IB: rework memlock limit handling code
PATCH v2 1/1 IB: rework memlock limit handling code
This patch provides the uniform handling for RLIM_INFINITY value across the infiniband/rdma subsystem.
Currently in some cases the infinity constant is treated as an actual limit value, which could be misleading.
Let's also provide the single helper to check against process MEMLOCK limit while registering user memory region mappings.
---
v1 -> v2: rewritten commit message, rebased on recent upstream
drivers/infiniband/core/umem.c | 7 ++----- drivers/infiniband/hw/qib/qib_user_pages.c | 7 +++---- drivers/infiniband/hw/usnic/usnic_uiom.c | 6 ++---- drivers/infiniband/sw/siw/siw_mem.c | 6 +++--- drivers/infiniband/sw/siw/siw_verbs.c | 23 ++++++++++------------ include/rdma/ib_umem.h | 11 +++++++++++ 6 files changed, 31 insertions(+), 29 deletions(-)
drivers/infiniband/core/umem.c +2 -5
@@ -35,12 +35,12 @@#include <linux/mm.h> #include <linux/dma-mapping.h> -#include <linux/sched/signal.h> #include <linux/sched/mm.h> #include <linux/export.h> #include <linux/slab.h> #include <linux/pagemap.h> #include <linux/count_zeros.h> +#include <rdma/ib_umem.h> #include <rdma/ib_umem_odp.h> #include "uverbs.h"
@@ -150,7 +150,6 @@ struct ib_umem *ib_umem_get(struct ib_device *device, unsigned long addr, {struct ib_umem *umem;struct page **page_list; -unsigned long lock_limit;unsigned long new_pinned;unsigned long cur_base;unsigned long dma_attr =0 ;
@@ -200,10 +199,8 @@ struct ib_umem *ib_umem_get(struct ib_device *device, unsigned long addr, goto out; } - lock_limit =rlimit (RLIMIT_MEMLOCK ) >>PAGE_SHIFT ; - new_pinned =atomic64_add_return (npages, &mm->pinned_vm); -if (new_pinned > lock_limit && !capable (CAP_IPC_LOCK )) { +if (!ib_umem_check_rlimit_memlock (new_pinned)) {atomic64_sub (npages, &mm->pinned_vm); ret = -ENOMEM ; goto out;
drivers/infiniband/hw/qib/qib_user_pages.c +3 -4
@@ -32,8 +32,8 @@*/ #include <linux/mm.h> -#include <linux/sched/signal.h> #include <linux/device.h> +#include <rdma/ib_umem.h> #include "qib.h"
@@ -94,14 +94,13 @@ int qib_map_page(struct pci_dev *hwdev, struct page *page, dma_addr_t *daddr)int qib_get_user_pages (unsigned long start_page,size_t num_pages,struct page **p) { -unsigned long locked, lock_limit; +unsigned long locked;size_t got;int ret; - lock_limit =rlimit (RLIMIT_MEMLOCK ) >>PAGE_SHIFT ; locked =atomic64_add_return (num_pages, ¤t->mm->pinned_vm); -if (locked > lock_limit && !capable (CAP_IPC_LOCK )) { +if (!ib_umem_check_rlimit_memlock (locked)) { ret = -ENOMEM ; goto bail; }
drivers/infiniband/hw/usnic/usnic_uiom.c +2 -4
@@ -34,13 +34,13 @@#include <linux/mm.h> #include <linux/dma-mapping.h> -#include <linux/sched/signal.h> #include <linux/sched/mm.h> #include <linux/hugetlb.h> #include <linux/iommu.h> #include <linux/workqueue.h> #include <linux/list.h> #include <rdma/ib_verbs.h> +#include <rdma/ib_umem.h> #include "usnic_log.h" #include "usnic_uiom.h"
@@ -90,7 +90,6 @@ static int usnic_uiom_get_pages(unsigned long addr, size_t size, int writable,struct scatterlist *sg;struct usnic_uiom_chunk *chunk;unsigned long locked; -unsigned long lock_limit;unsigned long cur_base;unsigned long npages;int ret;
@@ -124,9 +123,8 @@ static int usnic_uiom_get_pages(unsigned long addr, size_t size, int writable,mmap_read_lock (mm); locked =atomic64_add_return (npages, ¤t->mm->pinned_vm); - lock_limit =rlimit (RLIMIT_MEMLOCK ) >>PAGE_SHIFT ; -if ((locked > lock_limit) && !capable (CAP_IPC_LOCK )) { +if (!ib_umem_check_rlimit_memlock (locked)) { ret = -ENOMEM ; goto out; }
drivers/infiniband/sw/siw/siw_mem.c +3 -3
@@ -5,6 +5,7 @@#include <linux/gfp.h> #include <rdma/ib_verbs.h> +#include <rdma/ib_umem.h> #include <linux/dma-mapping.h> #include <linux/slab.h> #include <linux/sched/mm.h>
@@ -367,7 +368,6 @@ struct siw_umem *siw_umem_get(u64 start, u64 len, bool writable)struct siw_umem *umem;struct mm_struct *mm_s;u64 first_page_va; -unsigned long mlock_limit;unsigned int foll_flags =FOLL_LONGTERM ;int num_pages, num_chunks, i, rv =0 ;
@@ -396,9 +396,9 @@ struct siw_umem *siw_umem_get(u64 start, u64 len, bool writable)mmap_read_lock (mm_s); - mlock_limit =rlimit (RLIMIT_MEMLOCK ) >>PAGE_SHIFT ; -if (atomic64_add_return (num_pages, &mm_s->pinned_vm) > mlock_limit) { +if (!ib_umem_check_rlimit_memlock ( +atomic64_add_return (num_pages, &mm_s->pinned_vm))) { rv = -ENOMEM ; goto out_sem_up; }
drivers/infiniband/sw/siw/siw_verbs.c +10 -13
