Skip to contentSkip to conversation

Fix kdump failures with crashkernel=high

PATCH Superseded Joerg Roedel 4 participants 28 messages Jan 6, 2015 – Jun 2, 2015

Subsystems (4)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

v2 cover Fix kdump failures with crashkernel=high

v1->v2:

* Updated comments based on feedback from Konrad
* Added Acked-bys
* Rebased to v3.19-rc3

Hi,

here is a patch-set to fix failed kdump kernel boots when the systems was booted with crashkernel=X,high. On those systems the kernel allocates only 72MiB of low-memory for DMA buffers, which showed to be too low on some systems.

The problem is that 64MiB of the low-memory is allocated by swiotlb, leaving 8MB for the page-allocator. But swiotlb tries to allocate DMA memory from the page-allocator first, which fails pretty fast in the boot sequence, causing warnings. This patch-set removes these warnings.

But even the 64MiB for swiotlb are eaten up on some systems, so that the default of low-memory allocated for the crash-kernel is increase from 72MB to 256MB (only changing the defaults, can still be overwritten by crashkernel=X,low).

This number comes from experiments on the affected systems, 128MiB low-memory was still not enough there, thus I set the value to 256MiB to fix the issues.

Any feedback appreciated.

Thanks,

Joerg

Joerg Roedel (3):
swiotlb: Warn on allocation failure in swiotlb_alloc_coherent x86, swiotlb: Try coherent allocations with __GFP_NOWARN x86, crash: Allocate enough low-mem when crashkernel=high

 arch/x86/kernel/pci-swiotlb.c |  8 ++++++++
 arch/x86/kernel/setup.c       |  5 ++++-
 lib/swiotlb.c                 | 11 +++++++++--
 3 files changed, 21 insertions(+), 3 deletions(-)
Signature · 14 lines
1.9.1

Joerg Roedel (3):
  swiotlb: Warn on allocation failure in swiotlb_alloc_coherent
  x86, swiotlb: Try coherent allocations with __GFP_NOWARN
  x86, crash: Allocate enough low-mem when crashkernel=high

 arch/x86/kernel/pci-swiotlb.c |  8 ++++++++
 arch/x86/kernel/setup.c       |  5 ++++-
 lib/swiotlb.c                 | 11 +++++++++--
 3 files changed, 21 insertions(+), 3 deletions(-)

-- 
1.9.1
#1 lore raw

PATCH v2 1/3 swiotlb: Warn on allocation failure in swiotlb_alloc_coherent

From: Joerg Roedel <[email protected]>

Print a warning when all allocation tries have been failed and the function is about to return NULL. This prepares for calling the function with __GFP_NOWARN to suppress allocation failure warnings before all fall-backs have failed.

---
 lib/swiotlb.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)
lib/swiotlb.c +9 -2
index 4abda07..e0e9212 100644
@@ -655,7 +655,7 @@ swiotlb_alloc_coherent(struct device *hwdev, size_t size,
 		 */
 		phys_addr_t paddr = map_single(hwdev, 0, size, DMA_FROM_DEVICE);
 		if (paddr == SWIOTLB_MAP_ERROR)
-			return NULL;
+			goto err_warn;
 
 		ret = phys_to_virt(paddr);
 		dev_addr = phys_to_dma(hwdev, paddr);
@@ -669,7 +669,7 @@ swiotlb_alloc_coherent(struct device *hwdev, size_t size,
 			/* DMA_TO_DEVICE to avoid memcpy in unmap_single */
 			swiotlb_tbl_unmap_single(hwdev, paddr,
 						 size, DMA_TO_DEVICE);
-			return NULL;
+			goto err_warn;
 		}
 	}
 
@@ -677,6 +677,13 @@ swiotlb_alloc_coherent(struct device *hwdev, size_t size,
 	memset(ret, 0, size);
 
 	return ret;
+
+err_warn:
+	pr_warn("swiotlb: coherent allocation failed for device %s size=%zu\n",
+		dev_name(hwdev), size);
+	dump_stack();
+
+	return NULL;
 }
 EXPORT_SYMBOL(swiotlb_alloc_coherent);
 
Signature · 1 line
1.9.1
maintainer
#2 lore raw

Re: [PATCH 1/3] swiotlb: Warn on allocation failure in swiotlb_alloc_coherent

Quoted 11 lines · Joerg Roedel

From: Joerg Roedel <[email protected]>

Print a warning when all allocation tries have been failed and the function is about to return NULL. This prepares for calling the function with __GFP_NOWARN to suppress allocation failure warnings before all fall-backs have failed.

---

...

lib/swiotlb.c swiotlb_alloc_coherent(struct device *hwdev, size_t size, · line 684

Quoted 9 lines
lib/swiotlb.c +5 -0
@@ -677,6 +677,13 @@ swiotlb_alloc_coherent(struct device *hwdev, size_t size,
 	memset(ret, 0, size);
 
 	return ret;
+
+err_warn:
+	pr_warn("swiotlb: coherent allocation failed for device %s size=%zu\n",
+		dev_name(hwdev), size);
+	dump_stack();

Are we really sure we want to be that noisy about it? What happens if that fails, we can't do DMA anymore or should we free some precious DMA memory, as a compromise?

