From 48cfb14f8b89d4d5b3df6c16f08b258686fb12ad Mon Sep 17 00:00:00 2001 From: Hannes Hering Date: Wed, 7 May 2008 14:43:36 +0200 Subject: ehea: Add DLPAR memory remove support The eHEA driver uses the recently modified walk_memory_resource for powerpc functionality to detect the memory layout. It further uses the memory hotplug notifiers to catch memory hotplug events. Signed-off-by: Hannes Hering Signed-off-by: Jeff Garzik --- drivers/net/ehea/ehea.h | 27 +++- drivers/net/ehea/ehea_main.c | 25 ++++ drivers/net/ehea/ehea_qmr.c | 286 +++++++++++++++++++++++++++++++------------ 3 files changed, 254 insertions(+), 84 deletions(-) (limited to 'drivers/net/ehea') diff --git a/drivers/net/ehea/ehea.h b/drivers/net/ehea/ehea.h index f5dacceab95b..fe872fbd671e 100644 --- a/drivers/net/ehea/ehea.h +++ b/drivers/net/ehea/ehea.h @@ -40,7 +40,7 @@ #include #define DRV_NAME "ehea" -#define DRV_VERSION "EHEA_0090" +#define DRV_VERSION "EHEA_0091" /* eHEA capability flags */ #define DLPAR_PORT_ADD_REM 1 @@ -118,6 +118,13 @@ #define EHEA_MR_ACC_CTRL 0x00800000 #define EHEA_BUSMAP_START 0x8000000000000000ULL +#define EHEA_INVAL_ADDR 0xFFFFFFFFFFFFFFFFULL +#define EHEA_DIR_INDEX_SHIFT 13 /* 8k Entries in 64k block */ +#define EHEA_TOP_INDEX_SHIFT (EHEA_DIR_INDEX_SHIFT * 2) +#define EHEA_MAP_ENTRIES (1 << EHEA_DIR_INDEX_SHIFT) +#define EHEA_MAP_SIZE (0x10000) /* currently fixed map size */ +#define EHEA_INDEX_MASK (EHEA_MAP_ENTRIES - 1) + #define EHEA_WATCH_DOG_TIMEOUT 10*HZ @@ -192,10 +199,20 @@ struct h_epas { set to 0 if unused */ }; -struct ehea_busmap { - unsigned int entries; /* total number of entries */ - unsigned int valid_sections; /* number of valid sections */ - u64 *vaddr; +/* + * Memory map data structures + */ +struct ehea_dir_bmap +{ + u64 ent[EHEA_MAP_ENTRIES]; +}; +struct ehea_top_bmap +{ + struct ehea_dir_bmap *dir[EHEA_MAP_ENTRIES]; +}; +struct ehea_bmap +{ + struct ehea_top_bmap *top[EHEA_MAP_ENTRIES]; }; struct ehea_qp; diff --git a/drivers/net/ehea/ehea_main.c b/drivers/net/ehea/ehea_main.c index f9bc21c74b59..d1b6d4e7495d 100644 --- a/drivers/net/ehea/ehea_main.c +++ b/drivers/net/ehea/ehea_main.c @@ -35,6 +35,7 @@ #include #include #include +#include #include #include @@ -3503,6 +3504,24 @@ void ehea_crash_handler(void) 0, H_DEREG_BCMC); } +static int ehea_mem_notifier(struct notifier_block *nb, + unsigned long action, void *data) +{ + switch (action) { + case MEM_OFFLINE: + ehea_info("memory has been removed"); + ehea_rereg_mrs(NULL); + break; + default: + break; + } + return NOTIFY_OK; +} + +static struct notifier_block ehea_mem_nb = { + .notifier_call = ehea_mem_notifier, +}; + static int ehea_reboot_notifier(struct notifier_block *nb, unsigned long action, void *unused) { @@ -3581,6 +3600,10 @@ int __init ehea_module_init(void) if (ret) ehea_info("failed registering reboot notifier"); + ret = register_memory_notifier(&ehea_mem_nb); + if (ret) + ehea_info("failed registering memory remove notifier"); + ret = crash_shutdown_register(&ehea_crash_handler); if (ret) ehea_info("failed registering crash handler"); @@ -3604,6 +3627,7 @@ int __init ehea_module_init(void) out3: ibmebus_unregister_driver(&ehea_driver); out2: + unregister_memory_notifier(&ehea_mem_nb); unregister_reboot_notifier(&ehea_reboot_nb); crash_shutdown_unregister(&ehea_crash_handler); out: @@ -3621,6 +3645,7 @@ static void __exit ehea_module_exit(void) ret = crash_shutdown_unregister(&ehea_crash_handler); if (ret) ehea_info("failed unregistering crash handler"); + unregister_memory_notifier(&ehea_mem_nb); kfree(ehea_fw_handles.arr); kfree(ehea_bcmc_regs.arr); ehea_destroy_busmap(); diff --git a/drivers/net/ehea/ehea_qmr.c b/drivers/net/ehea/ehea_qmr.c index d522e905f460..140f05baafd8 100644 --- a/drivers/net/ehea/ehea_qmr.c +++ b/drivers/net/ehea/ehea_qmr.c @@ -31,8 +31,8 @@ #include "ehea_phyp.h" #include "ehea_qmr.h" +struct ehea_bmap *ehea_bmap = NULL; -struct ehea_busmap ehea_bmap = { 0, 0, NULL }; static void *hw_qpageit_get_inc(struct hw_queue *queue) @@ -559,125 +559,253 @@ int ehea_destroy_qp(struct ehea_qp *qp) return 0; } -int ehea_create_busmap(void) +static inline int ehea_calc_index(unsigned long i, unsigned long s) { - u64 vaddr = EHEA_BUSMAP_START; - unsigned long high_section_index = 0; - int i; + return (i >> s) & EHEA_INDEX_MASK; +} - /* - * Sections are not in ascending order -> Loop over all sections and - * find the highest PFN to compute the required map size. - */ - ehea_bmap.valid_sections = 0; +static inline int ehea_init_top_bmap(struct ehea_top_bmap *ehea_top_bmap, + int dir) +{ + if(!ehea_top_bmap->dir[dir]) { + ehea_top_bmap->dir[dir] = + kzalloc(sizeof(struct ehea_dir_bmap), GFP_KERNEL); + if (!ehea_top_bmap->dir[dir]) + return -ENOMEM; + } + return 0; +} - for (i = 0; i < NR_MEM_SECTIONS; i++) - if (valid_section_nr(i)) - high_section_index = i; +static inline int ehea_init_bmap(struct ehea_bmap *ehea_bmap, int top, int dir) +{ + if(!ehea_bmap->top[top]) { + ehea_bmap->top[top] = + kzalloc(sizeof(struct ehea_top_bmap), GFP_KERNEL); + if (!ehea_bmap->top[top]) + return -ENOMEM; + } + return ehea_init_top_bmap(ehea_bmap->top[top], dir); +} - ehea_bmap.entries = high_section_index + 1; - ehea_bmap.vaddr = vmalloc(ehea_bmap.entries * sizeof(*ehea_bmap.vaddr)); +static int ehea_create_busmap_callback(unsigned long pfn, + unsigned long nr_pages, void *arg) +{ + unsigned long i, mr_len, start_section, end_section; + start_section = (pfn * PAGE_SIZE) / EHEA_SECTSIZE; + end_section = start_section + ((nr_pages * PAGE_SIZE) / EHEA_SECTSIZE); + mr_len = *(unsigned long *)arg; - if (!ehea_bmap.vaddr) + ehea_bmap = kzalloc(sizeof(struct ehea_bmap), GFP_KERNEL); + if (!ehea_bmap) return -ENOMEM; - for (i = 0 ; i < ehea_bmap.entries; i++) { - unsigned long pfn = section_nr_to_pfn(i); + for (i = start_section; i < end_section; i++) { + int ret; + int top, dir, idx; + u64 vaddr; + + top = ehea_calc_index(i, EHEA_TOP_INDEX_SHIFT); + dir = ehea_calc_index(i, EHEA_DIR_INDEX_SHIFT); + + ret = ehea_init_bmap(ehea_bmap, top, dir); + if(ret) + return ret; - if (pfn_valid(pfn)) { - ehea_bmap.vaddr[i] = vaddr; - vaddr += EHEA_SECTSIZE; - ehea_bmap.valid_sections++; - } else - ehea_bmap.vaddr[i] = 0; + idx = i & EHEA_INDEX_MASK; + vaddr = EHEA_BUSMAP_START + mr_len + i * EHEA_SECTSIZE; + + ehea_bmap->top[top]->dir[dir]->ent[idx] = vaddr; } + mr_len += nr_pages * PAGE_SIZE; + *(unsigned long *)arg = mr_len; + return 0; } +static unsigned long ehea_mr_len; + +static DEFINE_MUTEX(ehea_busmap_mutex); + +int ehea_create_busmap(void) +{ + int ret; + mutex_lock(&ehea_busmap_mutex); + ehea_mr_len = 0; + ret = walk_memory_resource(0, 1ULL << MAX_PHYSMEM_BITS, &ehea_mr_len, + ehea_create_busmap_callback); + mutex_unlock(&ehea_busmap_mutex); + return ret; +} + void ehea_destroy_busmap(void) { - vfree(ehea_bmap.vaddr); + int top, dir; + mutex_lock(&ehea_busmap_mutex); + if (!ehea_bmap) + goto out_destroy; + + for (top = 0; top < EHEA_MAP_ENTRIES; top++) { + if (!ehea_bmap->top[top]) + continue; + + for (dir = 0; dir < EHEA_MAP_ENTRIES; dir++) { + if (!ehea_bmap->top[top]->dir[dir]) + continue; + + kfree(ehea_bmap->top[top]->dir[dir]); + } + + kfree(ehea_bmap->top[top]); + } + + kfree(ehea_bmap); + ehea_bmap = NULL; +out_destroy: + mutex_unlock(&ehea_busmap_mutex); } u64 ehea_map_vaddr(void *caddr) { - u64 mapped_addr; - unsigned long index = __pa(caddr) >> SECTION_SIZE_BITS; - - if (likely(index < ehea_bmap.entries)) { - mapped_addr = ehea_bmap.vaddr[index]; - if (likely(mapped_addr)) - mapped_addr |= (((unsigned long)caddr) - & (EHEA_SECTSIZE - 1)); - else - mapped_addr = -1; - } else - mapped_addr = -1; - - if (unlikely(mapped_addr == -1)) - if (!test_and_set_bit(__EHEA_STOP_XFER, &ehea_driver_flags)) - schedule_work(&ehea_rereg_mr_task); - - return mapped_addr; + int top, dir, idx; + unsigned long index, offset; + + if (!ehea_bmap) + return EHEA_INVAL_ADDR; + + index = virt_to_abs(caddr) >> SECTION_SIZE_BITS; + top = (index >> EHEA_TOP_INDEX_SHIFT) & EHEA_INDEX_MASK; + if (!ehea_bmap->top[top]) + return EHEA_INVAL_ADDR; + + dir = (index >> EHEA_DIR_INDEX_SHIFT) & EHEA_INDEX_MASK; + if (!ehea_bmap->top[top]->dir[dir]) + return EHEA_INVAL_ADDR; + + idx = index & EHEA_INDEX_MASK; + if (!ehea_bmap->top[top]->dir[dir]->ent[idx]) + return EHEA_INVAL_ADDR; + + offset = (unsigned long)caddr & (EHEA_SECTSIZE - 1); + return ehea_bmap->top[top]->dir[dir]->ent[idx] | offset; +} + +static inline void *ehea_calc_sectbase(int top, int dir, int idx) +{ + unsigned long ret = idx; + ret |= dir << EHEA_DIR_INDEX_SHIFT; + ret |= top << EHEA_TOP_INDEX_SHIFT; + return abs_to_virt(ret << SECTION_SIZE_BITS); +} + +static u64 ehea_reg_mr_section(int top, int dir, int idx, u64 *pt, + struct ehea_adapter *adapter, + struct ehea_mr *mr) +{ + void *pg; + u64 j, m, hret; + unsigned long k = 0; + u64 pt_abs = virt_to_abs(pt); + + void *sectbase = ehea_calc_sectbase(top, dir, idx); + + for (j = 0; j < (EHEA_PAGES_PER_SECTION / EHEA_MAX_RPAGE); j++) { + + for (m = 0; m < EHEA_MAX_RPAGE; m++) { + pg = sectbase + ((k++) * EHEA_PAGESIZE); + pt[m] = virt_to_abs(pg); + } + hret = ehea_h_register_rpage_mr(adapter->handle, mr->handle, 0, + 0, pt_abs, EHEA_MAX_RPAGE); + + if ((hret != H_SUCCESS) + && (hret != H_PAGE_REGISTERED)) { + ehea_h_free_resource(adapter->handle, mr->handle, + FORCE_FREE); + ehea_error("register_rpage_mr failed"); + return hret; + } + } + return hret; +} + +static u64 ehea_reg_mr_sections(int top, int dir, u64 *pt, + struct ehea_adapter *adapter, + struct ehea_mr *mr) +{ + u64 hret = H_SUCCESS; + int idx; + + for (idx = 0; idx < EHEA_MAP_ENTRIES; idx++) { + if (!ehea_bmap->top[top]->dir[dir]->ent[idx]) + continue; + + hret = ehea_reg_mr_section(top, dir, idx, pt, adapter, mr); + if ((hret != H_SUCCESS) && (hret != H_PAGE_REGISTERED)) + return hret; + } + return hret; +} + +static u64 ehea_reg_mr_dir_sections(int top, u64 *pt, + struct ehea_adapter *adapter, + struct ehea_mr *mr) +{ + u64 hret = H_SUCCESS; + int dir; + + for (dir = 0; dir < EHEA_MAP_ENTRIES; dir++) { + if (!ehea_bmap->top[top]->dir[dir]) + continue; + + hret = ehea_reg_mr_sections(top, dir, pt, adapter, mr); + if ((hret != H_SUCCESS) && (hret != H_PAGE_REGISTERED)) + return hret; + } + return hret; } int ehea_reg_kernel_mr(struct ehea_adapter *adapter, struct ehea_mr *mr) { int ret; u64 *pt; - void *pg; - u64 hret, pt_abs, i, j, m, mr_len; + u64 hret; u32 acc_ctrl = EHEA_MR_ACC_CTRL; - mr_len = ehea_bmap.valid_sections * EHEA_SECTSIZE; + unsigned long top; - pt = kzalloc(PAGE_SIZE, GFP_KERNEL); + pt = kzalloc(PAGE_SIZE, GFP_KERNEL); if (!pt) { ehea_error("no mem"); ret = -ENOMEM; goto out; } - pt_abs = virt_to_abs(pt); - hret = ehea_h_alloc_resource_mr(adapter->handle, - EHEA_BUSMAP_START, mr_len, - acc_ctrl, adapter->pd, + hret = ehea_h_alloc_resource_mr(adapter->handle, EHEA_BUSMAP_START, + ehea_mr_len, acc_ctrl, adapter->pd, &mr->handle, &mr->lkey); + if (hret != H_SUCCESS) { ehea_error("alloc_resource_mr failed"); ret = -EIO; goto out; } - for (i = 0 ; i < ehea_bmap.entries; i++) - if (ehea_bmap.vaddr[i]) { - void *sectbase = __va(i << SECTION_SIZE_BITS); - unsigned long k = 0; - - for (j = 0; j < (EHEA_PAGES_PER_SECTION / - EHEA_MAX_RPAGE); j++) { - - for (m = 0; m < EHEA_MAX_RPAGE; m++) { - pg = sectbase + ((k++) * EHEA_PAGESIZE); - pt[m] = virt_to_abs(pg); - } - - hret = ehea_h_register_rpage_mr(adapter->handle, - mr->handle, - 0, 0, pt_abs, - EHEA_MAX_RPAGE); - if ((hret != H_SUCCESS) - && (hret != H_PAGE_REGISTERED)) { - ehea_h_free_resource(adapter->handle, - mr->handle, - FORCE_FREE); - ehea_error("register_rpage_mr failed"); - ret = -EIO; - goto out; - } - } - } + if (!ehea_bmap) { + ehea_h_free_resource(adapter->handle, mr->handle, FORCE_FREE); + ehea_error("no busmap available"); + ret = -EIO; + goto out; + } + + for (top = 0; top < EHEA_MAP_ENTRIES; top++) { + if (!ehea_bmap->top[top]) + continue; + + hret = ehea_reg_mr_dir_sections(top, pt, adapter, mr); + if((hret != H_PAGE_REGISTERED) && (hret != H_SUCCESS)) + break; + } if (hret != H_SUCCESS) { ehea_h_free_resource(adapter->handle, mr->handle, FORCE_FREE); -- cgit v1.2.3 From 7fb1c2ac8ecaf0883f2fcb38dfc9ec2d15cee11d Mon Sep 17 00:00:00 2001 From: Brian King Date: Wed, 14 May 2008 09:48:25 -0500 Subject: ehea: Fix use after free on reboot Fixes the following use after free oops: ehea: Reboot: freeing all eHEA resources Unable