@@ -12,6 +12,7 @@#include <rdma/iw_cm.h> #include <rdma/ib_verbs.h> +#include <rdma/ib_umem.h> #include <rdma/ib_user_verbs.h> #include <rdma/uverbs_ioctl.h>
@@ -1321,8 +1322,8 @@ struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len,struct siw_umem *umem =NULL ;struct siw_ureq_reg_mr ureq;struct siw_device *sdev =to_siw_dev (pd->device); - -unsigned long mem_limit =rlimit (RLIMIT_MEMLOCK ); +unsigned long num_pages = + (PAGE_ALIGN (len + (start & ~PAGE_MASK ))) >>PAGE_SHIFT ;int rv; siw_dbg_pd(pd, "start: 0x%pK, va: 0x%pK, len: %llu\n",
@@ -1338,19 +1339,15 @@ struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len, rv = -EINVAL ; goto err_out; } -if (mem_limit !=RLIM_INFINITY ) { -unsigned long num_pages = - (PAGE_ALIGN (len + (start & ~PAGE_MASK ))) >>PAGE_SHIFT ; - mem_limit >>=PAGE_SHIFT ; -if (num_pages > mem_limit - current->mm->locked_vm) { -siw_dbg_pd (pd,"pages req %lu, max %lu, lock %lu\n" , - num_pages, mem_limit, - current->mm->locked_vm); - rv = -ENOMEM ; - goto err_out; - } +if (!ib_umem_check_rlimit_memlock (num_pages + current->mm->locked_vm)) { +siw_dbg_pd (pd,"pages req %lu, max %lu, lock %lu\n" , + num_pages,rlimit (RLIMIT_MEMLOCK ), + current->mm->locked_vm); + rv = -ENOMEM ; + goto err_out; } + umem =siw_umem_get (start, len,ib_access_writable (rights));if (IS_ERR (umem)) { rv =PTR_ERR (umem);
include/rdma/ib_umem.h +11 -0
@@ -11,6 +11,7 @@#include <linux/scatterlist.h> #include <linux/workqueue.h> #include <rdma/ib_verbs.h> +#include <linux/sched/signal.h> struct ib_ucontext ;struct ib_umem_odp ;
@@ -71,6 +72,16 @@ static inline size_t ib_umem_num_pages(struct ib_umem *umem)return ib_umem_num_dma_blocks (umem,PAGE_SIZE ); } +static inline bool ib_umem_check_rlimit_memlock (unsigned long value) +{ +unsigned long lock_limit =rlimit (RLIMIT_MEMLOCK ); + +if (lock_limit ==RLIM_INFINITY ||capable (CAP_IPC_LOCK )) +return true; + +return value <=PFN_DOWN (lock_limit); +} +static inline void __rdma_umem_block_iter_start (struct ib_block_iter *biter,struct ib_umem *umem,unsigned long pgsz)
Signature · 1 line
2.39.3
Quoted 21 lines · Maxim Samoylov
This patch provides the uniform handling for
RLIM_INFINITYvalue across the infiniband/rdma subsystem.Currently in some cases the infinity constant is treated as an actual limit value, which could be misleading.
Let's also provide the single helper to check against process MEMLOCK limit while registering user memory region mappings.
Signed-off-by Maxim Samoylov <[email protected]>---
v1 -> v2: rewritten commit message, rebased on recent upstream
drivers/infiniband/core/umem.c | 7 ++----- drivers/infiniband/hw/qib/qib_user_pages.c | 7 +++---- drivers/infiniband/hw/usnic/usnic_uiom.c | 6 ++---- drivers/infiniband/sw/siw/siw_mem.c | 6 +++--- drivers/infiniband/sw/siw/siw_verbs.c | 23 ++++++++++------------ include/rdma/ib_umem.h | 11 +++++++++++ 6 files changed, 31 insertions(+), 29 deletions(-)
<...>
Quoted 24 lines
drivers/infiniband/sw/siw/siw_verbs.c +2
-10@@ -1321,8 +1322,8 @@ struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len,struct siw_umem *umem =NULL ;struct siw_ureq_reg_mr ureq;struct siw_device *sdev =to_siw_dev (pd->device); - -unsigned long mem_limit =rlimit (RLIMIT_MEMLOCK ); +unsigned long num_pages = + (PAGE_ALIGN (len + (start & ~PAGE_MASK ))) >>PAGE_SHIFT ;int rv; siw_dbg_pd(pd, "start: 0x%pK, va: 0x%pK, len: %llu\n",@@ -1338,19 +1339,15 @@ struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len, rv = -EINVAL ; goto err_out; } -if (mem_limit !=RLIM_INFINITY ) { -unsigned long num_pages = - (PAGE_ALIGN (len + (start & ~PAGE_MASK ))) >>PAGE_SHIFT ; - mem_limit >>=PAGE_SHIFT ; -if (num_pages > mem_limit - current->mm->locked_vm) { -siw_dbg_pd (pd,"pages req %lu, max %lu, lock %lu\n" , - num_pages, mem_limit, - current->mm->locked_vm);
drivers/infiniband/sw/siw/siw_verbs.c struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len, · line 1349
Quoted 10 lines
drivers/infiniband/sw/siw/siw_verbs.c +6
-3- rv = -ENOMEM ; - goto err_out; - } +if (!ib_umem_check_rlimit_memlock (num_pages + current->mm->locked_vm)) { +siw_dbg_pd (pd,"pages req %lu, max %lu, lock %lu\n" , + num_pages,rlimit (RLIMIT_MEMLOCK ), + current->mm->locked_vm); + rv = -ENOMEM ; + goto err_out; }
Sorry for late response, but why does this hunk exist in first place?
drivers/infiniband/sw/siw/siw_verbs.c struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len, · line 1351
Quoted 2 lines
drivers/infiniband/sw/siw/siw_verbs.c +1
-0+ umem =siw_umem_get (start, len,ib_access_writable (rights));
This should be ib_umem_get().