Signature · 5 lines
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.
--
#3 lore raw

Re: [PATCH 1/3] swiotlb: Warn on allocation failure in swiotlb_alloc_coherent

Quoted 10 lines · Borislav Petkov

lib/swiotlb.c swiotlb_alloc_coherent(struct device *hwdev, size_t size, · line 684

Quoted 5 lines · Joerg Roedel
lib/swiotlb.c +5 -0
+
+err_warn:
+	pr_warn("swiotlb: coherent allocation failed for device %s size=%zu\n",
+		dev_name(hwdev), size);
+	dump_stack();

Are we really sure we want to be that noisy about it? What happens if that fails, we can't do DMA anymore or should we free some precious DMA memory, as a compromise?

Hmm, I don't think there is a way to request drivers to free DMA memory. What this patch does is to keep the same noisiness as before, just not from the page allocator but from the very end of the DMA allocation process, when everything failed.
We can of course discuss whether this is too noisy and put some kind of ratelimit around.

Joerg

#4 lore raw

PATCH v2 2/3 x86, swiotlb: Try coherent allocations with __GFP_NOWARN

In reply to Joerg Roedel

From: Joerg Roedel <[email protected]>

When we boot a kdump kernel in high memory, there is by default only 72MB of low memory available. The swiotlb code takes 64MB of it (by default) so that there are only 8MB left to allocate from. On systems with many devices this causes page allocator warnings from dma_generic_alloc_coherent():

systemd-udevd: page allocation failure: order:0, mode:0x280d4

CPU: 0 PID: 197 Comm: systemd-udevd Tainted: G        W       3.12.28-4-default #1
Hardware name: HP ProLiant DL980 G7, BIOS P66 07/30/2012
 ffff8800781335e0 ffffffff8150b1db 00000000000280d4 ffffffff8113af90
 0000000000000000 0000000000000000 ffff88007efdbb00 0000000100000000
 0000000000000000 0000000000000000 0000000000000000 0000000000000001
Call Trace:
 [<ffffffff8100467d>] dump_trace+0x7d/0x2d0
 [<ffffffff81004964>] show_stack_log_lvl+0x94/0x170
 [<ffffffff81005d91>] show_stack+0x21/0x50
 [<ffffffff8150b1db>] dump_stack+0x41/0x51
 [<ffffffff8113af90>] warn_alloc_failed+0xf0/0x160
 [<ffffffff8150763a>] __alloc_pages_slowpath+0x72f/0x796
 [<ffffffff8113ee7a>] __alloc_pages_nodemask+0x1ea/0x210
 [<ffffffff81008256>] dma_generic_alloc_coherent+0x96/0x140
 [<ffffffff8103fccc>] x86_swiotlb_alloc_coherent+0x1c/0x50
 [<ffffffffa048ae5b>] ttm_dma_pool_alloc_new_pages+0xab/0x320 [ttm]
 [<ffffffffa048bc6e>] ttm_dma_populate+0x3ce/0x640 [ttm]
 [<ffffffffa0486>] ttm_tt_bind+0x36/0x60 [ttm]
 [<ffffffffa0484faf>] ttm_bo_handle_move_mem+0x55f/0x5c0 [ttm]
 [<ffffffffa0485be5>] ttm_bo_move_buffer+0x105/0x130 [ttm]
 [<ffffffffa0485cd1>] ttm_bo_validate+0xc1/0x130 [ttm]
 [<ffffffffa0485f8b>] ttm_bo_init+0x24b/0x400 [ttm]
 [<ffffffffa054f8bc>] radeon_bo_create+0x16c/0x200 [radeon]
 [<ffffffffa0563c8e>] radeon_ring_init+0x11e/0x2b0 [radeon]
 [<ffffffffa056c143>] r100_cp_init+0x123/0x5b0 [radeon]
 [<ffffffffa056e8e4>] r100_startup+0x194/0x230 [radeon]
 [<ffffffffa056ece3>] r100_init+0x223/0x410 [radeon]
 [<ffffffffa053495f>] radeon_device_init+0x6af/0x830 [radeon]
 [<ffffffffa0536979>] radeon_driver_load_kms+0x89/0x180 [radeon]
 [<ffffffffa04eeb31>] drm_get_pci_dev+0x121/0x2f0 [drm]
 [<ffffffff812d3ec9>ocal_pci_probe+0x39/0x60
 [<ffffffff812d51e9>] pci_device_probe+0xa9/0x120
 [<ffffffff8139871d>] driver_probe_device+0x9d/0x3d0
 [<ffffffff81398b1b>] __driver_attach+0x8b/0x90
 [<ffffffff8139667b>] bus_for_each_dev+0x5b/0x90
 [<ffffffff81397cd8>] bus_add_driver+0x1f8/0x2c0
 [<ffffffff8139911b>] driver_register+0x5b/0xe0
 [<ffffffff810002c2>] do_one_initcall+0xf2/0x1a0
 [<ffffffff810c8837>] load_module+0x1207/0x1c70
 [<ffffffff810c93f5>] SYSC_finit_module+0x75/0xa0
 [<ffffffff81519329>] system_call_fastpath+0x16/0x1b
 [<00007fac533d2789>] 0x7fac533d2788

After these warnings the code enters a fall-back path and allocated directly from the swiotlb aperture in the end. So remove these warnings as this is not a fatal error.