to handle kernel paging request for data at address 0x6b6b6b6b6b6b6c5b Faulting instruction address: 0xd000000000354488 cpu 0x0: Vector: 300 (Data Access) at [c00000002ec6f310] pc: d000000000354488: .ehea_shutdown_single_port+0x50/0x78 [ehea] lr: d00000000035447c: .ehea_shutdown_single_port+0x44/0x78 [ehea] sp: c00000002ec6f590 msr: 8000000000009032 dar: 6b6b6b6b6b6b6c5b dsisr: 40000000 current = 0xc0000000281412e0 paca = 0xc0000000006df300 pid = 10930, comm = reboot enter ? for help [c00000002ec6f590] d00000000035d64c .ehea_remove+0x44/0x124 [ehea] (unreliable) [c00000002ec6f630] c000000000319f88 .of_platform_device_remove+0x40/0x58 [c00000002ec6f6a0] c000000000291018 .__device_release_driver+0xb0/0xf0 [c00000002ec6f730] c000000000291120 .driver_detach+0xc8/0xfc [c00000002ec6f7c0] c00000000028fe24 .bus_remove_driver+0xb4/0x114 [c00000002ec6f850] c000000000291768 .driver_unregister+0x54/0x74 [c00000002ec6f8e0] c00000000031a0c8 .of_unregister_driver+0x14/0x28 [c00000002ec6f950] c000000000023ba0 .ibmebus_unregister_driver+0x10/0x24 [c00000002ec6f9c0] d000000000354180 .ehea_reboot_notifier+0x30/0x4c [ehea] [c00000002ec6fa40] c0000000003c95a8 .notifier_call_chain+0x5c/0xcc [c00000002ec6fae0] c000000000082cd4 .__blocking_notifier_call_chain+0x70/0xb0 [c00000002ec6fb90] c000000000075cf8 .kernel_restart_prepare+0x24/0x58 [c00000002ec6fc10] c000000000075f0c .kernel_restart+0x20/0x6c [c00000002ec6fc90] c000000000078674 .sys_reboot+0x1d4/0x290 [c00000002ec6fe30] c0000000000086ac syscall_exit+0x0/0x40 Signed-off-by: Jeff Garzik --- drivers/net/ehea/ehea_main.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'drivers/net/ehea') diff --git a/drivers/net/ehea/ehea_main.c b/drivers/net/ehea/ehea_main.c index d1b6d4e7495d..8645224da1cf 100644 --- a/drivers/net/ehea/ehea_main.c +++ b/drivers/net/ehea/ehea_main.c @@ -3178,11 +3178,12 @@ out_err: static void ehea_shutdown_single_port(struct ehea_port *port) { + struct ehea_adapter *adapter = port->adapter; unregister_netdev(port->netdev); ehea_unregister_port(port); kfree(port->mc_list); free_netdev(port->netdev); - port->adapter->active_ports--; + adapter->active_ports--; } static int ehea_setup_ports(struct ehea_adapter *adapter) -- cgit v1.2.3 From 25c16fffa8ed82d3ef31980d76ff95d3c6430f00 Mon Sep 17 00:00:00 2001 From: Joe Perches Date: Mon, 12 May 2008 14:38:17 -0700 Subject: drivers/net/ehea - remove unnecessary memset after kzalloc Signed-off-by: Joe Perches Signed-off-by: Jeff Garzik --- drivers/net/ehea/ehea_main.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'drivers/net/ehea') diff --git a/drivers/net/ehea/ehea_main.c b/drivers/net/ehea/ehea_main.c index 8645224da1cf..287a61918739 100644 --- a/drivers/net/ehea/ehea_main.c +++ b/drivers/net/ehea/ehea_main.c @@ -2213,8 +2213,6 @@ static void ehea_vlan_rx_register(struct net_device *dev, goto out; } - memset(cb1->vlan_filter, 0, sizeof(cb1->vlan_filter)); - hret = ehea_h_modify_ehea_port(adapter->handle, port->logical_port_id, H_PORT_CB1, H_PORT_CB1_ALL, cb1); if (hret != H_SUCCESS) -- cgit v1.2.3 From 00aaea2f95d73d4e2b5e45cf77c3cbb16c59e87f