From 2d6f0d0cd94f9b8b24102300d8dd9cbbd1688826 Mon Sep 17 00:00:00 2001 From: James Bottomley Date: Sun, 4 May 2008 22:34:49 -0500 Subject: [PATCH 1/7] [SCSI] gdth: fix timer handling The global timer handling is problematic in that if someone unbinds a PCI gdth instance, the BUG_ON() in the timer will cause a panic. Fix this by making the timer start and stop depending on whether there are instances present. This should also permit binding and unbinding to work. Signed-off-by: James Bottomley --- drivers/scsi/gdth.c | 38 ++++++++++++++++++++++++++------------ 1 file changed, 26 insertions(+), 12 deletions(-) diff --git a/drivers/scsi/gdth.c b/drivers/scsi/gdth.c index 8e2e964af668..16785a2ad035 100644 --- a/drivers/scsi/gdth.c +++ b/drivers/scsi/gdth.c @@ -3724,6 +3724,8 @@ static void gdth_log_event(gdth_evt_data *dvr, char *buffer) } #ifdef GDTH_STATISTICS +static unchar gdth_timer_running; + static void gdth_timeout(ulong data) { ulong32 i; @@ -3731,7 +3733,10 @@ static void gdth_timeout(ulong data) gdth_ha_str *ha; ulong flags; - BUG_ON(list_empty(&gdth_instances)); + if(unlikely(list_empty(&gdth_instances))) { + gdth_timer_running = 0; + return; + } ha = list_first_entry(&gdth_instances, gdth_ha_str, list); spin_lock_irqsave(&ha->smp_lock, flags); @@ -3751,6 +3756,22 @@ static void gdth_timeout(ulong data) add_timer(&gdth_timer); spin_unlock_irqrestore(&ha->smp_lock, flags); } + +static void gdth_timer_init(void) +{ + if (gdth_timer_running) + return; + gdth_timer_running = 1; + TRACE2(("gdth_detect(): Initializing timer !\n")); + gdth_timer.expires = jiffies + HZ; + gdth_timer.data = 0L; + gdth_timer.function = gdth_timeout; + add_timer(&gdth_timer); +} +#else +static inline void gdth_timer_init(void) +{ +} #endif static void __init internal_setup(char *str,int *ints) @@ -4735,6 +4756,7 @@ static int __init gdth_isa_probe_one(ulong32 isa_bios) if (error) goto out_free_coal_stat; list_add_tail(&ha->list, &gdth_instances); + gdth_timer_init(); scsi_scan_host(shp); @@ -4865,6 +4887,7 @@ static int __init gdth_eisa_probe_one(ushort eisa_slot) if (error) goto out_free_coal_stat; list_add_tail(&ha->list, &gdth_instances); + gdth_timer_init(); scsi_scan_host(shp); @@ -5011,6 +5034,7 @@ static int gdth_pci_probe_one(gdth_pci_str *pcistr, list_add_tail(&ha->list, &gdth_instances); pci_set_drvdata(ha->pdev, ha); + gdth_timer_init(); scsi_scan_host(shp); @@ -5110,6 +5134,7 @@ static int __init gdth_init(void) /* initializations */ gdth_polling = TRUE; gdth_clear_events(); + init_timer(&gdth_timer); /* As default we do not probe for EISA or ISA controllers */ if (probe_eisa_isa) { @@ -5138,17 +5163,6 @@ static int __init gdth_init(void) TRACE2(("gdth_detect() %d controller detected\n", gdth_ctr_count)); - if (list_empty(&gdth_instances)) - return -ENODEV; - -#ifdef GDTH_STATISTICS - TRACE2(("gdth_detect(): Initializing timer !\n")); - init_timer(&gdth_timer); - gdth_timer.expires = jiffies + HZ; - gdth_timer.data = 0L; - gdth_timer.function = gdth_timeout; - add_timer(&gdth_timer); -#endif major = register_chrdev(0,"gdth", &gdth_fops); register_reboot_notifier(&gdth_notifier); gdth_polling = FALSE; From a85591fd0baf4ed3f03ee1aaac6a985e400cf089 Mon Sep 17 00:00:00 2001 From: James Bottomley Date: Sun, 4 May 2008 22:35:58 -0500 Subject: [PATCH 2/7] [SCSI] gdth: fix Error: Driver 'gdth' is already registered, aborting... This message appears on modprobe/rmmod/modprobe of the driver. It's caused because if the driver has no instances, it returns an error from gdth_init, which causes the module to fail to load. Unfortunately, the module's pci driver is still registered at this point. Fix this by making gdth behave like a modern driver and insert even if it doesn't find any instances (in case of hot plug or software driven binding). Signed-off-by: James Bottomley --- drivers/scsi/gdth.