---
 arch/x86/kernel/pci-swiotlb.c | 8 ++++++++
 1 file changed, 8 insertions(+)
arch/x86/kernel/pci-swiotlb.c +8 -0
index 77dd0ad..79b2291 100644
@@ -20,6 +20,14 @@ void *x86_swiotlb_alloc_coherent(struct device *hwdev, size_t size,
 {
 	void *vaddr;
 
+	/*
+	 * When booting a kdump kernel in high memory these allocations are very
+	 * likely to fail, as there are by default only 8MB of low memory to
+	 * allocate from. So disable the warnings from the allocator when this
+	 * happens.  SWIOTLB also implements fall-backs for failed allocations.
+	 */
+	flags |= __GFP_NOWARN;
+
 	vaddr = dma_generic_alloc_coherent(hwdev, size, dma_handle, flags,
 					   attrs);
 	if (vaddr)
Signature · 1 line
1.9.1
maintainer
#5 lore raw

Re: [PATCH 2/3] x86, swiotlb: Try coherent allocations with __GFP_NOWARN

Quoted 43 lines · Joerg Roedel

From: Joerg Roedel <[email protected]>

When we boot a kdump kernel in high memory, there is by default only 72MB of low memory available. The swiotlb code takes 64MB of it (by default) so that there are only 8MB left to allocate from. On systems with many devices this causes page allocator warnings from dma_generic_alloc_coherent():

systemd-udevd: page allocation failure: order:0, mode:0x280d4

CPU: 0 PID: 197 Comm: systemd-udevd Tainted: G        W       3.12.28-4-default #1
Hardware name: HP ProLiant DL980 G7, BIOS P66 07/30/2012
 ffff8800781335e0 ffffffff8150b1db 00000000000280d4 ffffffff8113af90
 0000000000000000 0000000000000000 ffff88007efdbb00 0000000100000000
 0000000000000000 0000000000000000 0000000000000000 0000000000000001
Call Trace:
 [<ffffffff8100467d>] dump_trace+0x7d/0x2d0
 [<ffffffff81004964>] show_stack_log_lvl+0x94/0x170
 [<ffffffff81005d91>] show_stack+0x21/0x50
 [<ffffffff8150b1db>] dump_stack+0x41/0x51
 [<ffffffff8113af90>] warn_alloc_failed+0xf0/0x160
 [<ffffffff8150763a>] __alloc_pages_slowpath+0x72f/0x796
 [<ffffffff8113ee7a>] __alloc_pages_nodemask+0x1ea/0x210
 [<ffffffff81008256>] dma_generic_alloc_coherent+0x96/0x140
 [<ffffffff8103fccc>] x86_swiotlb_alloc_coherent+0x1c/0x50
 [<ffffffffa048ae5b>] ttm_dma_pool_alloc_new_pages+0xab/0x320 [ttm]
 [<ffffffffa048bc6e>] ttm_dma_populate+0x3ce/0x640 [ttm]
 [<ffffffffa0486>] ttm_tt_bind+0x36/0x60 [ttm]
 [<ffffffffa0484faf>] ttm_bo_handle_move_mem+0x55f/0x5c0 [ttm]
 [<ffffffffa0485be5>] ttm_bo_move_buffer+0x105/0x130 [ttm]
 [<ffffffffa0485cd1>] ttm_bo_validate+0xc1/0x130 [ttm]
 [<ffffffffa0485f8b>] ttm_bo_init+0x24b/0x400 [ttm]
 [<ffffffffa054f8bc>] radeon_bo_create+0x16c/0x200 [radeon]
 [<ffffffffa0563c8e>] radeon_ring_init+0x11e/0x2b0 [radeon]
 [<ffffffffa056c143>] r100_cp_init+0x123/0x5b0 [radeon]
 [<ffffffffa056e8e4>] r100_startup+0x194/0x230 [radeon]
 [<ffffffffa056ece3>] r100_init+0x223/0x410 [radeon]
 [<ffffffffa053495f>] radeon_device_init+0x6af/0x830 [radeon]
 [<ffffffffa0536979>] radeon_driver_load_kms+0x89/0x180 [radeon]
 [<ffffffffa04eeb31>] drm_get_pci_dev+0x121/0x2f0 [drm]
 [<ffffffff812d3ec9>ocal_pci_probe+0x39/0x60
 [<ffffffff812d51e9>] pci_device_probe+0xa9/0x120
 [<ffffffff8139871d>] driver_probe_device+0x9d/0x3d0
 [<ffffffff81398b1b>] __driver_attach+0x8b/0x90
Quoted 8 lines
 [<ffffffff8139667b>] bus_for_each_dev+0x5b/0x90
 [<ffffffff81397cd8>] bus_add_driver+0x1f8/0x2c0
 [<ffffffff8139911b>] driver_register+0x5b/0xe0
 [<ffffffff810002c2>] do_one_initcall+0xf2/0x1a0
 [<ffffffff810c8837>] load_module+0x1207/0x1c70
 [<ffffffff810c93f5>] SYSC_finit_module+0x75/0xa0
 [<ffffffff81519329>] system_call_fastpath+0x16/0x1b
 [<00007fac533d2789>] 0x7fac533d2788

Please remove those addresses and offsets and leave only the function names in the call trace - the rest is useless clutter only as those addresses don't mean anything on else but this kernel.

Quoted 14 lines

After these warnings the code enters a fall-back path and allocated directly from the swiotlb aperture in the end. So remove these warnings as this is not a fatal error.

---
 arch/x86/kernel/pci-swiotlb.c | 8 ++++++++
 1 file changed, 8 insertions(+)
arch/x86/kernel/pci-swiotlb.c +0 -0
index 77dd0ad..79b2291 100644

arch/x86/kernel/pci-swiotlb.c void *x86_swiotlb_alloc_coherent(struct device *hwdev, size_t size, · line 29

Quoted 10 lines
arch/x86/kernel/pci-swiotlb.c +7 -0
@@ -20,6 +20,14 @@ void *x86_swiotlb_alloc_coherent(struct device *hwdev, size_t size,
 {
 	void *vaddr;
 
+	/*
+	 * When booting a kdump kernel in high memory these allocations are very
+	 * likely to fail, as there are by default only 8MB of low memory to
+	 * allocate from. So disable the warnings from the allocator when this
+	 * happens.  SWIOTLB also implements fall-backs for failed allocations.
+	 */
+	flags |= __GFP_NOWARN;

Ok, so this practically does all allocations __GFP_NOWARN now. Shouldn't you be doing this before swiotlb_alloc_coherent() and not before dma_generic_alloc_coherent()?