Thanks
Quoted 60 lines · Leon Romanovsky
Quoted 21 lines · Maxim Samoylov
This patch provides the uniform handling for
RLIM_INFINITYvalue across the infiniband/rdma subsystem.Currently in some cases the infinity constant is treated as an actual limit value, which could be misleading.
Let's also provide the single helper to check against process MEMLOCK limit while registering user memory region mappings.
Signed-off-by Maxim Samoylov<[email protected]>---
v1 -> v2: rewritten commit message, rebased on recent upstream
drivers/infiniband/core/umem.c | 7 ++----- drivers/infiniband/hw/qib/qib_user_pages.c | 7 +++---- drivers/infiniband/hw/usnic/usnic_uiom.c | 6 ++---- drivers/infiniband/sw/siw/siw_mem.c | 6 +++--- drivers/infiniband/sw/siw/siw_verbs.c | 23 ++++++++++------------ include/rdma/ib_umem.h | 11 +++++++++++ 6 files changed, 31 insertions(+), 29 deletions(-)<...>
Quoted 24 lines
drivers/infiniband/sw/siw/siw_verbs.c +2
-10@@ -1321,8 +1322,8 @@ struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len,struct siw_umem *umem =NULL ;struct siw_ureq_reg_mr ureq;struct siw_device *sdev =to_siw_dev (pd->device); - -unsigned long mem_limit =rlimit (RLIMIT_MEMLOCK ); +unsigned long num_pages = + (PAGE_ALIGN (len + (start & ~PAGE_MASK ))) >>PAGE_SHIFT ;int rv; siw_dbg_pd(pd, "start: 0x%pK, va: 0x%pK, len: %llu\n",@@ -1338,19 +1339,15 @@ struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len, rv = -EINVAL ; goto err_out; } -if (mem_limit !=RLIM_INFINITY ) { -unsigned long num_pages = - (PAGE_ALIGN (len + (start & ~PAGE_MASK ))) >>PAGE_SHIFT ; - mem_limit >>=PAGE_SHIFT ; -if (num_pages > mem_limit - current->mm->locked_vm) { -siw_dbg_pd (pd,"pages req %lu, max %lu, lock %lu\n" , - num_pages, mem_limit, - current->mm->locked_vm);drivers/infiniband/sw/siw/siw_verbs.c struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len, · line 1349
Quoted 10 lines
drivers/infiniband/sw/siw/siw_verbs.c +6
-3- rv = -ENOMEM ; - goto err_out; - } +if (!ib_umem_check_rlimit_memlock (num_pages + current->mm->locked_vm)) { +siw_dbg_pd (pd,"pages req %lu, max %lu, lock %lu\n" , + num_pages,rlimit (RLIMIT_MEMLOCK ), + current->mm->locked_vm); + rv = -ENOMEM ; + goto err_out; }Sorry for late response, but why does this hunk exist in first place?
Quoted 3 lines
drivers/infiniband/sw/siw/siw_verbs.c struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len, · line 1351
Quoted 2 lines
drivers/infiniband/sw/siw/siw_verbs.c +1
-0+ umem =siw_umem_get (start, len,ib_access_writable (rights));This should be
ib_umem_get().
IMO, it deserves a separate patch, and replace siw_umem_get with ib_umem_get is not straightforward given siw_mem has two types of memory (pbl and umem).
Thanks,
Guoqing
Quoted 61 lines · Guoqing Jiang
Quoted 58 lines · Leon Romanovsky
Quoted 21 lines · Maxim Samoylov
This patch provides the uniform handling for
RLIM_INFINITYvalue across the infiniband/rdma subsystem.Currently in some cases the infinity constant is treated as an actual limit value, which could be misleading.
Let's also provide the single helper to check against process MEMLOCK limit while registering user memory region mappings.
Signed-off-by Maxim Samoylov<[email protected]>---
v1 -> v2: rewritten commit message, rebased on recent upstream
drivers/infiniband/core/umem.c | 7 ++----- drivers/infiniband/hw/qib/qib_user_pages.c | 7 +++---- drivers/infiniband/hw/usnic/usnic_uiom.c | 6 ++---- drivers/infiniband/sw/siw/siw_mem.c | 6 +++--- drivers/infiniband/sw/siw/siw_verbs.c | 23 ++++++++++------------ include/rdma/ib_umem.h | 11 +++++++++++ 6 files changed, 31 insertions(+), 29 deletions(-)<...>
Quoted 22 lines
drivers/infiniband/sw/siw/siw_verbs.c +2
-10@@ -1321,8 +1322,8 @@ struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len,struct siw_umem *umem =NULL ;struct siw_ureq_reg_mr ureq;struct siw_device *sdev =to_siw_dev (pd->device); - -unsigned long mem_limit =rlimit (RLIMIT_MEMLOCK ); +unsigned long num_pages = + (PAGE_ALIGN (len + (start & ~PAGE_MASK ))) >>PAGE_SHIFT ;int rv; siw_dbg_pd(pd, "start: 0x%pK, va: 0x%pK, len: %llu\n",@@ -1338,19 +1339,15 @@ struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len, rv = -EINVAL ; goto err_out; } -if (mem_limit !=RLIM_INFINITY ) { -unsigned long num_pages = - (PAGE_ALIGN (len + (start & ~PAGE_MASK ))) >>PAGE_SHIFT ; - mem_limit >>=PAGE_SHIFT ; -if (num_pages > mem_limit - current->mm->locked_vm) { -siw_dbg_pd (pd,"pages req %lu, max %lu, lock %lu\n" , - num_pages, mem_limit, - current->mm->locked_vm);drivers/infiniband/sw/siw/siw_verbs.c struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len, · line 1348
Quoted 10 lines
drivers/infiniband/sw/siw/siw_verbs.c +6
-3- rv = -ENOMEM ; - goto err_out; - } +if (!ib_umem_check_rlimit_memlock (num_pages + current->mm->locked_vm)) { +siw_dbg_pd (pd,"pages req %lu, max %lu, lock %lu\n" , + num_pages,rlimit (RLIMIT_MEMLOCK ), + current->mm->locked_vm); + rv = -ENOMEM ; + goto err_out; }Sorry for late response, but why does this hunk exist in first place?