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-) diff --git a/drivers/scsi/gdth.c b/drivers/scsi/gdth.c index 16785a2ad035..46771d4c81bd 100644 --- a/drivers/scsi/gdth.c +++ b/drivers/scsi/gdth.c @@ -550,7 +550,6 @@ static int __init gdth_search_isa(ulong32 bios_adr) #endif /* CONFIG_ISA */ #ifdef CONFIG_PCI -static bool gdth_pci_registered; static bool gdth_search_vortex(ushort device) { @@ -5157,8 +5156,13 @@ static int __init gdth_init(void) #ifdef CONFIG_PCI /* scanning for PCI controllers */ - if (pci_register_driver(&gdth_pci_driver) == 0) - gdth_pci_registered = true; + if (pci_register_driver(&gdth_pci_driver)) { + gdth_ha_str *ha; + + list_for_each_entry(ha, &gdth_instances, list) + gdth_remove_one(ha); + return -ENODEV; + } #endif /* CONFIG_PCI */ TRACE2(("gdth_detect() %d controller detected\n", gdth_ctr_count)); @@ -5181,8 +5185,7 @@ static void __exit gdth_exit(void) #endif #ifdef CONFIG_PCI - if (gdth_pci_registered) - pci_unregister_driver(&gdth_pci_driver); + pci_unregister_driver(&gdth_pci_driver); #endif list_for_each_entry(ha, &gdth_instances, list) From 4cf1043593db6a337f10e006c23c69e5fc93e722 Mon Sep 17 00:00:00 2001 From: Mike Christie Date: Wed, 7 May 2008 20:43:52 -0500 Subject: [PATCH 3/7] [SCSI] libiscsi regression in 2.6.25: fix nop timer handling The following patch fixes a bug in the iscsi nop processing. The target sends iscsi nops to ping the initiator and the initiator has to send nops to reply and can send nops to ping the target. In 2.6.25 we moved the nop processing to the kernel to handle problems when the userspace daemon is not up, but the target is pinging us, and to handle when scsi commands timeout, but the transport may be the cause (we can send a nop to check the transport). When we added this code we added a bug where if the transport timer wakes at the exact same time we are supposed to check for a nop timeout we drop the session instead of checking the transport. This patch checks if a iscsi ping is outstanding and if the ping has timed out, to determine if we need to signal a connection problem. Signed-off-by: Mike Christie Cc: Stable Tree Signed-off-by: James Bottomley --- drivers/scsi/libiscsi.c | 17 +++++++++-------- 1 file changed, 9 insertions(+), 8 deletions(-) diff --git a/drivers/scsi/libiscsi.c b/drivers/scsi/libiscsi.c index 010c1b9b178c..98164f3c3517 100644 --- a/drivers/scsi/libiscsi.c +++ b/drivers/scsi/libiscsi.c @@ -1453,19 +1453,20 @@ static void iscsi_check_transport_timeouts(unsigned long data) { struct iscsi_conn *conn = (struct iscsi_conn *)data; struct iscsi_session *session = conn->session; - unsigned long timeout, next_timeout = 0, last_recv; + unsigned long recv_timeout, next_timeout = 0, last_recv; spin_lock(&session->lock); if (session->state != ISCSI_STATE_LOGGED_IN) goto done; - timeout = conn->recv_timeout; - if (!timeout) + recv_timeout = conn->recv_timeout; + if (!recv_timeout) goto done; - timeout *= HZ; + recv_timeout *= HZ; last_recv = conn->last_recv; - if (time_before_eq(last_recv + timeout + (conn->ping_timeout * HZ), + if (conn->ping_mtask && + time_before_eq(conn->last_ping + (conn->ping_timeout * HZ), jiffies)) { iscsi_conn_printk(KERN_ERR, conn, "ping timeout of %d secs " "expired, last rx %lu, last ping %lu, " @@ -1476,15 +1477,15 @@ static void iscsi_check_transport_timeouts(unsigned long data) return; } - if (time_before_eq(last_recv + timeout, jiffies)) { + if (time_before_eq(last_recv + recv_timeout, jiffies)) { if (time_before_eq(conn->last_ping, last_recv)) { /* send a ping to try to provoke some traffic */ debug_scsi("Sending nopout as ping on conn %p\n", conn); iscsi_send_nopout(conn, NULL); } - next_timeout = last_recv + timeout + (conn->ping_timeout * HZ); + next_timeout = conn->last_ping + (conn->ping_timeout * HZ); } else - next_timeout = last_recv + timeout; + next_timeout = last_recv + recv_timeout; debug_scsi("Setting next tmo %lu\n", next_timeout); mod_timer(&conn->transport_timer, next_timeout); From c8611f975403dd20e6503aff8aded5dcb718f75b Mon Sep 17 00:00:00 2001 From: Mike Christie Date: Thu, 8 May 2008 20:15:34 -0500 Subject: [PATCH 4/7] [SCSI] libiscsi regression in 2.6.25: fix setting of recv timer If the ping tmo is longer than the recv tmo then we could miss a window where we were supposed to check the recv tmo. This happens because the ping code will set the next timeout for the ping timeout, and if the ping executes quickly there will be a long chunk of time before the timer wakes up again. This patch has the ping processing code kick off a recv tmo check when getting a nop in response to our ping. Signed-off-by: Mike Christie Cc: Stable Tree Signed-off-by: James Bottomley --- drivers/scsi/libiscsi.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/drivers/scsi/libiscsi.c b/drivers/scsi/libiscsi.c index 98164f3c3517..b43bf1d60dac 100644 --- a/drivers/scsi/libiscsi.c +++ b/drivers/scsi/libiscsi.c @@ -730,7 +730,9 @@ static int __iscsi_complete_pdu(struct iscsi_conn *conn, struct iscsi_hdr *hdr, if (iscsi_recv_pdu(conn->cls_conn, hdr, data, datalen)) rc = ISCSI_ERR_CONN_FAILED; - } + } else + mod_timer(&conn->transport_timer, + jiffies + conn->recv_timeout); iscsi_free_mgmt_task(conn, mtask); break; default: @@ -1478,11 +1480,9 @@ static void iscsi_check_transport_timeouts(unsigned long data) } if (time_before_eq(last_recv + recv_timeout, jiffies)) { - if (time_before_eq(conn->last_ping, last_recv)) { - /* send a ping to try to provoke some traffic */ - debug_scsi("Sending nopout as ping on conn %p\n", conn); - iscsi_send_nopout(conn, NULL); - } + /* send a ping to try to provoke some traffic */ + debug_scsi("Sending nopout as ping on conn %p\n", conn); + iscsi_send_nopout(conn, NULL); next_timeout = conn->last_ping + (conn->ping_timeout * HZ); } else next_timeout = last_recv + recv_timeout; From ad2fa42d044b98469449880474a9662fb689f7f9 Mon Sep 17 00:00:00 2001 From: James Bottomley Date: Sat, 10 May 2008 08:33:58 -0500 Subject: [PATCH 5/7] [SCSI] aha152x: fix init suspiciously returned 1, it should follow 0/-E convention Reported-by: Frank de Jong > [1.] One line summary of the problem: > linux-2.6.25.3, aha152x'->init suspiciously returned 1, it should > follow 0/-E convention. The module / driver works okay. Unloading the > module is impossible. The driver is apparently returning 0 on failure and 1 on success. That's a bit unfortunate. Fix it by altering to -ENODEV and 0. Cc: Stable Tree Signed-off-by: James Bottomley --- drivers/scsi/aha152x.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/scsi/aha152x.c b/drivers/scsi/aha152x.c index f5215fd4b73d..f0c4ffceabbe 100644 --- a/drivers/scsi/aha152x.c +++ b/drivers/scsi/aha152x.c @@ -3830,7 +3830,7 @@ static int __init aha152x_init(void) iounmap(p); } if (!ok && setup_count == 0) - return 0; + return -ENODEV; printk(KERN_INFO "aha152x: BIOS test: passed, "); #else @@ -3909,7 +3909,7 @@ static int __init aha152x_init(void) #endif } - return 1; + return 0; } static void __exit aha152x_exit(void) From 64976a0387835a7ac61bbe2a99b27ccae34eac5d Mon Sep 17 00:00:00 2001 From: James Bottomley Date: Sat, 10 May 2008 14:08:40 -0500 Subject: [PATCH 6/7] [SCSI] aha152x: Fix oops on module removal Reported-by: Frank de Jong > after trying to unload the module: > BUG: unable to handle kernel paging request at 00100100 > IP: [] :aha152x:aha152x_exit+0x47/0x6a > *pde = 00000000 > Oops: 0000 [#1] PREEMPT SMP > Modules linked in: aha152x(-) w83781d hwmon_vid tun ne 8390 bonding > usb_storage snd_usb_audio snd_usb_lib snd_rawmidi pwc snd_seq_device > compat_ioctl32 snd_hwdep videodev v4l1_compat 3c59x mii intel_agp > agpgart snd_pcm_oss snd_pcm snd_timer snd_page_alloc snd_mixer_oss snd > > Pid: 2837, comm: rmmod Not tainted (2.6.25.3 #1) > EIP: 0060:[] EFLAGS: 00210212 CPU: 0 > EIP is at aha152x_exit+0x47/0x6a [aha152x] > EAX: 