Quoted 3 lines
arch/x86/kernel/pci-swiotlb.c +0 -0
 	vaddr = dma_generic_alloc_coherent(hwdev, size, dma_handle, flags,
 					   attrs);
 	if (vaddr)
Signature · 5 lines
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.
--
#6 lore raw

Re: [PATCH 2/3] x86, swiotlb: Try coherent allocations with __GFP_NOWARN

Quoted 11 lines · Borislav Petkov
Quoted 10 lines · Joerg Roedel

From: Joerg Roedel <[email protected]>

arch/x86/kernel/pci-swiotlb.c +0 -0
index 77dd0ad..79b2291 100644
@@ -20,6 +20,14 @@ void *x86_swiotlb_alloc_coherent(struct device *hwdev, size_t size,
 {
 	void *vaddr;
 
Quoted 11 lines

arch/x86/kernel/pci-swiotlb.c void *x86_swiotlb_alloc_coherent(struct device *hwdev, size_t size, · line 29

Quoted 7 lines
arch/x86/kernel/pci-swiotlb.c +7 -0
+	/*
+	 * When booting a kdump kernel in high memory these allocations are very
+	 * likely to fail, as there are by default only 8MB of low memory to
+	 * allocate from. So disable the warnings from the allocator when this
+	 * happens.  SWIOTLB also implements fall-backs for failed allocations.
+	 */
+	flags |= __GFP_NOWARN;

Ok, so this practically does all allocations __GFP_NOWARN now. Shouldn't you be doing this before swiotlb_alloc_coherent() and not before dma_generic_alloc_coherent()?

I think this patch mainly suppress warning from buddy allocation failure because it tried buddy allocation several times before the final try of bounce buffer allocation. Buddy allocation failure will call dump_stack.

Quoted 16 lines
Quoted 4 lines
arch/x86/kernel/pci-swiotlb.c +0 -0
 	vaddr = dma_generic_alloc_coherent(hwdev, size, dma_handle, flags,
 					   attrs);
 	if (vaddr)
Signature · 10 lines
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.
--
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
#7 lore raw

Re: [PATCH 2/3] x86, swiotlb: Try coherent allocations with __GFP_NOWARN

Quoted 8 lines · Baoquan He
Quoted 3 lines

Ok, so this practically does all allocations __GFP_NOWARN now. Shouldn't you be doing this before swiotlb_alloc_coherent() and not before dma_generic_alloc_coherent()?

I think this patch mainly suppress warning from buddy allocation failure because it tried buddy allocation several times before the final try of bounce buffer allocation. Buddy allocation failure will call dump_stack.

Yes, exactly. The default low-memory available to the page-allocator with crashkernel=high is 8MB. This is up pretty fast and then we start to get warnings, even when there is still memory left in the swiotlb space. The __GFP_WARN is there to suppress the warnings from the page-allocator, so it has to be set before it is called.

Joerg

#8 lore raw

PATCH v2 3/3 x86, crash: Allocate enough low-mem when crashkernel=high

In reply to Joerg Roedel

From: Joerg Roedel <[email protected]>

When the crashkernel is loaded above 4GiB in memory the first kernel only allocates 72MiB of low-memory for the DMA requirements of the second kernel. On systems with many devices this is not enough and causes device driver initialization errors and failed crash dumps. Set this default value to 256MiB to make sure there is enough memory available for DMA.

---
 arch/x86/kernel/setup.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
arch/x86/kernel/setup.c +4 -1
index ab4734e..d6e6a6d 100644
@@ -536,8 +536,11 @@ static void __init reserve_crashkernel_low(void)
 		 *	swiotlb overflow buffer: now is hardcoded to 32k.
 		 *		We round it to 8M for other buffers that
 		 *		may need to stay low too.
+		 *		Also make sure we allocate enough extra memory
+		 *		low memory so that we don't run out of DMA
+		 *		buffers for 32bit devices.
 		 */
-		low_size = swiotlb_size_or_default() + (8UL<<20);
+		low_size = max(swiotlb_size_or_default() + (8UL<<20), 256UL<<20);
 		auto_set = true;
 	} else {
 		/* passed with crashkernel=0,low ? */
Signature · 1 line
1.9.1
#9 lore raw

Re: [PATCH 3/3] x86, crash: Allocate enough low-mem when crashkernel=high

Hi Joerg,

Yeah, it does happen if too many devices. I guess the reason no reports come to us on rhel is we always use a auto mechsanism. We always try to allocate from below 896M, if failed try below 4G, if failed try above 4G.

This could be solved in 2 ways:

1) We could optimize the distro shell scripts which build the initramfs for kdump. Only include those devices which are necessary in kdump kernel. E.g for net dump only that NIC which connect to dump target will be brought up, this can decrease the dma memory requirement.

2) increase low-mem when crashkernel=high. But we have to be careful to do this. We implement crashkernel=high not only for the unhappiness crashkernel reservation is limited below 4G, but dma/dma32 memory space is precious on some systems. If set crashkernel=high still too much low memory has to be reserved by default, it's important to find the balance. So if we have to increase the default low-mem, how much memory is enough, why 256M? why not 128M/192M/320M/384M? And if 256M works on your system, what if another person say it does't work because there are more devices on his system?

