+ ret = ENOMEM;
+ goto bail;
+ }
+
+ dev->n_mcast_grps_allocated++;
list_add_tail_rcu(&mqp->list, &mcast->qp_list);
@@ -205,17 +219,18 @@ static int ipath_mcast_add(struct ipath_
rb_link_node(&mcast->rb_node, pn, n);
rb_insert_color(&mcast->rb_node, &mcast_tree);
+ ret = 0;
+
+bail:
spin_unlock_irqrestore(&mcast_lock, flags);
- ret = 0;
-
-bail:
return ret;
}
int ipath_multicast_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
{
struct ipath_qp *qp = to_iqp(ibqp);
+ struct ipath_ibdev *dev = to_idev(ibqp->device);
struct ipath_mcast *mcast;
struct ipath_mcast_qp *mqp;
int ret;
@@ -235,7 +250,7 @@ int ipath_multicast_attach(struct ib_qp
ret = -ENOMEM;
goto bail;
}
- switch (ipath_mcast_add(mcast, mqp)) {
+ switch (ipath_mcast_add(dev, mcast, mqp)) {
case ESRCH:
/* Neither was used: can't attach the same QP twice. */
ipath_mcast_qp_free(mqp);
@@ -245,6 +260,12 @@ int ipath_multicast_attach(struct ib_qp
case EEXIST: /* The mcast wasn't used */
ipath_mcast_free(mcast);
break;
+ case ENOMEM:
+ /* Exceeded the maximum number of mcast groups. */
+ ipath_mcast_qp_free(mqp);
+ ipath_mcast_free(mcast);
+ ret = -ENOMEM;
+ goto bail;
default:
break;
}
@@ -258,6 +279,7 @@ int ipath_multicast_detach(struct ib_qp
int ipath_multicast_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
{
struct ipath_qp *qp = to_iqp(ibqp);
+ struct ipath_ibdev *dev = to_idev(ibqp->device);
struct ipath_mcast *mcast = NULL;
struct ipath_mcast_qp *p, *tmp;
struct rb_node *n;
@@ -296,6 +318,7 @@ int ipath_multicast_detach(struct ib_qp
* link until we are sure there are no list walkers.
*/
list_del_rcu(&p->list);
+ mcast->n_attached--;
/* If this was the last attached QP, remove the GID too. */
if (list_empty(&mcast->qp_list)) {
@@ -319,6 +342,7 @@ int ipath_multicast_detach(struct ib_qp
atomic_dec(&mcast->refcount);
wait_event(mcast->wait, !atomic_read(&mcast->refcount));
ipath_mcast_free(mcast);
+ dev->n_mcast_grps_allocated--;
}
ret = 0;
^ permalink raw reply
* [PATCH 3 of 53] ipath - report max MR and QP sizes based on table sizes
From: Bryan O'Sullivan @ 2006-05-12 23:42 UTC (permalink / raw)
To: rdreier; +Cc: openib-general, linux-kernel
In-Reply-To: <patchbomb.1147477365@eng-12.pathscale.com>
Report max MR based on the lkey table size.
Report max QP based on the QP table size.
Signed-off-by: Bryan O'Sullivan <bos@pathscale.com>
diff -r 3ab7a7b10bf2 -r 5d5e1e641b16 drivers/infiniband/hw/ipath/ipath_verbs.c
--- a/drivers/infiniband/hw/ipath/ipath_verbs.c Fri May 12 15:55:27 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.c Fri May 12 15:55:27 2006 -0700
@@ -583,12 +583,12 @@ static int ipath_query_device(struct ib_
props->sys_image_guid = dev->sys_image_guid;
props->max_mr_size = ~0ull;
- props->max_qp = 0xffff;
+ props->max_qp = dev->qp_table.max;
props->max_qp_wr = 0xffff;
props->max_sge = 255;
props->max_cq = 0xffff;
props->max_cqe = 0xffff;
- props->max_mr = 0xffff;
+ props->max_mr = dev->lk_table.max;
props->max_pd = 0xffff;
props->max_qp_rd_atom = 1;
props->max_qp_init_rd_atom = 1;
^ permalink raw reply
* [PATCH 4 of 53] ipath - cap number of PDs that can be allocated
From: Bryan O'Sullivan @ 2006-05-12 23:42 UTC (permalink / raw)
To: rdreier; +Cc: openib-general, linux-kernel
In-Reply-To: <patchbomb.1147477365@eng-12.pathscale.com>
Put an arbitrary cap on the maximum number of PDs that can be allocated
for a device. This is arbitrary because the number we support
is constrained only by system memory and what kmalloc can give us.
Nevertheless, if we don't have a limit, some third-party OpenIB stress
tests fail. The limit can be changed on the fly using a module parameter.