Quoted 6 lines
Quoted 3 lines
drivers/infiniband/sw/siw/siw_verbs.c struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len, · line 1351
Quoted 2 lines
drivers/infiniband/sw/siw/siw_verbs.c +1
-0+ umem =siw_umem_get (start, len,ib_access_writable (rights));This should be
ib_umem_get().IMO, it deserves a separate patch, and replace
siw_umem_getwithib_umem_getis not straightforward givensiw_memhas two types of memory (pbl and umem).
The thing is that once you convince yourself that SIW should use ib_umem_get(), the same question will arise for other parts of this patch where ib_umem_check_rlimit_memlock() is used.
And if we eliminate them all, there won't be a need for this new API call at all.
Thanks
Quoted 3 lines
Thanks,
Guoqing
Quoted 54 lines · Leon Romanovsky
Quoted 53 lines · Guoqing Jiang
Quoted 50 lines · Leon Romanovsky
Quoted 21 lines · Maxim Samoylov
This patch provides the uniform handling for
RLIM_INFINITYvalue across the infiniband/rdma subsystem.Currently in some cases the infinity constant is treated as an actual limit value, which could be misleading.
Let's also provide the single helper to check against process MEMLOCK limit while registering user memory region mappings.
Signed-off-by Maxim Samoylov<[email protected]>---
v1 -> v2: rewritten commit message, rebased on recent upstream
drivers/infiniband/core/umem.c | 7 ++----- drivers/infiniband/hw/qib/qib_user_pages.c | 7 +++---- drivers/infiniband/hw/usnic/usnic_uiom.c | 6 ++---- drivers/infiniband/sw/siw/siw_mem.c | 6 +++--- drivers/infiniband/sw/siw/siw_verbs.c | 23 ++++++++++------------ include/rdma/ib_umem.h | 11 +++++++++++ 6 files changed, 31 insertions(+), 29 deletions(-)<...>
Quoted 26 lines
drivers/infiniband/sw/siw/siw_verbs.c +3
-13@@ -1321,8 +1322,8 @@ struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len,struct siw_umem *umem =NULL ;struct siw_ureq_reg_mr ureq;struct siw_device *sdev =to_siw_dev (pd->device); - -unsigned long mem_limit =rlimit (RLIMIT_MEMLOCK ); +unsigned long num_pages = + (PAGE_ALIGN (len + (start & ~PAGE_MASK ))) >>PAGE_SHIFT ;int rv; siw_dbg_pd(pd, "start: 0x%pK, va: 0x%pK, len: %llu\n",@@ -1338,19 +1339,15 @@ struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len, rv = -EINVAL ; goto err_out; } -if (mem_limit !=RLIM_INFINITY ) { -unsigned long num_pages = - (PAGE_ALIGN (len + (start & ~PAGE_MASK ))) >>PAGE_SHIFT ; - mem_limit >>=PAGE_SHIFT ; -if (num_pages > mem_limit - current->mm->locked_vm) { -siw_dbg_pd (pd,"pages req %lu, max %lu, lock %lu\n" , - num_pages, mem_limit, - current->mm->locked_vm); - rv = -ENOMEM ; - goto err_out; - } +if (!ib_umem_check_rlimit_memlock (num_pages + current->mm->locked_vm)) {
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
drivers/infiniband/sw/siw/siw_verbs.c struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len, · line 1349
Quoted 6 lines
drivers/infiniband/sw/siw/siw_verbs.c +5
-0+siw_dbg_pd (pd,"pages req %lu, max %lu, lock %lu\n" , + num_pages,rlimit (RLIMIT_MEMLOCK ), + current->mm->locked_vm); + rv = -ENOMEM ; + goto err_out; }Sorry for late response, but why does this hunk exist in first place?
Trailing newline, will definitely drop it.
Quoted 7 lines
Quoted 6 lines
Quoted 3 lines
drivers/infiniband/sw/siw/siw_verbs.c struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len, · line 1351
Quoted 2 lines
drivers/infiniband/sw/siw/siw_verbs.c +1
-0+ umem =siw_umem_get (start, len,ib_access_writable (rights));This should be
ib_umem_get().IMO, it deserves a separate patch, and replace
siw_umem_getwithib_umem_getis not straightforward givensiw_memhas two types of memory (pbl and umem).
Quoted 8 lines
The thing is that once you convince yourself that SIW should use
ib_umem_get(), the same question will arise for other parts of this patch whereib_umem_check_rlimit_memlock()is used.And if we eliminate them all, there won't be a need for this new API call at all.
Thanks
Hi!
So, as for 31.10.2023 I still see siw_umem_get() call used in linux-rdma repo in "for-next" branch.
AFAIU this helper call is used only in a single place and could potentially be replaced with ib_umem_get() as Leon suggests.
But should we perform it right inside this memlock helper patch?
I can submit later another patch with siw_umem_get() replaced if necessary.
Quoted 3 lines
Quoted 3 lines
Thanks,
Guoqing
Quoted 79 lines · Maxim Samoylov
Quoted 54 lines · Leon Romanovsky
Quoted 53 lines · Guoqing Jiang
Quoted 50 lines · Leon Romanovsky
Quoted 21 lines · Maxim Samoylov
This patch provides the uniform handling for
RLIM_INFINITYvalue across the infiniband/rdma subsystem.Currently in some cases the infinity constant is treated as an actual limit value, which could be misleading.
Let's also provide the single helper to check against process MEMLOCK limit while registering user memory region mappings.