Anyway, I understand the requirement, but we need find out how much memory can satisfy most of systems.

Thanks
Baoquan

Quoted 40 lines · Joerg Roedel

From: Joerg Roedel <[email protected]>

When the crashkernel is loaded above 4GiB in memory the first kernel only allocates 72MiB of low-memory for the DMA requirements of the second kernel. On systems with many devices this is not enough and causes device driver initialization errors and failed crash dumps. Set this default value to 256MiB to make sure there is enough memory available for DMA.

---
 arch/x86/kernel/setup.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
arch/x86/kernel/setup.c +4 -1
index ab4734e..d6e6a6d 100644
@@ -536,8 +536,11 @@ static void __init reserve_crashkernel_low(void)
 		 *	swiotlb overflow buffer: now is hardcoded to 32k.
 		 *		We round it to 8M for other buffers that
 		 *		may need to stay low too.
+		 *		Also make sure we allocate enough extra memory
+		 *		low memory so that we don't run out of DMA
+		 *		buffers for 32bit devices.
 		 */
-		low_size = swiotlb_size_or_default() + (8UL<<20);
+		low_size = max(swiotlb_size_or_default() + (8UL<<20), 256UL<<20);
 		auto_set = true;
 	} else {
 		/* passed with crashkernel=0,low ? */
Signature · 7 lines
1.9.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
#10 lore raw

Re: [PATCH 3/3] x86, crash: Allocate enough low-mem when crashkernel=high

Hi Baoquan,

thanks for your reply.

Quoted 10 lines · Baoquan He

2) increase low-mem when crashkernel=high. But we have to be careful to do this. We implement crashkernel=high not only for the unhappiness crashkernel reservation is limited below 4G, but dma/dma32 memory space is precious on some systems. If set crashkernel=high still too much low memory has to be reserved by default, it's important to find the balance. So if we have to increase the default low-mem, how much memory is enough, why 256M? why not 128M/192M/320M/384M? And if 256M works on your system, what if another person say it does't work because there are more devices on his system?

Quoted 2 lines

Anyway, I understand the requirement, but we need find out how much memory can satisfy most of systems.

Yes, I totally agree that it is tough to find a good default here. I used 256MB because this is what was required on the system the failed kdumps were reported on.

But probably we can agree that 72MB is not enough (given that 64MB are taken away by swiotlb already), and increase it to a value we think by now is sufficient for most systems.

Btw, the issue was also reported on machines with only a few devices, the reason there is that device drivers allocate more dma memory by default on intilization. Maybe we should handle that as a driver regression in the future, forcing them to allocate more dma-memory on-demand and not on initialization.

Joerg

#11 lore raw

Re: [PATCH 3/3] x86, crash: Allocate enough low-mem when crashkernel=high

Quoted 18 lines · Joerg Roedel

Hi Baoquan,

thanks for your reply.

Quoted 10 lines · Baoquan He

2) increase low-mem when crashkernel=high. But we have to be careful to do this. We implement crashkernel=high not only for the unhappiness crashkernel reservation is limited below 4G, but dma/dma32 memory space is precious on some systems. If set crashkernel=high still too much low memory has to be reserved by default, it's important to find the balance. So if we have to increase the default low-mem, how much memory is enough, why 256M? why not 128M/192M/320M/384M? And if 256M works on your system, what if another person say it does't work because there are more devices on his system?

Quoted 2 lines

Anyway, I understand the requirement, but we need find out how much memory can satisfy most of systems.

Quoted 7 lines

Yes, I totally agree that it is tough to find a good default here. I used 256MB because this is what was required on the system the failed kdumps were reported on.

But probably we can agree that 72MB is not enough (given that 64MB are taken away by swiotlb already), and increase it to a value we think by now is sufficient for most systems.

Yeah, and I got report from user about this issue too. It should be fixed. Like I said, the 1st suggestion mainly will goes to the area of initramfs making tools, currently maybe dracut which is used widely. This may cause many changes. Hence increasing low mem is a better idea.

Before I said 256M may not be a good value, that's because in your patch cover you said this number comes from experiments on the affected systems, and 128M was still not enough, then you set it to 256M. This may be a little rush. I think the step size to increase should be 32M, after all previously people only take 64M and 8M, enlarge it on a step size of 128M only one time, it can't be seen as patient and careful. If it failed on 224M but succeed on 256M, then 256M may be not enough. I would like to say 32M is better, then we can make a good evaluate.

I will ask user reported this issue to help test and see what value will be satisfy their system.

Anyway, I think this patch is helpful and necessary.

Quoted 6 lines

Btw, the issue was also reported on machines with only a few devices, the reason there is that device drivers allocate more dma memory by default on intilization. Maybe we should handle that as a driver regression in the future, forcing them to allocate more dma-memory on-demand and not on initialization.

Yeah, agree. In that case it shoube be handled as a regression.

Thanks
Baoquan

#12 lore raw

Re: [PATCH 3/3] x86, crash: Allocate enough low-mem when crashkernel=high

Hi Baoquan,

Quoted 8 lines · Baoquan He

Before I said 256M may not be a good value, that's because in your patch cover you said this number comes from experiments on the affected systems, and 128M was still not enough, then you set it to 256M. This may be a little rush. I think the step size to increase should be 32M, after all previously people only take 64M and 8M, enlarge it on a step size of 128M only one time, it can't be seen as patient and careful. If it failed on 224M but succeed on 256M, then 256M may be not enough. I would like to say 32M is better, then we can make a good evaluate.