Signed-off-by: Bryan O'Sullivan <bos@pathscale.com>
diff -r 5d5e1e641b16 -r 300f0aa6f034 drivers/infiniband/hw/ipath/ipath_verbs.c
--- a/drivers/infiniband/hw/ipath/ipath_verbs.c Fri May 12 15:55:27 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.c Fri May 12 15:55:27 2006 -0700
@@ -54,6 +54,11 @@ unsigned int ib_ipath_debug; /* debug ma
unsigned int ib_ipath_debug; /* debug mask */
module_param_named(debug, ib_ipath_debug, uint, S_IWUSR | S_IRUGO);
MODULE_PARM_DESC(debug, "Verbs debug mask");
+
+static unsigned int ib_ipath_max_pds = 0xFFFF;
+module_param_named(max_pds, ib_ipath_max_pds, uint, S_IWUSR | S_IRUGO);
+MODULE_PARM_DESC(max_pds,
+ "Maximum number of protection domains to support");
MODULE_LICENSE("GPL");
MODULE_AUTHOR("PathScale <support@pathscale.com>");
@@ -589,7 +594,7 @@ static int ipath_query_device(struct ib_
props->max_cq = 0xffff;
props->max_cqe = 0xffff;
props->max_mr = dev->lk_table.max;
- props->max_pd = 0xffff;
+ props->max_pd = ib_ipath_max_pds;
props->max_qp_rd_atom = 1;
props->max_qp_init_rd_atom = 1;
/* props->max_res_rd_atom */
@@ -743,8 +748,23 @@ static struct ib_pd *ipath_alloc_pd(stru
struct ib_ucontext *context,
struct ib_udata *udata)
{
+ struct ipath_ibdev *dev = to_idev(ibdev);
struct ipath_pd *pd;
struct ib_pd *ret;
+
+ /*
+ * This is actually totally arbitrary. Some correctness tests
+ * assume there's a maximum number of PDs that can be allocated.
+ * We don't actually have this limit, but we fail the test if
+ * we allow allocations of more than we report for this value.
+ */
+
+ if (dev->n_pds_allocated == ib_ipath_max_pds) {
+ ret = ERR_PTR(-ENOMEM);
+ goto bail;
+ }
+
+ dev->n_pds_allocated++;
pd = kmalloc(sizeof *pd, GFP_KERNEL);
if (!pd) {
@@ -764,6 +784,9 @@ static int ipath_dealloc_pd(struct ib_pd
static int ipath_dealloc_pd(struct ib_pd *ibpd)
{
struct ipath_pd *pd = to_ipd(ibpd);
+ struct ipath_ibdev *dev = to_idev(ibpd->device);
+
+ dev->n_pds_allocated--;
kfree(pd);
diff -r 5d5e1e641b16 -r 300f0aa6f034 drivers/infiniband/hw/ipath/ipath_verbs.h
--- a/drivers/infiniband/hw/ipath/ipath_verbs.h Fri May 12 15:55:27 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_verbs.h Fri May 12 15:55:27 2006 -0700
@@ -431,6 +431,7 @@ struct ipath_ibdev {
__be64 sys_image_guid; /* in network order */
__be64 gid_prefix; /* in network order */
__be64 mkey;
+ u32 n_pds_allocated; /* number of PDs allocated for device */
u64 ipath_sword; /* total dwords sent (sample result) */
u64 ipath_rword; /* total dwords received (sample result) */
u64 ipath_spkts; /* total packets sent (sample result) */
^ permalink raw reply
* [PATCH 0 of 53] ipath driver updates for 2.6.17-rc4
From: Bryan O'Sullivan @ 2006-05-12 23:42 UTC (permalink / raw)
To: rdreier; +Cc: openib-general, linux-kernel
[-- Attachment #1: Type: text/plain, Size: 255 bytes --]
Hi, Roland -
Here is a series of patches to bring the ipath driver up to date. I
believe you may already have two of them (but I've included them just
in case), but the others should all be new.
They apply on top of Linus's current -git.
Cheers,
<b
^ permalink raw reply
* [PATCH 1 of 53] ipath - fix spinlock recursion bug
From: Bryan O'Sullivan @ 2006-05-12 23:42 UTC (permalink / raw)
To: rdreier; +Cc: openib-general, linux-kernel
In-Reply-To: <patchbomb.1147477365@eng-12.pathscale.com>
The local loopback path for RC can lock the rkey table lock without
blocking interrupts. The receive interrupt path can then call
ipath_rkey_ok() and deadlock. Since the lock only protects a 64 bit read,
the lock isn't needed.
Signed-off-by: Bryan O'Sullivan <bos@pathscale.com>
diff -r 89f7c69a68bf -r 9b9f24aab350 drivers/infiniband/hw/ipath/ipath_keys.c
--- a/drivers/infiniband/hw/ipath/ipath_keys.c Fri May 12 15:55:27 2006 -0700
+++ b/drivers/infiniband/hw/ipath/ipath_keys.c Fri May 12 15:55:27 2006 -0700
@@ -136,9 +136,7 @@ int ipath_lkey_ok(struct ipath_lkey_tabl
ret = 1;
goto bail;
}
- spin_lock(&rkt->lock);
mr = rkt->table[(sge->lkey >> (32 - ib_ipath_lkey_table_size))];
- spin_unlock(&rkt->lock);
if (unlikely(mr == NULL || mr->lkey != sge->lkey)) {
ret = 0;
goto bail;
@@ -184,8 +182,6 @@ bail:
* @acc: access flags
*
* Return 1 if successful, otherwise 0.