Signed-off-by Maxim Samoylov<[email protected]>---
v1 -> v2: rewritten commit message, rebased on recent upstream
drivers/infiniband/core/umem.c | 7 ++----- drivers/infiniband/hw/qib/qib_user_pages.c | 7 +++---- drivers/infiniband/hw/usnic/usnic_uiom.c | 6 ++---- drivers/infiniband/sw/siw/siw_mem.c | 6 +++--- drivers/infiniband/sw/siw/siw_verbs.c | 23 ++++++++++------------ include/rdma/ib_umem.h | 11 +++++++++++ 6 files changed, 31 insertions(+), 29 deletions(-)<...>
Quoted 26 lines
drivers/infiniband/sw/siw/siw_verbs.c +3
-13@@ -1321,8 +1322,8 @@ struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len,struct siw_umem *umem =NULL ;struct siw_ureq_reg_mr ureq;struct siw_device *sdev =to_siw_dev (pd->device); - -unsigned long mem_limit =rlimit (RLIMIT_MEMLOCK ); +unsigned long num_pages = + (PAGE_ALIGN (len + (start & ~PAGE_MASK ))) >>PAGE_SHIFT ;int rv; siw_dbg_pd(pd, "start: 0x%pK, va: 0x%pK, len: %llu\n",@@ -1338,19 +1339,15 @@ struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len, rv = -EINVAL ; goto err_out; } -if (mem_limit !=RLIM_INFINITY ) { -unsigned long num_pages = - (PAGE_ALIGN (len + (start & ~PAGE_MASK ))) >>PAGE_SHIFT ; - mem_limit >>=PAGE_SHIFT ; -if (num_pages > mem_limit - current->mm->locked_vm) { -siw_dbg_pd (pd,"pages req %lu, max %lu, lock %lu\n" , - num_pages, mem_limit, - current->mm->locked_vm); - rv = -ENOMEM ; - goto err_out; - } +if (!ib_umem_check_rlimit_memlock (num_pages + current->mm->locked_vm)) {Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
drivers/infiniband/sw/siw/siw_verbs.c struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len, · line 1349
Quoted 6 lines
drivers/infiniband/sw/siw/siw_verbs.c +5
-0+siw_dbg_pd (pd,"pages req %lu, max %lu, lock %lu\n" , + num_pages,rlimit (RLIMIT_MEMLOCK ), + current->mm->locked_vm); + rv = -ENOMEM ; + goto err_out; }Sorry for late response, but why does this hunk exist in first place?
Trailing newline, will definitely drop it.
Quoted 13 lines
Quoted 6 lines
Quoted 3 lines
drivers/infiniband/sw/siw/siw_verbs.c struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len, · line 1351
Quoted 2 lines
drivers/infiniband/sw/siw/siw_verbs.c +1
-0+ umem =siw_umem_get (start, len,ib_access_writable (rights));This should be
ib_umem_get().IMO, it deserves a separate patch, and replace
siw_umem_getwithib_umem_getis not straightforward givensiw_memhas two types of memory (pbl and umem).The thing is that once you convince yourself that SIW should use
ib_umem_get(), the same question will arise for other parts of this patch whereib_umem_check_rlimit_memlock()is used.And if we eliminate them all, there won't be a need for this new API call at all.
Quoted 7 lines
Quoted 2 lines
Thanks
Hi!
So, as for 31.10.2023 I still see
siw_umem_get()call used in linux-rdma repo in "for-next" branch.
I hoped to hear some feedback from Bernard and Dennis.
Quoted 14 lines
AFAIU this helper call is used only in a single place and could potentially be replaced with
ib_umem_get()as Leon suggests.But should we perform it right inside this memlock helper patch?
I can submit later another patch with
siw_umem_get()replaced if necessary.Quoted 3 lines
Quoted 3 lines
Thanks,
Guoqing
Quoted 105 lines
-----Original Message-----
From: Leon Romanovsky <[email protected]>
Sent: Thursday, November 2, 2023 1:32 PM
To: Maxim Samoylov <[email protected]>; Bernard Metzler <[email protected]>; Dennis Dalessandro <[email protected]>
Cc: Guoqing Jiang <[email protected]>; [email protected]; Jason Gunthorpe <[email protected]>; Christian Benvenuti <[email protected]>; Vadim Fedorenko <[email protected]>
Subject: [EXTERNAL] Re: [PATCH v2] IB: rework memlock limit handling codeQuoted 18 lines · Maxim Samoylov
Quoted 17 lines · Leon Romanovsky
Quoted 16 lines · Guoqing Jiang
Quoted 13 lines · Leon Romanovsky
Quoted 12 lines · Maxim Samoylov
This patch provides the uniform handling for
RLIM_INFINITYvalue across the infiniband/rdma subsystem.Currently in some cases the infinity constant is treated as an actual limit value, which could be misleading.
Let's also provide the single helper to check against process MEMLOCK limit while registering user memory region mappings.