That makes sense. I also asked the customer to test intermediate values, we already know that it works with 256MB but also that 128MB are not enough. I will report back when I have the results of the intermediate values in 32MB steps.

Thanks,

Joerg

#13 lore raw

Re: [PATCH 3/3] x86, crash: Allocate enough low-mem when crashkernel=high

Hi Baoquan,

Quoted 4 lines · Joerg Roedel

That makes sense. I also asked the customer to test intermediate values, we already know that it works with 256MB but also that 128MB are not enough. I will report back when I have the results of the intermediate values in 32MB steps.

I got the results from the customer, and it turns out that a value of 192MB is sufficient to make the kdump succeed. It fails with 128MB and 160MB.

So I think we can settle in 192MB for now. What do you think?

Regards,

Joerg

#14 lore raw

Re: [PATCH 3/3] x86, crash: Allocate enough low-mem when crashkernel=high

Quoted 8 lines · Joerg Roedel

Hi Baoquan,

Quoted 4 lines · Joerg Roedel

That makes sense. I also asked the customer to test intermediate values, we already know that it works with 256MB but also that 128MB are not enough. I will report back when I have the results of the intermediate values in 32MB steps.

Quoted 5 lines

I got the results from the customer, and it turns out that a value of 192MB is sufficient to make the kdump succeed. It fails with 128MB and 160MB.

So I think we can settle in 192MB for now. What do you think?

Hi Joerg,

Sorry for late reply.

So that machine need eat memory between 160M and 192M. Then how about setting it as 256M? Since 192M seems very close to the brink, and setting it larger can keep enough space to extend for increasing dma device on larger machines, otherwise this value could be increased soon.

Wasting some low memory is better than kdump kernel can't bootup caused by not enough low memory. If someone grudge the low mem they can specify crashkernel=size,low manually.

Conclusively, I like 256M since the testing data showed it's sufficient now and should be save for a long time.

Thanks
Baoquan

#15 lore raw

Re: [PATCH 3/3] x86, crash: Allocate enough low-mem when crashkernel=high

Hi Baoquan,

Quoted 2 lines · Baoquan He

Conclusively, I like 256M since the testing data showed it's sufficient now and should be save for a long time.

Thanks, I am fine with 256MB too, so can I have your Acked-by on this series? I will rebase and resend it then after the merge window in the hope it gets queued.

Joerg

#16 lore raw

Re: [PATCH 3/3] x86, crash: Allocate enough low-mem when crashkernel=high

Quoted 9 lines · Joerg Roedel

Hi Baoquan,

Quoted 2 lines · Baoquan He

Conclusively, I like 256M since the testing data showed it's sufficient now and should be save for a long time.

Thanks, I am fine with 256MB too, so can I have your Acked-by on this series? I will rebase and resend it then after the merge window in the hope it gets queued.

Sure. Have acked it in reply to cover letter, feel free to add my Acked-by in your resend.

Thanks a lot!

Thanks
Baoquan

Quoted 9 lines

Joerg

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

maintainer
#17 lore raw

Re: [PATCH 3/3] x86, crash: Allocate enough low-mem when crashkernel=high

In reply to Joerg Roedel
Quoted 8 lines · Joerg Roedel

From: Joerg Roedel <[email protected]>

When the crashkernel is loaded above 4GiB in memory the first kernel only allocates 72MiB of low-memory for the DMA requirements of the second kernel. On systems with many devices this is not enough and causes device driver initialization errors and failed crash dumps. Set this default value to 256MiB to make sure there is enough memory

This upper limit of 256 looks arbitrary. Are we going to raise it a couple of years from now if it becomes insufficient then?

It probably won't be easy but is there some more reliable way to allocate enough memory for DMA on a say per-system basis or whatever...? Probably not but let me ask it anyway.

Signature · 5 lines
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.
--
#18 lore raw

Re: [PATCH 3/3] x86, crash: Allocate enough low-mem when crashkernel=high

Quoted 10 lines · Borislav Petkov
Quoted 8 lines · Joerg Roedel

From: Joerg Roedel <[email protected]>

When the crashkernel is loaded above 4GiB in memory the first kernel only allocates 72MiB of low-memory for the DMA requirements of the second kernel. On systems with many devices this is not enough and causes device driver initialization errors and failed crash dumps. Set this default value to 256MiB to make sure there is enough memory

Quoted 2 lines

This upper limit of 256 looks arbitrary. Are we going to raise it a couple of years from now if it becomes insufficient then?

Yes, it is arbitrary. I am open for suggestions on what might be a proper value to satisfy most systems.

Quoted 3 lines

It probably won't be easy but is there some more reliable way to allocate enough memory for DMA on a say per-system basis or whatever...? Probably not but let me ask it anyway.

Well, there is no easy way. But we could collect information from the loaded drivers on boot about how many dma-memory they allocate and base our allocation on that. Or we solve it in user-space by some more cleverness in creating the kernel command-line for crashkernel=high.

But besides that, I think the first two patches of this set make sense anyway. I understand that the third one is debatable.

Joerg

maintainer
#19 lore raw

Re: [PATCH 3/3] x86, crash: Allocate enough low-mem when crashkernel=high

Quoted 3 lines · Joerg Roedel

Well, there is no easy way. But we could collect information from the loaded drivers on boot about how many dma-memory they allocate and base our allocation on that.

That sounds like a nifty idea to me.