00000001 EBX: 000ffdc4 ECX: f7c517a8 EDX: 00000001 > ESI: 00000000 EDI: 00000003 EBP: e7880000 ESP: e7881f58 > DS: 007b ES: 007b FS: 00d8 GS: 0033 SS: 0068 > Process rmmod (pid: 2837, ti=e7880000 task=f27eb580 task.ti=e7880000) > Stack: fba03700 c01419d2 31616861 00783235 e795ee70 c0157709 b7f24000 e79ae000 > c0158271 ffffffff b7f25000 e79ae004 e795e370 b7f25000 e795e37c e795e370 > 009ae000 fba03700 00000880 e7881fa8 00000000 bf93ec20 bf93ec20 c0102faa > Call Trace: > [] sys_delete_module+0x112/0x1a0 > [] remove_vma+0x39/0x50 > [] do_munmap+0x181/0x1f0 > [] sysenter_past_esp+0x5f/0x85 > [] rsc_parse+0x0/0x3c0 The problem is that the driver calls aha152x_release() under a list_for_each_entry(). Unfortunately, aha152x_release() deletes from the list in question. Fix this by using list_for_each_entry_safe(). Cc: Stable Tree Signed-off-by: James Bottomley --- drivers/scsi/aha152x.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/scsi/aha152x.c b/drivers/scsi/aha152x.c index f0c4ffceabbe..1dca1775f4b1 100644 --- a/drivers/scsi/aha152x.c +++ b/drivers/scsi/aha152x.c @@ -3914,9 +3914,9 @@ static int __init aha152x_init(void) static void __exit aha152x_exit(void) { - struct aha152x_hostdata *hd; + struct aha152x_hostdata *hd, *tmp; - list_for_each_entry(hd, &aha152x_host_list, host_list) { + list_for_each_entry_safe(hd, tmp, &aha152x_host_list, host_list) { struct Scsi_Host *shost = container_of((void *)hd, struct Scsi_Host, hostdata); aha152x_release(shost); From af5741c6de4f4a1d8608b0f00867c77cb7123635 Mon Sep 17 00:00:00 2001 From: Jeremy Higdon Date: Sun, 11 May 2008 23:17:03 -0700 Subject: [PATCH 7/7] [SCSI] qla1280: Fix queue depth problem The qla1280 driver was ANDing the output value of mailbox register 0 with (1 << target-number) to determine whether to enable queueing on the target in question. But mailbox register 0 has the status code for the mailbox command (in this case, Set Target Parameters). Potential values are: /* * ISP mailbox command complete status codes */ So clearly that is in error. I can't think what the author of that line was looking for in a mailbox register, so I just eliminated the AND. flag is used later in the function, and I think that the later usage was also wrong, though it was used to set values that aren't used. Oh well, an overhaul of this driver is not what I want to do now -- just a bugfix. After the fix, I found that my disks were getting a queue depth of 255, which is far too many. Most SCSI disks are limited to 32 or 64. In any case, there's no point, queueing up a bunch of commands to the adapter that will just result in queue full or starve other targets from being issued commands due to running out of internal memory. So I dropped default queue depth to 32 (from which 1 is subtracted elsewhere, giving net of 31). I tested with a Seagate ST336753LC, and results look good, so I'm satisfied with this patch. Signed-off-by: Jeremy Higdon Acked-by: Jes Sorensen Cc: Stable Tree Signed-off-by: James Bottomley --- drivers/scsi/qla1280.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/scsi/qla1280.c b/drivers/scsi/qla1280.c index fa060932d2b4..51e2f299dbbb 100644 --- a/drivers/scsi/qla1280.c +++ b/drivers/scsi/qla1280.c @@ -2007,7 +2007,7 @@ qla1280_set_defaults(struct scsi_qla_host *ha) nv->bus[bus].config_2.req_ack_active_negation = 1; nv->bus[bus].config_2.data_line_active_negation = 1; nv->bus[bus].selection_timeout = 250; - nv->bus[bus].max_queue_depth = 256; + nv->bus[bus].max_queue_depth = 32; if (IS_ISP1040(ha)) { nv->bus[bus].bus_reset_delay = 3; @@ -2051,7 +2051,7 @@ qla1280_config_target(struct scsi_qla_host *ha, int bus, int target) status = qla1280_mailbox_command(ha, 0x0f, mb); /* Save Tag queuing enable flag. */ - flag = (BIT_0 << target) & mb[0]; + flag = (BIT_0 << target); if (nv->bus[bus].target[target].parameter.tag_queuing) ha->bus_settings[bus].qtag_enables |= flag;