Mon Sep 17 00:00:00 2001 From: Jan-Bernd Themann Date: Mon, 9 Jun 2008 15:17:37 +0100 Subject: ehea: set mac address fix eHEA has to call firmware functions in order to change the mac address of a logical port. This patch checks if the logical port is up when calling the register / deregister mac address calls. If the port is down these firmware calls would fail and are therefore not executed. Signed-off-by: Jan-Bernd Themann Signed-off-by: Jeff Garzik --- drivers/net/ehea/ehea_main.c | 16 ++++++++++------ 1 file changed, 10 insertions(+), 6 deletions(-) (limited to 'drivers/net/ehea') diff --git a/drivers/net/ehea/ehea_main.c b/drivers/net/ehea/ehea_main.c index 287a61918739..faae01dc1c4b 100644 --- a/drivers/net/ehea/ehea_main.c +++ b/drivers/net/ehea/ehea_main.c @@ -1766,16 +1766,20 @@ static int ehea_set_mac_addr(struct net_device *dev, void *sa) mutex_lock(&ehea_bcmc_regs.lock); /* Deregister old MAC in pHYP */ - ret = ehea_broadcast_reg_helper(port, H_DEREG_BCMC); - if (ret) - goto out_upregs; + if (port->state == EHEA_PORT_UP) { + ret = ehea_broadcast_reg_helper(port, H_DEREG_BCMC); + if (ret) + goto out_upregs; + } port->mac_addr = cb0->port_mac_addr << 16; /* Register new MAC in pHYP */ - ret = ehea_broadcast_reg_helper(port, H_REG_BCMC); - if (ret) - goto out_upregs; + if (port->state == EHEA_PORT_UP) { + ret = ehea_broadcast_reg_helper(port, H_REG_BCMC); + if (ret) + goto out_upregs; + } ret = 0; -- cgit v1.2.3 From 4bb073c0e32a0862bdb5215d11af19f6c0180c98 Mon Sep 17 00:00:00 2001 From: "David S. Miller" Date: Thu, 12 Jun 2008 02:22:02 -0700 Subject: net: Eliminate flush_scheduled_work() calls while RTNL is held. If the RTNL is held when we invoke flush_scheduled_work() we could deadlock. One such case is linkwatch, it is a work struct which tries to grab the RTNL semaphore. The most common case are net driver ->stop() methods. The simplest conversion is to instead use cancel_{delayed_}work_sync() explicitly on the various work struct the driver uses. This is an OK transformation because these work structs are doing things like resetting the chip, restarting link negotiation, and so forth. And if we're bringing down the device, we're about to turn the chip off and reset it anways. So if we cancel a pending work event, that's fine here. Some drivers were working around this deadlock by using a msleep() polling loop of some sort, and those cases are converted to instead use cancel_{delayed_}work_sync() as well. Signed-off-by: David S. Miller --- drivers/net/ehea/ehea_main.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'drivers/net/ehea') diff --git a/drivers/net/ehea/ehea_main.c b/drivers/net/ehea/ehea_main.c index faae01dc1c4b..075fd547421e 100644 --- a/drivers/net/ehea/ehea_main.c +++ b/drivers/net/ehea/ehea_main.c @@ -2605,7 +2605,8 @@ static int ehea_stop(struct net_device *dev) if (netif_msg_ifdown(port)) ehea_info("disabling port %s", dev->name); - flush_scheduled_work(); + cancel_work_sync(&port->reset_task); + mutex_lock(&port->port_lock); netif_stop_queue(dev); port_napi_disable(port); -- cgit v1.2.3