Quoted 2 lines

Or we solve it in user-space by some more cleverness in creating the kernel command-line for crashkernel=high.

I'd say, we should try to do as much as possible automatically.

Quoted 2 lines

But besides that, I think the first two patches of this set make sense anyway. I understand that the third one is debatable.

Right, but since they fix a real problem, maybe we should take them now until a better fix is done? Yes, no?

Signature · 5 lines
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.
--
#20 lore raw

Re: [PATCH 3/3] x86, crash: Allocate enough low-mem when crashkernel=high

Quoted 5 lines · Borislav Petkov
Quoted 2 lines · Joerg Roedel

Or we solve it in user-space by some more cleverness in creating the kernel command-line for crashkernel=high.

I'd say, we should try to do as much as possible automatically.

Thats hard to do, without any information about the driver needs we only have the amount of devices in the system to base any heuristic on.

Quoted 5 lines
Quoted 2 lines

But besides that, I think the first two patches of this set make sense anyway. I understand that the third one is debatable.

Right, but since they fix a real problem, maybe we should take them now until a better fix is done? Yes, no?

Yes. Given that we have no data about what would work on most systems, we can only change the value to a number that fixes know problems and then act on possible regressions caused by the change (and that change is pretty easy to revert if needed).

Joerg

maintainer
#21 lore raw

Re: [PATCH 3/3] x86, crash: Allocate enough low-mem when crashkernel=high

Quoted 4 lines · Joerg Roedel

Yes. Given that we have no data about what would work on most systems, we can only change the value to a number that fixes know problems and then act on possible regressions caused by the change (and that change is pretty easy to revert if needed).

... and I think we can safely concentrate on kdump-specific usage anyway and de-prioritize the others. Like not adding in the radeon module to initramfs and such. I don't think we want to show a funky graphic while dumping the first kernel.

Signature · 5 lines
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.
--
#22 lore raw
In reply to Joerg Roedel

Ping.

Quoted 59 lines · Joerg Roedel

v1->v2:

* Updated comments based on feedback from Konrad
* Added Acked-bys
* Rebased to v3.19-rc3

Hi,

here is a patch-set to fix failed kdump kernel boots when the systems was booted with crashkernel=X,high. On those systems the kernel allocates only 72MiB of low-memory for DMA buffers, which showed to be too low on some systems.

The problem is that 64MiB of the low-memory is allocated by swiotlb, leaving 8MB for the page-allocator. But swiotlb tries to allocate DMA memory from the page-allocator first, which fails pretty fast in the boot sequence, causing warnings. This patch-set removes these warnings.

But even the 64MiB for swiotlb are eaten up on some systems, so that the default of low-memory allocated for the crash-kernel is increase from 72MB to 256MB (only changing the defaults, can still be overwritten by crashkernel=X,low).

This number comes from experiments on the affected systems, 128MiB low-memory was still not enough there, thus I set the value to 256MiB to fix the issues.

Any feedback appreciated.

Thanks,

Joerg

Joerg Roedel (3):
swiotlb: Warn on allocation failure in swiotlb_alloc_coherent x86, swiotlb: Try coherent allocations with __GFP_NOWARN x86, crash: Allocate enough low-mem when crashkernel=high

 arch/x86/kernel/pci-swiotlb.c |  8 ++++++++
 arch/x86/kernel/setup.c       |  5 ++++-
 lib/swiotlb.c                 | 11 +++++++++--
 3 files changed, 21 insertions(+), 3 deletions(-)
Signature · 14 lines
1.9.1

Joerg Roedel (3):
  swiotlb: Warn on allocation failure in swiotlb_alloc_coherent
  x86, swiotlb: Try coherent allocations with __GFP_NOWARN
  x86, crash: Allocate enough low-mem when crashkernel=high

 arch/x86/kernel/pci-swiotlb.c |  8 ++++++++
 arch/x86/kernel/setup.c       |  5 ++++-
 lib/swiotlb.c                 | 11 +++++++++--
 3 files changed, 21 insertions(+), 3 deletions(-)

-- 
1.9.1
maintainer
#23 lore raw
In reply to Joerg Roedel

Adding some more people to CC.

Quoted 67 lines · Joerg Roedel

v1->v2:

* Updated comments based on feedback from Konrad
* Added Acked-bys
* Rebased to v3.19-rc3

Hi,

here is a patch-set to fix failed kdump kernel boots when the systems was booted with crashkernel=X,high. On those systems the kernel allocates only 72MiB of low-memory for DMA buffers, which showed to be too low on some systems.

The problem is that 64MiB of the low-memory is allocated by swiotlb, leaving 8MB for the page-allocator. But swiotlb tries to allocate DMA memory from the page-allocator first, which fails pretty fast in the boot sequence, causing warnings. This patch-set removes these warnings.

But even the 64MiB for swiotlb are eaten up on some systems, so that the default of low-memory allocated for the crash-kernel is increase from 72MB to 256MB (only changing the defaults, can still be overwritten by crashkernel=X,low).

This number comes from experiments on the affected systems, 128MiB low-memory was still not enough there, thus I set the value to 256MiB to fix the issues.

Any feedback appreciated.