- *
- * The QP r_rq.lock should be held.
*/
int ipath_rkey_ok(struct ipath_ibdev *dev, struct ipath_sge_state *ss,
u32 len, u64 vaddr, u32 rkey, int acc)
@@ -196,9 +192,7 @@ int ipath_rkey_ok(struct ipath_ibdev *de
size_t off;
int ret;
- spin_lock(&rkt->lock);
mr = rkt->table[(rkey >> (32 - ib_ipath_lkey_table_size))];
- spin_unlock(&rkt->lock);
if (unlikely(mr == NULL || mr->lkey != rkey)) {
ret = 0;
goto bail;
^ permalink raw reply
* Re: [BUG 2.6.17-git] kmem_cache_create: duplicate cache scsi_cmd_cache
From: Al Viro @ 2006-05-12 23:37 UTC (permalink / raw)
To: Linus Torvalds
Cc: Erik Mouw, Or Gerlitz, linux-scsi, axboe,
Linux Kernel Mailing List
In-Reply-To: <20060512232131.GV27946@ftp.linux.org.uk>
On Sat, May 13, 2006 at 12:21:31AM +0100, Al Viro wrote:
> The problem is with disk->driverfs_dev, not disk itself. Block layer
> has no fscking business touching it after del_gendisk() - if nothing else,
> we might have _no_ underlying object at all from the very beginning.
>
> So anything that wants events related to partitions, let alone mounting,
> can't expect to see PHYSDEV... crap. Moreover, it can bloody well
> get to PHYSDEV... itself *if* it wants to and if it's there. There's
> a reason why we have that symlink in sys/block/<device> and userland can
> bloody well access it on its own.
BTW, the best option is to kill bdev_uevent() again. Short of that,
skip PHYSDEV mess if disk doesn't have GENHD_FL_UP.
^ permalink raw reply
* vutubeflorifidy
From: Dr. David Frankson @ 2006-05-12 23:36 UTC (permalink / raw)
To: sparclinux
You want new medication online http://rofujo.bpif.net/
^ permalink raw reply
* Re: [patch] smbus unhiding kills thermal management
From: Zwane Mwaikambo @ 2006-05-12 23:34 UTC (permalink / raw)
To: Carl-Daniel Hailfinger
Cc: Pavel Machek, Andrew Morton, kernel list, trenn, thoenig, stable
In-Reply-To: <44645FC2.80500@gmx.net>
On Fri, 12 May 2006, Carl-Daniel Hailfinger wrote:
> Pavel Machek wrote:
> > Do not enable the SMBus device on Asus boards if suspend
> > is used. We do not reenable the device on resume, leading to all sorts
> > of undesirable effects, the worst being a total fan failure after
> > resume on Samsung P35 laptop.
> >
> > Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
> > Signed-off-by: Pavel Machek <pavel@suse.cz>
>
> This is probably also -stable material.
Isn't it inevitable that we're going to have to rerun quirks on resume on
some hardware?
^ permalink raw reply
* Bigphysarea vs. kernel 2.4.32 and PPC405GPr
From: Stephen Williams @ 2006-05-12 23:31 UTC (permalink / raw)
To: linuxppc-embedded
I'm trying to use the bigphysarea patch to help me allocate big
physical chunks of memory for use by some custom embedded devices.
I've applied the bigphysarea-2.4.20 patch and built, no trouble,
and I've got it to config for my PPC. I can see at boot time that
the bigphysarea is getting its pages.
I'm trying to use a mmap call to the driver to map a chunk of this
memory into the process. The mmap for the driver has this:
pgprot_val(vma->vm_page_prot) |= _PAGE_NO_CACHE;
vma->vm_flags |= VM_RESERVED;
npages = (vma->vm_end - vma->vm_start) / PAGE_SIZE;
heap_map_base = bigphysarea_alloc_pages(npages, 1, GFP_KERNEL);
printk("XXXX Map base=%p, %ld pages\n", heap_map_base, npages);
rc = remap_page_range(vma->vm_start,
(unsigned long)heap_map_base,
npages*PAGE_SIZE,
vma->vm_page_prot);
The mmap returns without an error, but any access to the mapped
region gets me an immediate "Oops: machine check, sig: 7". I don't
see it. Where is the stupid mistake that I'm invariably making?
--
Steve Williams "The woods are lovely, dark and deep.
steve at icarus.com But I have promises to keep,
http://www.icarus.com and lines to code before I sleep,
http://www.picturel.com And lines to code before I sleep."
^ permalink raw reply
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.