Signed-off-by Maxim Samoylov<[email protected]>---
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
v1 -> v2: rewritten commit message, rebased on recent upstream
drivers/infiniband/core/umem.c | 7 ++----- drivers/infiniband/hw/qib/qib_user_pages.c | 7 +++---- drivers/infiniband/hw/usnic/usnic_uiom.c | 6 ++---- drivers/infiniband/sw/siw/siw_mem.c | 6 +++--- drivers/infiniband/sw/siw/siw_verbs.c | 23 ++++++++++------------
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 2 lines
include/rdma/ib_umem.h | 11 +++++++++++ 6 files changed, 31 insertions(+), 29 deletions(-)<...>
Review struct ib_mr *siw_reg_user_mr(struct ib_pd
Quoted 1 line
(quoted diff) +0
-0@@ -1321,8 +1322,8 @@ struct ib_mr *siw_reg_user_mr(struct ib_pd*pd, u64 start, u64 len,
Quoted 10 lines
Quoted 10 lines
Quoted 10 lines
Quoted 10 lines
drivers/infiniband/sw/siw/siw_verbs.c struct ib_mr *siw_reg_user_mr(struct ib_pd
Quoted 10 lines
drivers/infiniband/sw/siw/siw_verbs.c +2
-2struct siw_umem *umem =NULL ;struct siw_ureq_reg_mr ureq;struct siw_device *sdev =to_siw_dev (pd->device); - -unsigned long mem_limit =rlimit (RLIMIT_MEMLOCK ); +unsigned long num_pages = + (PAGE_ALIGN (len + (start & ~PAGE_MASK ))) >>PAGE_SHIFT ;int rv; siw_dbg_pd(pd, "start: 0x%pK, va: 0x%pK, len: %llu\n",@@ -1338,19 +1339,15 @@ struct ib_mr *siw_reg_user_mr(struct ib_pd*pd, u64 start, u64 len,
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
drivers/infiniband/sw/siw/siw_verbs.c struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len,
Quoted 6 lines
drivers/infiniband/sw/siw/siw_verbs.c +0
-3rv = -EINVAL ; goto err_out; } -if (mem_limit !=RLIM_INFINITY ) { -unsigned long num_pages = - (PAGE_ALIGN(len + (start & ~PAGE_MASK))) >>
PAGE_SHIFT;Quoted 19 lines
Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
drivers/infiniband/sw/siw/siw_verbs.c struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len,
Quoted 9 lines
drivers/infiniband/sw/siw/siw_verbs.c +1
-8- mem_limit >>=PAGE_SHIFT ; -if (num_pages > mem_limit - current->mm->locked_vm) { -siw_dbg_pd (pd,"pages req %lu, max %lu, lock %lu\n" , - num_pages, mem_limit, - current->mm->locked_vm); - rv = -ENOMEM ; - goto err_out; - } + if (!ib_umem_check_rlimit_memlock(num_pages + current->mm-
locked_vm)) {Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
drivers/infiniband/sw/siw/siw_verbs.c struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len, · line 1349
Quoted 6 lines
drivers/infiniband/sw/siw/siw_verbs.c +5
-0+siw_dbg_pd (pd,"pages req %lu, max %lu, lock %lu\n" , + num_pages,rlimit (RLIMIT_MEMLOCK ), + current->mm->locked_vm); + rv = -ENOMEM ; + goto err_out; }Sorry for late response, but why does this hunk exist in first place?
Quoted 7 lines
Trailing newline, will definitely drop it.
Quoted 5 lines
Quoted 5 lines
Quoted 3 lines
drivers/infiniband/sw/siw/siw_verbs.c struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len, · line 1351
Quoted 2 lines
drivers/infiniband/sw/siw/siw_verbs.c +1
-0+ umem =siw_umem_get (start, len,ib_access_writable (rights));This should be
ib_umem_get().IMO, it deserves a separate patch, and replace
siw_umem_getwith
ib_umem_getQuoted 1 line
Quoted 1 line
Quoted 1 line
is not straightforward given
siw_memhas two types of memory (pbl andumem).
Quoted 2 lines
Quoted 2 lines
The thing is that once you convince yourself that SIW should use
ib_umem_get(),Quoted 4 lines
Quoted 4 lines
the same question will arise for other parts of this patch where
ib_umem_check_rlimit_memlock()is used.And if we eliminate them all, there won't be a need for this new API
call at all.
Quoted 4 lines
Quoted 3 lines
Thanks
Quoted 6 lines
Quoted 4 lines
Hi!
So, as for 31.10.2023 I still see
siw_umem_get()call used in linux-rdma repo in "for-next" branch.I hoped to hear some feedback from Bernard and Dennis.
Sorry for the late reply.
Not using ib_umem_get(), siw intentionally implements a simpler routine to get its user pages pinned and organized as a two dimensional list. It not only saves memory for the structure holding the pages, but also makes it very easy to resume sending or receiving data at any address without traversing ib_umem linearly.
What we could do is using ib_umem_get() to get the page list and use that list to organize pages more efficiently for a software rdma driver. I think this is what rxe implements today.
It comes with the penalty of extra data structures being allocated a software rdma driver does not care about.
Quoted 15 lines
Quoted 15 lines
AFAIU this helper call is used only in a single place and could potentially be replaced with
ib_umem_get()as Leon suggests.But should we perform it right inside this memlock helper patch?
I can submit later another patch with
siw_umem_get()replaced if necessary.Quoted 3 lines
Quoted 3 lines
Thanks,
Guoqing
Dennis Dalessandro
Acked-byQuoted 6 lines · Leon Romanovsky
Quoted 3 lines
So, as for 31.10.2023 I still see
siw_umem_get()call used in linux-rdma repo in "for-next" branch.I hoped to hear some feedback from Bernard and Dennis.
Sorry about that. I thought I did respond about qib.
Quoted 9 lines · Dennis Dalessandro
Quoted 6 lines · Leon Romanovsky
Quoted 3 lines
So, as for 31.10.2023 I still see
siw_umem_get()call used in linux-rdma repo in "for-next" branch.I hoped to hear some feedback from Bernard and Dennis.
Sorry about that. I thought I did respond about qib.
Dennis, qib probably needs to use ib_umem_get too.
Quoted 2 lines
Bernard Metzler
Acked-byQuoted 57 lines
-----Original Message-----
From: Leon Romanovsky <[email protected]>
Sent: Thursday, November 2, 2023 1:32 PM
To: Maxim Samoylov <[email protected]>; Bernard Metzler <[email protected]>; Dennis Dalessandro <[email protected]>
Cc: Guoqing Jiang <[email protected]>; [email protected]; Jason Gunthorpe <[email protected]>; Christian Benvenuti <[email protected]>; Vadim Fedorenko <[email protected]>
Subject: [EXTERNAL] Re: [PATCH v2] IB: rework memlock limit handling codeQuoted 18 lines · Maxim Samoylov
Quoted 17 lines · Leon Romanovsky
Quoted 16 lines · Guoqing Jiang
Quoted 13 lines · Leon Romanovsky
Quoted 12 lines · Maxim Samoylov
This patch provides the uniform handling for
RLIM_INFINITYvalue across the infiniband/rdma subsystem.Currently in some cases the infinity constant is treated as an actual limit value, which could be misleading.