Thanks,

Joerg

Joerg Roedel (3):
swiotlb: Warn on allocation failure in swiotlb_alloc_coherent x86, swiotlb: Try coherent allocations with __GFP_NOWARN x86, crash: Allocate enough low-mem when crashkernel=high

 arch/x86/kernel/pci-swiotlb.c |  8 ++++++++
 arch/x86/kernel/setup.c       |  5 ++++-
 lib/swiotlb.c                 | 11 +++++++++--
 3 files changed, 21 insertions(+), 3 deletions(-)
Signature · 20 lines
1.9.1

Joerg Roedel (3):
  swiotlb: Warn on allocation failure in swiotlb_alloc_coherent
  x86, swiotlb: Try coherent allocations with __GFP_NOWARN
  x86, crash: Allocate enough low-mem when crashkernel=high

 arch/x86/kernel/pci-swiotlb.c |  8 ++++++++
 arch/x86/kernel/setup.c       |  5 ++++-
 lib/swiotlb.c                 | 11 +++++++++--
 3 files changed, 21 insertions(+), 3 deletions(-)

-- 
1.9.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
Signature · 5 lines
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.
--
Acked-by on the v2 series
#24 lore raw
In reply to Joerg Roedel

This patch is very helpful and necessary since several users complained about the failure caused by not enough low mem. And the default value 256M is suitable since the testing data showed it's sufficient now and should be save for a long time.

And it also makes sense to supress the warning from buddy allocation failure which will call dump_stack in x86_swiotlb_alloc_coherent. Since it tried buddy allocation several times before the final try of bounce buffer allocation.

So ack the whole patch set.

Hi Joerg,

Thanks for your effort on this issue.

Could you please also update the cover letter or patch log to tell how 256M comes out with the later test result? I think it is convincing and helpful to understand.

Thanks
Baoquan

Quoted 65 lines · Joerg Roedel

v1->v2:

* Updated comments based on feedback from Konrad
* Added Acked-bys
* Rebased to v3.19-rc3

Hi,

here is a patch-set to fix failed kdump kernel boots when the systems was booted with crashkernel=X,high. On those systems the kernel allocates only 72MiB of low-memory for DMA buffers, which showed to be too low on some systems.

The problem is that 64MiB of the low-memory is allocated by swiotlb, leaving 8MB for the page-allocator. But swiotlb tries to allocate DMA memory from the page-allocator first, which fails pretty fast in the boot sequence, causing warnings. This patch-set removes these warnings.

But even the 64MiB for swiotlb are eaten up on some systems, so that the default of low-memory allocated for the crash-kernel is increase from 72MB to 256MB (only changing the defaults, can still be overwritten by crashkernel=X,low).

This number comes from experiments on the affected systems, 128MiB low-memory was still not enough there, thus I set the value to 256MiB to fix the issues.

Any feedback appreciated.

Thanks,

Joerg

Joerg Roedel (3):
swiotlb: Warn on allocation failure in swiotlb_alloc_coherent x86, swiotlb: Try coherent allocations with __GFP_NOWARN x86, crash: Allocate enough low-mem when crashkernel=high

 arch/x86/kernel/pci-swiotlb.c |  8 ++++++++
 arch/x86/kernel/setup.c       |  5 ++++-
 lib/swiotlb.c                 | 11 +++++++++--
 3 files changed, 21 insertions(+), 3 deletions(-)
Signature · 20 lines
1.9.1

Joerg Roedel (3):
  swiotlb: Warn on allocation failure in swiotlb_alloc_coherent
  x86, swiotlb: Try coherent allocations with __GFP_NOWARN
  x86, crash: Allocate enough low-mem when crashkernel=high

 arch/x86/kernel/pci-swiotlb.c |  8 ++++++++
 arch/x86/kernel/setup.c       |  5 ++++-
 lib/swiotlb.c                 | 11 +++++++++--
 3 files changed, 21 insertions(+), 3 deletions(-)

-- 
1.9.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
#25 lore raw

Hi Baoquan,

Quoted 5 lines · Baoquan He

This patch is very helpful and necessary since several users complained about the failure caused by not enough low mem. And the default value 256M is suitable since the testing data showed it's sufficient now and should be save for a long time.

Quoted 8 lines

And it also makes sense to supress the warning from buddy allocation failure which will call dump_stack in x86_swiotlb_alloc_coherent. Since it tried buddy allocation several times before the final try of bounce buffer allocation.

So ack the whole patch set.

Thanks a lot!

Quoted 7 lines

Hi Joerg,

Thanks for your effort on this issue.

Could you please also update the cover letter or patch log to tell how 256M comes out with the later test result? I think it is convincing and helpful to understand.

Sure thing, will update the patch description before I resend the series.

Regards,

Joerg

#26 lore raw
Quoted 21 lines · Joerg Roedel

Hi Baoquan,

Quoted 5 lines · Baoquan He

This patch is very helpful and necessary since several users complained about the failure caused by not enough low mem. And the default value 256M is suitable since the testing data showed it's sufficient now and should be save for a long time.

Quoted 8 lines

And it also makes sense to supress the warning from buddy allocation failure which will call dump_stack in x86_swiotlb_alloc_coherent. Since it tried buddy allocation several times before the final try of bounce buffer allocation.

So ack the whole patch set.

Thanks a lot!

Quoted 2 lines

Hi Joerg,

Quoted 8 lines
Quoted 5 lines

Thanks for your effort on this issue.

Could you please also update the cover letter or patch log to tell how 256M comes out with the later test result? I think it is convincing and helpful to understand.

Sure thing, will update the patch description before I resend the series.

Hi Joerg,

Ping!

About this patchset, what's your plan?

Thanks
Baoquan

#27 lore raw

Hi Baoquan

Quoted 3 lines · Baoquan He

Ping!

About this patchset, what's your plan?

Sorry, this went out of my scope. Thanks for reminding me, I'll send a new version soon.

Joerg