Let's also provide the single helper to check against process MEMLOCK limit while registering user memory region mappings.
Signed-off-by Maxim Samoylov<[email protected]>---
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
v1 -> v2: rewritten commit message, rebased on recent upstream
drivers/infiniband/core/umem.c | 7 ++----- drivers/infiniband/hw/qib/qib_user_pages.c | 7 +++---- drivers/infiniband/hw/usnic/usnic_uiom.c | 6 ++---- drivers/infiniband/sw/siw/siw_mem.c | 6 +++--- drivers/infiniband/sw/siw/siw_verbs.c | 23 ++++++++++------------
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 2 lines
include/rdma/ib_umem.h | 11 +++++++++++ 6 files changed, 31 insertions(+), 29 deletions(-)<...>
Review struct ib_mr *siw_reg_user_mr(struct ib_pd
Quoted 1 line
(quoted diff) +0
-0@@ -1321,8 +1322,8 @@ struct ib_mr *siw_reg_user_mr(struct ib_pd*pd, u64 start, u64 len,
Quoted 10 lines
Quoted 10 lines
Quoted 10 lines
Quoted 10 lines
drivers/infiniband/sw/siw/siw_verbs.c struct ib_mr *siw_reg_user_mr(struct ib_pd
Quoted 10 lines
drivers/infiniband/sw/siw/siw_verbs.c +2
-2struct siw_umem *umem =NULL ;struct siw_ureq_reg_mr ureq;struct siw_device *sdev =to_siw_dev (pd->device); - -unsigned long mem_limit =rlimit (RLIMIT_MEMLOCK ); +unsigned long num_pages = + (PAGE_ALIGN (len + (start & ~PAGE_MASK ))) >>PAGE_SHIFT ;int rv; siw_dbg_pd(pd, "start: 0x%pK, va: 0x%pK, len: %llu\n",@@ -1338,19 +1339,15 @@ struct ib_mr *siw_reg_user_mr(struct ib_pd*pd, u64 start, u64 len,
Quoted 24 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
drivers/infiniband/sw/siw/siw_verbs.c struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len,
Quoted 6 lines
drivers/infiniband/sw/siw/siw_verbs.c +0
-3rv = -EINVAL ; goto err_out; } -if (mem_limit !=RLIM_INFINITY ) { -unsigned long num_pages = - (PAGE_ALIGN(len + (start & ~PAGE_MASK))) >>
PAGE_SHIFT;Quoted 17 lines
Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
drivers/infiniband/sw/siw/siw_verbs.c struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len,
Quoted 9 lines
drivers/infiniband/sw/siw/siw_verbs.c +1
-8- mem_limit >>=PAGE_SHIFT ; -if (num_pages > mem_limit - current->mm->locked_vm) { -siw_dbg_pd (pd,"pages req %lu, max %lu, lock %lu\n" , - num_pages, mem_limit, - current->mm->locked_vm); - rv = -ENOMEM ; - goto err_out; - } + if (!ib_umem_check_rlimit_memlock(num_pages + current->mm-
locked_vm)) {Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
drivers/infiniband/sw/siw/siw_verbs.c struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len, · line 1349
Quoted 6 lines
drivers/infiniband/sw/siw/siw_verbs.c +5
-0+siw_dbg_pd (pd,"pages req %lu, max %lu, lock %lu\n" , + num_pages,rlimit (RLIMIT_MEMLOCK ), + current->mm->locked_vm); + rv = -ENOMEM ; + goto err_out; }Sorry for late response, but why does this hunk exist in first place?
If using ib_umem_get() for siw, as I sent as for-next patch yesterday, we can drop that logic completely, since we now have it in ib_umem_get(). It was only there because of not using ib_umem_get().
I can resend my pending for-next patch as a patch to current, also removing memlock check (I simply forgot to remove it). Not sure if it would obsolete this patch here completely. Leon, please advise.
Otherwise:
Quoted 45 lines
Quoted 9 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
Trailing newline, will definitely drop it.
Quoted 5 lines
Quoted 5 lines
Quoted 3 lines
drivers/infiniband/sw/siw/siw_verbs.c struct ib_mr *siw_reg_user_mr(struct ib_pd *pd, u64 start, u64 len, · line 1351
Quoted 2 lines
drivers/infiniband/sw/siw/siw_verbs.c +1
-0+ umem =siw_umem_get (start, len,ib_access_writable (rights));This should be
ib_umem_get().IMO, it deserves a separate patch, and replace
siw_umem_getwith
ib_umem_getQuoted 1 line
Quoted 1 line
Quoted 1 line
is not straightforward given
siw_memhas two types of memory (pbl andumem).
Quoted 2 lines
Quoted 2 lines
The thing is that once you convince yourself that SIW should use
ib_umem_get(),Quoted 4 lines
Quoted 4 lines
the same question will arise for other parts of this patch where
ib_umem_check_rlimit_memlock()is used.And if we eliminate them all, there won't be a need for this new API
call at all.
Quoted 8 lines
Quoted 3 lines
Thanks
Hi!
So, as for 31.10.2023 I still see
siw_umem_get()call used in linux-rdma repo in "for-next" branch.I hoped to hear some feedback from Bernard and Dennis.
Quoted 14 lines
AFAIU this helper call is used only in a single place and could potentially be replaced with
ib_umem_get()as Leon suggests.But should we perform it right inside this memlock helper patch?
I can submit later another patch with
siw_umem_get()replaced if necessary.Quoted 3 lines
Quoted 3 lines
Thanks,
Guoqing
Quoted 88 lines · Bernard Metzler
Quoted 57 lines
-----Original Message-----
From: Leon Romanovsky <[email protected]>
Sent: Thursday, November 2, 2023 1:32 PM
To: Maxim Samoylov <[email protected]>; Bernard Metzler <[email protected]>; Dennis Dalessandro <[email protected]>
Cc: Guoqing Jiang <[email protected]>; [email protected]; Jason Gunthorpe <[email protected]>; Christian Benvenuti <[email protected]>; Vadim Fedorenko <[email protected]>
Subject: [EXTERNAL] Re: [PATCH v2] IB: rework memlock limit handling codeQuoted 18 lines · Maxim Samoylov
Quoted 17 lines · Leon Romanovsky
Quoted 16 lines · Guoqing Jiang
Quoted 13 lines · Leon Romanovsky
Quoted 12 lines · Maxim Samoylov
This patch provides the uniform handling for
RLIM_INFINITYvalue across the infiniband/rdma subsystem.Currently in some cases the infinity constant is treated as an actual limit value, which could be misleading.
Let's also provide the single helper to check against process MEMLOCK limit while registering user memory region mappings.
Signed-off-by Maxim Samoylov<[email protected]>---
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
v1 -> v2: rewritten commit message, rebased on recent upstream
drivers/infiniband/core/umem.c | 7 ++----- drivers/infiniband/hw/qib/qib_user_pages.c | 7 +++---- drivers/infiniband/hw/usnic/usnic_uiom.c | 6 ++---- drivers/infiniband/sw/siw/siw_mem.c | 6 +++--- drivers/infiniband/sw/siw/siw_verbs.c | 23 ++++++++++------------
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 2 lines
include/rdma/ib_umem.h | 11 +++++++++++ 6 files changed, 31 insertions(+), 29 deletions(-)<...>
Review struct ib_mr *siw_reg_user_mr(struct ib_pd
Quoted 1 line
(quoted diff) +0
-0@@ -1321,8 +1322,8 @@ struct ib_mr *siw_reg_user_mr(struct ib_pd*pd, u64 start, u64 len,
Quoted 10 lines
Quoted 10 lines
Quoted 10 lines
Quoted 10 lines
Review struct ib_mr *siw_reg_user_mr(struct ib_pd
Quoted 10 lines
(quoted diff) +2
-2struct siw_umem *umem = NULL; struct siw_ureq_reg_mr ureq; struct siw_device *sdev = to_siw_dev(pd->device); - - unsigned long mem_limit = rlimit(RLIMIT_MEMLOCK); + unsigned long num_pages = + (PAGE_ALIGN(len + (start & ~PAGE_MASK))) >> PAGE_SHIFT; int rv; siw_dbg_pd(pd, "start: 0x%pK, va: 0x%pK, len: %llu\n",@@ -1338,19 +1339,15 @@ struct ib_mr *siw_reg_user_mr(struct ib_pd*pd, u64 start, u64 len,
Quoted 24 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Review struct ib_mr *siw_reg_user_mr(struct ib_pd · line 1343
Quoted 6 lines
(quoted diff) +0
-3rv = -EINVAL; goto err_out; } - if (mem_limit != RLIM_INFINITY) { - unsigned long num_pages = - (PAGE_ALIGN(len + (start & ~PAGE_MASK))) >>
PAGE_SHIFT;Quoted 17 lines
Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Review struct ib_mr *siw_reg_user_mr(struct ib_pd · line 1342
Quoted 9 lines
(quoted diff) +1
-8- mem_limit >>= PAGE_SHIFT; - if (num_pages > mem_limit - current->mm->locked_vm) { - siw_dbg_pd(pd, "pages req %lu, max %lu, lock %lu\n", - num_pages, mem_limit, - current->mm->locked_vm); - rv = -ENOMEM; - goto err_out; - } + if (!ib_umem_check_rlimit_memlock(num_pages + current->mm-
locked_vm)) {Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Review struct ib_mr *siw_reg_user_mr(struct ib_pd · line 1348
Quoted 6 lines
(quoted diff) +5
-0+ siw_dbg_pd(pd, "pages req %lu, max %lu, lock %lu\n", + num_pages, rlimit(RLIMIT_MEMLOCK), + current->mm->locked_vm); + rv = -ENOMEM; + goto err_out; }Sorry for late response, but why does this hunk exist in first place?
If using
ib_umem_get()for siw, as I sent as for-next patch yesterday, we can drop that logic completely, since we now have it inib_umem_get(). It was only there because of not usingib_umem_get().
Quoted 4 lines
I can resend my pending for-next patch as a patch to current, also removing memlock check (I simply forgot to remove it). Not sure if it would obsolete this patch here completely. Leon, please advise.
We are in the middle of merge window, so won't take any patches except bug fixes.
So please, resend your patch after after merge window ends.
Thanks
Quoted 51 lines
Otherwise:
Acked-by Bernard Metzler <[email protected]>Quoted 45 lines
Quoted 9 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
Trailing newline, will definitely drop it.
Quoted 5 lines
Quoted 5 lines
Quoted 3 lines
Review struct ib_mr *siw_reg_user_mr(struct ib_pd · line 1350
Quoted 2 lines
(quoted diff) +1
-0+ umem = siw_umem_get(start, len, ib_access_writable(rights));This should be
ib_umem_get().IMO, it deserves a separate patch, and replace
siw_umem_getwith
ib_umem_getQuoted 1 line
Quoted 1 line
Quoted 1 line
is not straightforward given
siw_memhas two types of memory (pbl andumem).
Quoted 2 lines
Quoted 2 lines
The thing is that once you convince yourself that SIW should use
ib_umem_get(),Quoted 4 lines
Quoted 4 lines
the same question will arise for other parts of this patch where
ib_umem_check_rlimit_memlock()is used.And if we eliminate them all, there won't be a need for this new API
call at all.
Quoted 8 lines
Quoted 3 lines
Thanks
Hi!
So, as for 31.10.2023 I still see
siw_umem_get()call used in linux-rdma repo in "for-next" branch.I hoped to hear some feedback from Bernard and Dennis.
Quoted 14 lines
AFAIU this helper call is used only in a single place and could potentially be replaced with
ib_umem_get()as Leon suggests.But should we perform it right inside this memlock helper patch?
I can submit later another patch with
siw_umem_get()replaced if necessary.Quoted 3 lines
Quoted 3 lines
Thanks,
Guoqing